ALTITUDE AND MOOD DISTURBANCES: PERU By Manuel M. Catacora A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Epidemiology – Doctor of Philosophy 2024 ABSTRACT My research examines the association between the altitude of the participant's residence and the presence of an actively depressed mood, explicitly examining whether individuals experienced a depressed mood during the four weeks before an interview assessment during a population-based epidemiological survey. I explore this association at the community and individual levels. Additionally, I investigated whether using coca leaves might moderate the observed relationship between altitude and depressed mood. In this dissertation, my epidemiological estimates are from nine yearly cross-sectional epidemiological studies conducted in Peru. These studies encompassed 15 regions in Peru and included community residents in areas situated at different altitudes above sea level. The main results of my dissertation can be summarized as follows: Study 1: At the community level (i.e., district level), those populations located at higher altitudes also had higher estimated prevalence proportions of depressed mood. These estimates are from a comparison of direct estimates of survey data analysis, using standardized estimates (i.e., with direct adjustment for age and sex), and finally using estimators corresponding to the Small Area Estimation methodology (i.e., Fay-Herriot model). Study 2: Estimated at the individual level, an association between altitude and depressed mood was found and remained stable after logistic regression models adjusted for other covariates of interest (i.e., age, sex, mother tongue, marital status, and length of residency). By stratifying the analysis according to the length of stay in the current community, the findings suggested a causal possibility. The estimated effect of high altitude on a person's mood may be more pronounced after a minimum of two years of residing at such heights. Study 3: In assessing coca-leaf consumption as a moderator of the altitude-depression association, I found an absence of evidence that coca-leaf product use in the month before the survey modifies the relationship between altitude and depressed mood. In fact, if anything, individuals who use coca might be more likely to experience active depressive moods, even at higher altitudes. However, the study's cross-sectional nature prevents the evaluation of the temporal aspect of this association. As for limitations, the cross-sectional character of the data, with no specification of the age of onset of depressed mood relative to the time of residency at the altitude of the current dwelling, is a crucial limitation. Also, this study did not evaluate other community characteristics that might be covarying with the altitude-depression relationship. This new evidence on the altitude-depression association draws attention to a need for more research along these lines. The research might gain increased importance during the 21st century intervals of global warming and associated climate change. Given that an estimated 500 million individuals reside in high-altitude regions around the globe, the association that links altitude and the occurrence of depressed mood deserves continuing attention in our epidemiological research projects. Copyright by MANUEL M. CATACORA 2024 I dedicate this dissertation to Manuel and Aus, my beloved parents. v ACKNOWLEDGMENTS I wish to thank those who, without their professional or personal support, this dissertation would not have been possible. First, I want to express my immense gratitude to my mentor, Professor James C. Anthony. His dedicated, wise, and kind guidance was my most valuable gift during this academic journey. I cannot imagine a better mentor than him. I also want to thank my guidance committee members, Dr. David Barondess, Dr. German Alvarado, and Dr. Olga Vsevolozhskaya. As a result of their support, I have completed this dissertation. I will acknowledge fellowship and instructional materials support from the following research training grants awarded to my advisor by the National Institutes of Health National Institute on Drug Abuse and the Fogarty International Center (NIH, NIDA, FIC): R25DA030310, T32DA021129, and D43TW005819. Also, I want to acknowledge the scholarship for dissertation completion from the MSU Office of Vice President for Research and Innovation, the Vice President for Graduate Studies, and the College of Human Medicine. Additionally, I sincerely thank Dr. Javier Saavedra and the research team of the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health. I would like to extend my gratitude to Marina Piazza and all the T32/D43 pre- and post-doc fellows. I also must acknowledge my family. My wife, sons, daughter, and sister have been a source of strength that enabled me to complete this journey. Finally, all my infinite thanks to God since I owe everything to him. vi TABLE OF CONTENTS LIST OF ABBREVIATIONS ......................................................................................................................... viii CHAPTER 1: INTRODUCTION, SPECIFIC AIMS, AND HYPOTHESES ....................................................... 1 CHAPTER 2: OVERVIEW OF CONCEPTS AND BACKGROUND ................................................................ 6 CHAPTER 3: ESTIMATING THE ALTITUDE-DEPRESSED MOOD ASSOCIATION AT THE DISTRICT LEVEL .......................................................................................................................................................... 23 CHAPTER 4: ESTIMATING THE ALTITUDE-DEPRESSED MOOD ASSOCIATION AT THE INDIVIDUAL LEVEL .................................................................................................................................... 57 CHAPTER 5: THE ESTIMATED EFFECT OF COCA-LEAF USE IN THE ASSOCIATION BETWEEN ALTITUDE AND DEPRESSED MOOD ........................................................................................................ 76 CHAPTER 6: CONCLUSIONS AND FUTURE DIRECTIONS FOR RESEARCH ........................................... 94 REFERENCES ............................................................................................................................................ 103 APPENDIX A: DEPRESSED MOOD AND COCA USE QUESTIONS FROM EPIDEMIOLOGICAL SURVEYS OF THE PERUVIAN NATIONAL INSTITUTE OF MENTAL HEALTH “HONORIO DELGADO – HIDEYO NOGUCHI” (“NOGUCHI SURVEYS”) .................................................................. 114 APPENDIX B: CHAPTER 3 ....................................................................................................................... 118 APPENDIX C: CHAPTER 4 ....................................................................................................................... 143 APPENDIX D: CHAPTER 5 ....................................................................................................................... 147 vii LIST OF ABBREVIATIONS BC − Before Christ DM − Depressed Mood CLC − Coca Leaf Chewing SaO2 − Arterial Oxygen Saturation HAPE − High-altitude Pulmonary Edema AMS − Acute Mountain Sickness HACE − High-Altitude Cerebral Edema CMS − Chronic Mountain Sickness PH − Pulmonary Hypertension MDE − Major Depressive Episode MDD − Major Depressive Disorder DSM − Diagnostic and Statistical Manual of Mental Disorders ICD − International Classification of Diseases ATP − Adenosine Triphosphate REM − Rapid Eye Movement PHQ-9 − Patient Health Questionnaire-9 ENDES − Demographic and Family Health Survey US − United States HADS − Hospital Anxiety and Depression Scale WHO − World Health Organization SAE − Small Area Estimation viii IRB − Institutional Review Board CEPLAN − Peruvian National Center for Strategic Planning SRQ − Self Reporting Questionnaire MINI − Mini-International Neuropsychiatric Interview EBLUP − Empirical Best Linear Unbiased Prediction FH − Fay-Herriot SDG − Sustainable Development Goal INEI − Peruvian National Institute of Statistics and Informatics CEPLAN − Peruvian National Center for Strategic Planning HDI − Human Development Index TSL − Taylor Series Linearization GoF − Goodness of fit statistics GLM − Generalized Linear Models RCT − Randomized Clinical Trials DHS − Demographic and Health Surveys DIF − Differential Item Functioning MIMIC − Multiple Indicators Multiple Causes ix CHAPTER 1: INTRODUCTION, SPECIFIC AIMS, AND HYPOTHESES 1.1 Introduction Climate and the characteristics of the territory in which populations live have been believed to influence the characteristics and personalities of people since Hippocrates' time (5th century BC)(1). Among Peruvians, Hipólito Unanue, in 1806, reflected on the relationship between climate and mood, describing a melancholy mood as the dominant mood throughout the history of this country (2). An interesting fact about altitude and mental health occurred in 1842 when Johann J. Guggenbühl hypothesized that intellectually disabled children could be treated and cured at higher altitudes. Guggenbühl opened the Abendberg School 1,200 meters above sea level for children with intellectual disabilities (3). (Other studies on altitude are mentioned later in this report.) More recently, studies conducted in the 21st century suggest a potential association between the altitude of a residential area and the risk of experiencing depression (4,5) and suicide behavior (6–9). Most epidemiological studies on this topic have been conducted on populations in North America, specifically at elevations not exceeding 1500 meters above sea level. The Peruvian Andes region includes several cities and towns that reach altitudes of nearly 5,000 meters. Consequently, the Peruvian Andes provide an opportunity to investigate the potential relationship between altitude and depression in a broader context. Additionally, the Peruvian population traditionally consumes coca leaf, particularly in the Andean region (10,11). Chewing coca leaves allows the consumption of very small doses of 1 fourteen different alkaloids, cocaine being the most significant (12). For this reason, it is postulated that coca leaves might modulate the mood of those who consume them (13–15). In this context, the present research aims to analyze the relationship between altitude and the prevalence of depressed mood in the Peruvian population and evaluate the possible moderating effect of chewing coca leaf on this association. The secondary analysis of epidemiological survey databases from the "Honorio Delgado - Hideyo Noguchi" National Institute of Mental Health - "Noguchi Surveys" is conducted in this research. These 22 regional surveys include locations on the coast, highlands, and jungle. They were completed sequentially between 2003 and 2013. They were all implemented with a similar methodology that addresses sociodemographic variables, mental disorders, alcohol and other drug use (e.g., coca leaf consumption), intimate violence, and other mental health-related topics. In this dissertation, after describing relevant concepts and background in Chapter 2, I will describe three studies. First, in Chapter 3, a study at the community level of analysis will analyze the relationship between districts' altitude and the prevalence proportion of depressed mood (DM). In Chapter 4, the second study addresses the relationship between altitude of residence and other relevant covariates with the presence of DM at the individual level of analysis. In Chapter 5, I will explore the possible moderator effect of coca-leaf consumption in the relationship between altitude and DM, also at the individual level of analysis. Chapter 6 is offered as a description of the conclusions of this dissertation and future research that is needed. 2 1.2 Objective 1. Based on data from Peru's "Noguchi Surveys," this dissertation research project aims to contribute to new epidemiological evidence on the association between the altitude of residence and depressed mood and the possible moderator effect of coca leaf use on that suggested relationship. I will examine the association between altitude and mood at the community (district) and individual levels. 1.3 Specific Aims 1. Via my novel analyses of data from Peru's "Noguchi Surveys," I will estimate active depression prevalence proportions (depression at the time of the interview assessment) with the community district level as the unit of analysis, with hypoxia theory-based expectations that the proportions would be greater in communities at higher altitudes and smaller in communities at lower altitudes (e.g., by tertiles). After initial cross-tabular analyses, I will adjust for age and sex with stratified analyses and direct standardization methods. 2. Extending this initial work with data from Peru's “Noguchi Surveys” and focusing on individuals as the unit of analysis data rather than district-level data, I will estimate the relationship between altitude in the community district and the occurrence of active depression (at the time of the interview). According to hypoxia theory, there might be a lower asymptote for the prevalence proportion (or odds), an upward-turning slope, and then an upper asymptote. I will estimate this altitude-depression locational association using a generalized linear model (e.g., with a logit link), considering a possible need for 3 more than one slope, and with these covariates: participant's sex, age, and mother tongue. 3. Via my novel analyses of individual-level data from Peru's “Noguchi Surveys,” I will study whether the association between altitude and depression might be subject to modification by a participant's recent history of coca leaf chewing (CLC). Evidence on this question can be addressed with a model fit index approach, comparing model fit statistics before and after adding altitude*CLC product-term(s) to the model and with consideration of slope estimates. 1.4 Hypotheses Under Specific Aim number 1, the hypothesis can be stated in the null form: Depressed mood prevalence proportions (i.e., feeling sadness “always” or “almost always” in the last four weeks, as indicated by subjective report at the time of the interview assessment) at the community district level do not vary significantly among communities at different altitude levels (i.e., by deciles), before or after direct standardization methods to adjust for age and sex. The alternative hypothesis, with hypoxia theory-based expectations, is that the Depressed Mood prevalence proportion would be greater in communities at higher altitudes and lower in communities at lower altitudes. Under Specific Aim number 2, the hypothesis can be stated in the null form: Focusing on individual data, there is no association between altitude levels of residence and depressed mood at the time of the interview assessment and after adjusting for relevant 4 covariates (i.e., participant's sex, age, and mother tongue). The odds of having a depressed mood do not vary across different altitude levels (i.e., tertiles). The alternative hypothesis states that there is a relationship between altitude quantiles and depression (i.e., feels sad “always” or “almost always” in the last four weeks). According to hypoxia theory, the odds of having an active depressed mood could be higher at elevated altitudes of residence. Under Specific Aim number 3, the hypothesis can be stated in the null form: There is no evidence suggestive of a moderator effect of coca-leaf use (i.e., “last month's use") on the association between altitude and active depressed mood. The alternative hypothesis states that the association between altitude and depressed mood might be subject to modification by a participant's recent history of coca leaf chewing (CLC). 5 CHAPTER 2: OVERVIEW OF CONCEPTS AND BACKGROUND 2.1 Introduction to Chapter Two In this chapter’s four parts, I provide a framework of relevant research and concepts regarding the relationship between altitude and health problems. First, the altitude research and health from Hippocrates to the present will be addressed, considering notions of hypobaric hypoxia. Then, concepts about affective psychopathology and depressed mood pertinent to this dissertation will be reviewed. The third section will address background research, especially the Hypoxia Theory regarding depressive symptoms. Then, background research of epidemiological studies on the altitude/depression association will be addressed. 2.2 Concepts 2.2.1 Altitude Research from Hippocrates to Nowadays Diseases have been linked to climate, environmental conditions, and health since ancient times. A notable work on this subject was the book attributed to Hippocrates called “Of Airs, Waters, and Places” (5th century BC), considered by some scholars to be a collection of essay fragments and essays from multiple authors. Whether one person or many, the Greek physician Hippocrates is regarded as the father of Western medicine, and his extensive writings on medicine and health date back to the 5th century BC. In "Of Airs, Waters, and Places," Hippocrates argues that the environment in which a person lives significantly impacts their health. He argued that a region's climate, water quality, and air quality could influence the prevalence of diseases and illnesses (16). Although Hippocrates didn't mention altitude specifically, the Hippocratic essays introduced a mainstream theory of disease causation known as the “miasma theory” (17). In 6 particular, it was observed that elevations higher than low-lying swamps and marshes might be conducive to a reduced disease occurrence. Consequently, populations living high above miasmas might be less susceptible to diseases associated with them. Three centuries later, the influential physician Galen extended the work of Hippocrates. Galen’s medical approach also conceived the environment as one of the most significant factors for health (18). These ideas and the miasma theory were carried forward by many medical authorities for thousands of years. Even in the 19th century, the miasma theory and other environmental factors were addressed by William Farr. In 1843, in his Fifth Annual Report, he described how rurality was related to health outcomes. With a numerical approach, he described how population density was associated with higher mortality. He extended further, proposing an equation calculating mortality rates according to population density (19). Farr also endorsed the miasmatic theory of the disease. He found that cholera deaths were more frequent in places at low altitude levels. Contradictory evidence came from John Snow, who argued that water was the source of cholera. For many historians of epidemiology, Snow’s work qualifies him as the European father of modern environmental epidemiology (20,21). Another milestone in studying health-related environmental factors was the Great Smog of London in 1952. Thousands of deaths followed this combination of smoke and fog, finally leading to changes in the British legislation designed to reduce air contamination (20). During the 20th century and early 21st century, the epidemiological study of the environment and its effects on health has become increasingly intensive. Central themes in this 7 research include a growing interest in the harmful effects of living in large cities, changing lifestyles, and especially in the environmental changes caused by man (22). In this case, the urbanization process in the Peruvian Andes, related to developing new cities adjacent to mining areas, allowed clinicians from the early 20th century to conduct medical studies in high-altitude populations (23). Over the past century, much research has been undertaken regarding high-altitude medicine, particularly concerning the physiology of adaptation and the health problems that arise when this adaptation is lost. However, very little has been studied regarding the effects of high altitude on emotional and mental well-being. To delve into current knowledge regarding the potential effects of high altitude on emotional health, first, I will elaborate on theoretical aspects of hypoxia at high altitudes (i.e., hypobaric hypoxia). Altitude and hypobaric hypoxia The oxygen levels in the breathing air are related to the atmospheric pressure. At sea level, the atmospheric pressure is known as one atmospheric pressure or 760 mmHg. At this level, the corresponding oxygen pressure is around 149 mmHg, about 21% of the total atmospheric composition. When the altitude level goes up, the corresponding value of the atmospheric pressure declines in an exponential function (i.e., hypobaric conditions). Accordingly, the partial oxygen pressure decreases exponentially due to the decrease in atmospheric pressure (see Figure 2.1) when altitude levels increase. 8 Figure 2.1. Effective oxygen level (expressed as a percentage) plotted against altitude (in meters)1. ) % ( n e g y x O e v i t c e f f E 25 20 15 10 5 0 Note the marked negative exponential decline of effective oxygen levels (y-axis) across the altitude gradient (x-axis) 0 2000 4000 6000 Atitude in meters 8000 10000 The levels of environmentally effective oxygen pressure are directly related to arterial oxygen saturation (SaO2), the percentage of oxygenated hemoglobin usually measured noninvasively by finger pulse oximetry. The values of SaO2 also have an inverse relation with the altitude level (24,25). Hypobaric hypoxia is related to several health adaptative processes and health problems. Sometimes, the adaptative mechanisms do not appear adequately in people who ascend to high altitudes, and sometimes, the person who lives in the highlands loses their physiologic adaptative mechanisms. Both ways generate some health problems and diseases. 1 Footnote https://hypoxico.com/pages/altitude-to-oxygen-chart to figure 2.1: this figure was made using raw data found in 9 Acute altitude sickness For more than 100 years, an acute sickness related to ascending to high altitudes has been recognized by medical authorities. In 1927, Harold Crane described cases of patients with cough, blood in the sputum, and lung congestion in the Peruvian mining city of Cerro de Pasco. Interestingly, he described a quick recovery after descending to sea level (23). Some years later, Lizarraga described a series of cases with the term “acute soroche” and described these cases with the current medical condition known as pulmonary edema (26). Nowadays, this medical condition is called High-Altitude Pulmonary Edema (HAPE) and is one of the three more important acute diseases related to high-altitude exposure (27). The other two acute illnesses related to climbing at high altitudes are acute mountain sickness (AMS) and high-altitude cerebral edema (HACE). AMS should be suspected when a recently ascended person has a headache and one or more of the following symptoms: nausea, vomiting, anorexia, insomnia, or dizziness. HACE is diagnosed when some neurologic signs appear in a person with AMS or HAPE, such as altered consciousness, ataxia, or hallucinations (28). In general, these acute altitude sicknesses appear at elevations above 2,500 meters. However, cases have also been found at lower altitudes, but all three increase with higher altitudes (27). Cases are expected to occur from some hours to five days after the person arrives in high-altitude places. A study in the Alps found 9% AMS incidence at 2850 m. At 3050 and 3650 m, AMS incidence increased to 13% and 34-57%, respectively. Around 12% of cases at 3600m were 10 hospitalized for treatment (29). HAPE and HACE occur less frequently than AMS. In unacclimatized hikers at 4,243 meters, HAPE and HACE occurred in 2.5% and 1.8% of cases (29). Chronic Mountain Sickness (CMS) For adaptative purposes, chronic hypoxia causes the pulmonary arterial vessels to become more pressurized, counteracting the lower partial pressure of oxygen. This condition is called Pulmonary Hypertension (PH) and causes high lung blood pressure. These findings are correlated with the thickening of the pulmonary arterial muscular layers. Shortly after birth, pulmonary arterial musculature thickens differently in children born at sea level versus high altitudes (30,25). Several other adaptative mechanisms are described among healthy highlanders (i.e., long-term residents of high-altitude dwellings). For example, more ventilation in the Highlands and higher hemoglobin concentrations are well-defined and recognized. Erythropoietin (i.e., the hormone that induces red blood cell production) levels also increase in native highlanders. Some evidence indicates a genetic variation due to evolutionary processes in different native highland populations (31). Chronic mountain sickness (CMS) was described early in the 20th century by Carlos Monge Sr., and since then, it has been attributed to the loss of altitude adaptation capability (32). Hypoxia and polycythemia are increased in these cases, and pulmonary hypertension is more severe. It is a public health issue in mountainous regions around the world, including the Andes (30). As adaptability declines, alveolar ventilation decreases, particularly at night. As a result, residents suffer from greater levels of hypoxia than usual. As a result of this lower oxygenation, 11 erythropoietin increases, and polycythemia is exaggerated. Additionally, pulmonary hypertension increases, and in some cases, cardiac morphology (right ventricular hypertrophy) is altered. The most common symptoms include fatigue, sleep problems, headaches, dizziness, and mental fatigue. The treatment usually involves descending to lower altitudes or bleeding (30). According to Carlos Monge's studies, CMS disease increases with aging due to the progressive loss of the native's greater ventilatory capacity (33). Studies that measure the prevalence of CMS in large populations are rare. I have found reports on a CMS prevalence of around 5% for the Andes, observed at altitudes between 3,600 and 3,800 meters (34,35). In Tibet, lower prevalence values have been found for poorly understood reasons (36). Extensive research is being conducted on altitude's acute and chronic effects on the cardiovascular, respiratory, and renal systems. Neurological aspects, such as headache, sleep quality, and neurocognitive effects, have been the subject of increased research in recent decades. Also, genetic studies have shown differences between native highlanders and other populations regarding their adaptation to hypoxia. Explanations for the Tibetan and Andean population contrasts have included speculations about variations in evolutionary processes (31,37). Finally, as mentioned earlier in this chapter, some epidemiological evidence has emerged during the 21st century, suggesting that the altitude at which populations live may affect mood and suicidal behavior as a consequence of hypoxia, as described in later sections of this chapter. 12 2.2.2 Introductory Psychopathology of Affect (concepts of mood, depressed mood, and depressive disorder) The Bible contains some of the earliest descriptions of depressed mood. In the Old Testament, Saul is despondent after God's rejection and losing battles against enemy armies. He eventually requests to be killed and ends up committing suicide (38). In ancient Greece, the previously mentioned Hippocrates was the first to provide a clinical description of depression. In the Corpus Hippocraticum, the melancholic temperament is described as particularly prone to depressive episodes, especially during the autumn. This condition was attributed to an excess of black bile, one of the four fundamental elements of the human body. Hippocrates is credited with the following description: "Fear or sadness that lasts a long time means melancholia" (39). Areteus of Cappadocia (1st century AD) defined melancholia in two dimensions: (a) as an emotional state characterized by anguish and (b) as an intellectual state characterized by a delusional conception. It was centuries later, in Bright's treatise on melancholy (1586), that he described melancholia as being "for the most part sad and fearful accompanied by distrust, doubt, diffidence, and despair"(40). Consistent with these ideas, throughout history, from the earliest medical texts to the present day, deep sadness and its variants, such as hopelessness, sorrow, and emptiness, have been mentioned as the central features of depression (39). At present, we can understand “affectivity” as an integration of a set of states and tendencies that the individual experiences in a personal and immediate way with influences observed in behavioral manifestations, usually distributed in dual terms (joy-sadness, pleasure- 13 pain). For psychopathologists, affectivity covers a large set of experiences that define people's emotional lives (41). The American Psychological Association (APA) describes affectivity in these terms (42): n. the degree of a person’s response or susceptibility to pleasure, pain, and other emotional stimuli. Evaluation of affectivity is an important component of a psychological examination; the therapist or clinician may look for evidence of such reactions as blunted affect, inappropriate affect, loss of affect, ambivalence, depersonalization, elation, depression, or anxiety. (Quotation from page 28, APA Dictionary of Psychology, 2015). There are many ways in which affectivity manifests, but the most important ones are (41): • Emotions: sudden, intense, and acute reactions related to the psychophysiological concept of reaction. There is an abundant somatic correlation with symptoms or signs that might last for a short time. Fear, sadness, anxiety, and anger are examples of emotions. • Mood: a disposition or affective state more stable and persistent than emotions. Usually, it oscillates between happiness and sadness. It is also described as the sustained and continuous emotion that is subjectively experienced and can be observed by others. The most common mood states are those located on the joy- sadness axis, but mood can also be experienced as irritability, anger, or anxiety. In euthymia, the mood is considered to be in a normal state. 14 Depressed mood is one of the fundamental criteria for diagnosing depressive episodes (43,44). It is worthwhile to highlight the conceptual difference between mood and emotional states, which are more transitory. The mood is conceptualized as the subject's baseline affective state, established more slowly and progressively than emotions, and is more enduring and stable over time (41,45). In this line, Bleuler described that emotional states tend to prolong into persistent moods, influencing the person's entire experience. However, the mood is not as dependent on experiences as on the disposition it generates (46). Affect as a construct is broad and includes both emotions and moods. Within the Affect domain, emotions are defined as biobehavioral systems that present four components: 1) a subjective experience. 2) a physiological reaction. 3) an expressive component (e.g., facial expression), and 4) a behavioral response. These components occur as part of an intense and coordinated response that lasts for brief periods, usually seconds or minutes (47). Mood also represents a subjective experience of oneself and has an evaluative quality of feeling either positively or negatively. However, mood does not necessarily generate emotional responses since mood does not exhibit the four emotional components mentioned. In emotions, for example, anger involves dramatic manifestations of the four components. In contrast, the experience of an irritable mood does not necessarily involve the other three components (e.g., the face does not necessarily express anger). While emotions are brief and intense, moods can be less dramatic and last longer (41). Another distinction between emotions and moods is related to activations or triggers. Emotions can have identifiable triggers or events that activate a coordinated response. On the 15 other hand, moods are generally present continuously and without a clear trigger or reference event. They can dissipate without clear intervention or environmental change. That's why a person can experience a depressed or dysphoric mood without knowing why. One reason for this is that moods are strongly influenced by various endogenous processes, such as circadian rhythms (47). For affectivity research, evaluating mood states rather than emotions is advisable. However, emotions and moods are not mutually exclusive. Similar processes and components can affect both. The emotions we feel can influence our mood, and moods can alter our likelihood of experiencing specific emotions (47). Depressed mood and depressive episode A Depressed Mood (DM) and a Major Depressive Episode (MDE) are related. In short, a depressed mood is a core symptom or sign commonly experienced during a Major Depressive Episode. The mood state might not be conveyed by a person’s own subjective description of psychological well-being. In some instances, particularly noteworthy in younger children before the Piagetian stages of abstract thinking about one’s self, and in later life, there can be ‘depression equivalents’ more readily seen as signs of disturbed behavior and maladaptation. In children, the equivalents might include acting out, striking others, or striking oneself. In older adults, the equivalents might manifest in irritability or visible shifts in behavioral patterns away from activities that usually have functioned as reinforcers for the individual. A depressed mood refers to a sustained sadness or hopelessness (44). In general, the subjective feeling of sadness is present most of the time. It is a symptom of MDE but can also be a symptom of other mental health disorders, such as adaptative disorders or dysthymia. 16 For most of the past 40-50 years, the concept of a depressive episode has referred to a period of two or more weeks where a person experiences a combination of symptoms or signs of depression (or depression equivalents), such as a depressed mood (DM), feelings of worthlessness or guilt, fatigue, changes in appetite and sleep patterns, difficulty concentrating, and in severe cases, thoughts of self-harm or suicide (43,44). Since 1980, the American Psychiatric Association has defined a depressive episode as a necessary requirement to diagnose Major Depressive Disorder (MDD), a type of depression. Table 2.1 shows the full criteria for depressive episodes in DSM and ICD; note that depressed mood is the most important of the core symptoms in both diagnostic systems. 17 Table 2.1. A Diagnosable Depression as Defined in the Diagnostic and Statistical Manual (5th Edition) of the American Psychiatric Association and the International Classification of Diseases of the World Health Organization Glossary of Mental Disorders (10th edition)2. DSM-5 At least five symptoms in total for diagnosis • Depressed mood† • Markedly diminished interest or ICD-10 At least four symptoms in total for diagnosis • Depressed mood‡ • Loss of interest and enjoyment‡ • Reduced energy and diminished activity‡ • Reduced concentration and attention • Reduced self-esteem and self-confidence • Ideas of guilt and unworthiness • Bleak and pessimistic views of the future • Ideas or acts of self-harm or suicide. • Disturbed sleep • Diminished appetite pleasure† • Significant weight loss or weight gain • Insomnia or hypersomnia • Psychomotor agitation or retardation • Fatigue or loss of energy • Feelings of worthlessness or excessive or inappropriate guilt • Diminished ability to think or concentrate, or indecisiveness. • Recurrent thoughts of death, recurrent suicidal ideation, or a suicide attempt † Core symptoms of the DSM-5 criteria: at least one of these two must be present for diagnosis. ‡ Core symptoms of the ICD-10 criteria: at least two of these three must be present for diagnosis. 2.3 Background 2.3.1 The Hypoxia Theory of mood changes Multiple pathways might link hypoxia to mood disturbances. First, reduced serotonin function in neurotransmission pathways has been described. Also, inflammatory changes and reduced brain bioenergetics have been found. Finally, several sleep changes have been detected with hypobaric hypoxia and could have a role in the emergence of mood disturbances. 2 Footnote to Table 2.1: the original of this table was found in Park et. al., 2020 (48), which I then adapted by myself for this doctoral dissertation. 18 Studies with animal models exposed to chronic hypoxic environments found a significant decrease in serotonin synthesis due to reduced tryptophan hydroxylase activity (49). Also, it has been reported that the effects on mood secondary to exposure to high altitude may be more significant in women due to differences in tryptophan and serotonin metabolism (50). On the other hand, it has been postulated that hypobaric hypoxia causes inflammatory changes in cells related to depressive symptoms. Specifically, using animal models, it has been found that after exposure to moderate altitudes (i.e., 1,600 m), biological signs of cellular inflammation appear, such as 1) an increased granulocyte/lymphocyte ratio; 2) an increased number of circulating monocytes; and 3) an increased monocyte/lymphocyte ratio. The animal model found that this cellular inflammation was correlated with signs of depression (51). These animal models have also shown that chronic exposure to hypoxia generates inflammatory changes that could cause structural damage at the cellular level in the hippocampus (52). It has also been reported that these harmful effects are more intense in males and suggest that sex hormones are involved in the susceptibility to present brain damage due to chronic exposure to hypoxia (52). It was proposed that brain bioenergetics is also involved in the emergence of mood disturbances among people exposed to hypobaric hypoxia (8). The production of adenosine triphosphate (ATP) in hypoxic environments is diminished due to changes in the creatine kinase reaction (53). Deficiencies in ATP production and creatine kinase function are involved in neuronal dysfunction, which can correlate with mood disturbances (54). Sleep experimental studies with voluntary participants have investigated the effects of acute, subacute, and chronic hypoxia. For example, it has been clearly described that changes in 19 sleep quality and mood state appear very soon under artificial hypoxic conditions. Total sleep time, sleep quality, slow-wave sleep, and REM sleep are reduced (55). Considering that sleep quality and quantity are associated with the development of anxious and depressive symptoms, it would be expected that changes in mood state could also be observed rapidly from the early years of residency in the highlands. It has also been found that these sleep disturbances increase at higher altitudes (56). A longitudinal study of one year on a group of volunteers found that ascending to high altitudes caused mood changes even within the first year. These changes were related to the magnitude of altitude. It was also found that these changes were mediated by decreased sleep quality (57). Scientific literature finds a significant association between sleep problems induced by altitude and the emergence of cognitive and anxiety problems, which may later be associated with an increased occurrence of mood symptoms such as anxiety and depression. It has been identified that further research is needed on altitude-associated mood changes (58). Studies on the negative cognitive effects at high altitudes in lowlanders who ascend to greater heights have found a variation over residency time. An oscillating evolution has been described; in the beginning, there is a marked deterioration of some neuropsychological functions, then a period of acclimatization follows, and finally, in some cases, a deterioration can be seen again (59). However, there are doubts about whether the findings obtained in individuals who ascend in altitude can be applied to long-term residents who have lived in highlands for several generations (59). 20 2.3.2 Recent Research Pertinent to the Hypoxia Theory and mood disturbances The most relevant literature studying the relationship between altitude and mood disturbances started in this century. First, several studies using aggregated data in the United States found more significant suicide proportions among populations at higher altitudes (60,61,6,62). Then, this association was also found in other countries (63,64). After the reported suicide-altitude relationship, Delmastro 2011 found a positive correlation between depression prevalence proportion in the last year and the altitude of residence at the U.S.A. substate level (5). Then, some studies focusing on the individual level of analysis found the same positive association between depression and hypobaric hypoxia. Kious, in a cohort study and using the Patient Health Questionnaire-9 (PHQ-9), found that changing the altitude of residence to a higher altitude was associated with increased scores of depression, anxiety, and suicidality (4). Outside of the United States, three large epidemiological studies examined the relationship between altitude and depression with the inclusion of especially high-altitude communities. Most of these prior studies were conducted and then published during the years after I originally proposed the focus of my dissertation research project. These three studies were conducted, one in Nepal (65) and two in Peru, both published in 2022 (66,67). The case definitions for these prior studies are different, but all found an association between altitude and depressive symptoms. The study in Nepal was nationwide and tried to estimate anxiety and depression prevalence across the country. The Hospital Anxiety and Depression Scale (HADS) included seven questions about depressive symptoms during the last week before the survey. After using 21 score thresholds for defining cases, the multiple logistic regression modeling found that people living above 2,000 meters have a higher risk for depression (65). The two studies in Peru used data from the Demographic and Family Health Survey (ENDES in Spanish), which included PHQ-9 to evaluate depressive symptoms during the last two weeks before the assessment. These two studies defined different outcome variables. One considered the score of depressive symptoms, as elicited in a multi-item scale, and implemented Gamma and Quantile regression models because of scores overdispersion (66), and the other study imposed a threshold cut point on that summary score and analyzed using Poisson regressions (67). Both studies found that depressive symptoms are associated with altitude, the first using altitude as a continuous variable and the second using three categories of altitude levels. 22 CHAPTER 3: ESTIMATING THE ALTITUDE-DEPRESSED MOOD ASSOCIATION AT THE DISTRICT 3.1 Introduction LEVEL This study examines whether estimated prevalence proportions of depressed mood (DM) at the community district level might be greater in communities at higher altitudes across 223 districts from 33 provinces in Peru. For this purpose, nine epidemiological surveys from the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health - "Noguchi Surveys" have been consolidated. According to the World Health Organization (WHO), depression is the primary cause of the burden of disease globally in terms of nonfatal health outcomes. Also, it is the primary cause of suicide (68). Around 500 million people live in the highlands (69), and millions travel to the highland regions yearly. Confirming the Altitude-Depression relationship is crucial for public health reasons and raising awareness of the need for further research into the underlying mechanisms. The most prominent prior contributions to the study of the relationship between altitude and depression or suicide have been published during the last 20 years. During the early 2000s, reports were published regarding the correlation between suicide rates and high altitude (6,9,60,70). The study of DelMastro in 2011 was the first large epidemiological study on altitude and depression. It found a positive correlation between depression prevalence proportion in the previous year of the interview and the altitude of residence at the substate level in the US (5). Later, some studies focused on the individual level of analysis and found similar associations between altitude and depressive symptoms (4,65,66,71,72). 23 The Small Area Estimation (SAE) approach has been described since 1979 (73). It is used to obtain more precise population estimates for small-scale areas than those initially projected in the original studies (74). SAE methods achieve this by combining original data analysis with auxiliary data from other sources (e.g., government data). This auxiliary data addresses related variables but with more precise estimates. In this study, I will use the SAE Fay-Herriot model to calculate more accurate estimates of DM prevalence at the district level. For estimating DM prevalence proportions at the community district level variations across communities at higher and lower altitudes, I used the direct weighted estimates of survey data analysis, and the standardized estimates (i.e., with direct adjustment for age and sex). Finally, the estimators from the SAE methodology (i.e., Fay-Herriot model) are presented. 3.2 Materials and Methods This research includes cross-sectional analyses of nine yearly epidemiological studies in Peru. These surveys were completed sequentially between 2003 and 2013 by the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health of Peru. They included 15 regions of Peru, covering a broad range of altitudes above sea level. Each year, the surveys were implemented with almost similar research approaches and questionnaires, addressing sociodemographic variables, mental health problems, alcohol and other drug use (including coca leaf consumption), and other covariates. Population and Districts Under Study in Peru, 2003-2013 These household surveys are designed to represent four target groups of non- institutionalized community residents of Peru: adolescents, adults (i.e.,18 years or older, including elders), elderly individuals, and married or cohabitant women. (75). In this research, I 24 tapped data only from the adult population modules. Only household residents were included. The Peruvian National Institute of Statistics and Informatics provided the census clusters for the sampling procedures based on the most recent census survey of the Peruvian population available. It is important to note that the population under study was based on the residence of the Peruvian community members (i.e., where they resided). Each yearly survey examined one or more cities or rural areas. In total, 22 rural or urban areas (i.e., 22 strata) and 223 districts were surveyed (see Tables 1 and 2). The total aggregated target population of adults considered for the survey sampling procedures along these surveys was more than 9.8 million people. Description of the Sampling Approach and Aggregate Sample Size, Before Analysis Weights The sample size calculation for each yearly survey was made using the equation recommended by Lohr (76). It was considered a 95% confidence interval for the estimations and different expected prevalence percentages of mental disorders for each survey. For example, for the 2008 survey, it was considered 30%, and for 2012, it was set at 15% based on previous studies (77,78). Researchers from the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health used the same sample selection process for each surveyed area (i.e., urban or rural). It implemented a multistage random sampling process, carried out in three stages. Firstly, a group of census clusters was randomly selected (i.e., primary sampling units) within each stratum (i.e., the city or rural surveyed area). Usually, each census cluster includes around 80 dwelling units. Secondly, dwelling units within each census cluster were randomly selected. Finally, individuals from each target population within each dwelling unit were randomly chosen 25 using the Kish table approach. To accomplish this, the field staff member gathered a roster of dwelling unit residents (75,77). Usually, there is more than one census cluster in each district, but the district level was not necessarily considered for the staging of the sampling. Multiple institute reports have provided further details about the survey methods and research approaches employed by the “Noguchi” project. One convenient and easily accessible source of information is Dr. Victor Cruz's Master of Science thesis, which is available as an online document from Michigan State University (MSU) (77). 26 Ancash Apurimac Arequipa Ayacucho Urban Highland Urban Highland Cajamarca Urban Highland Huancavelica Huanuco Ica La Libertad Lima Loreto Pasco Piura San Martin Ucayali Table 3.1. “Noguchi Surveys”1 and regions across years. Population characteristics. Peru, 2003-2013. Region Year 2003 2004 2006 2007 2008 2009 2010 2012 2013 Urban Coast Urban Highland Rural Highland Rural Highland Rural Highland Urban Highland Urban Highland Total 3626 1746 1332 2324 2310 1716 Urban Coast Urban Coast Rural Highland Urban Coast Urban Rain Forest Urban Rain Forest Urban Rain Forest Rural Rain Forest Rural Rain Forest Metropolitan Coast Urban Highland 1496 1355 1222 6981 2486 Urban Highland 1469 1312 1340 2415 2965 33130 Total 1 “Noguchi Surveys”: epidemiological surveys from the "Honorio Delgado-Hideyo Noguchi" Peruvian National Institute of Mental Health. 3462 4445 3910 3031 6555 2331 2536 3895 27 Table 3.2. Survey areas, population, and samples. Survey Areas (Strata) Ancash Urban Highland Ayacucho Urban Highland Cajamarca Urban Highland Loreto Urban Rain Forest San Martin Urban Rain Forest Ucayali Urban Rain Forest Ancash Urban Coast (Chimbote) Arequipa Urban Highland Ica Urban Coast La Libertad Urban Coast (Trujillo) Piura Urban Coast Lima Rural Highland Ancash Rural Highland Ayacucho Rural Highland Cajamarca Rural Highland Loreto Rural Rain Forest Ucayali Rural Rain Forest Apurimac Urban Highland Huancavelica Urban Highland Metrop. Lima Coast Huanuco Urban Highland Pasco Urban Highland Total 1 PSU: primary sampling units Year 2003 2003 2003 2004 2004 2004 2006 2006 2006 2006 2006 2007 2008 2008 2008 2009 2009 2010 2010 2012 2013 2013 9 Number of PSUs1 117 117 117 117 117 117 117 117 117 117 117 131 70 70 70 40 40 99 75 700 172 125 2879 Total population considered Sample 1319 1301 1275 1306 1340 1264 1334 1332 1355 1222 1312 2536 973 1023 1035 1180 1151 1746 1716 4445 1496 1469 33130 53080 76556 101277 241022 67381 129419 401417 802309 223302 916716 409824 50708 49696 54971 137479 12648 13884 33442 25851 5899105 104166 39615 9843866 The participation levels for the Peruvian National Institute of Mental Health "Honorio Delgado - Hideyo Noguchi” surveys are described for each of the yearly surveys in Table 3.3. These values varied from 81.7% to 98.4%, most larger than 90%. The number of participants in each yearly survey ranged from 2,331 to 6,555. The total resulting sample consists of 33,130 participants. 28 Table 3.3. Surveys’ response rates. Planned households Rejection Absence 69 4212 74 Response rate (%) 96.3 Final sample 3895 Percent 11.76 2003 Ancash Ayacucho Cajamarca 2004 Loreto, San Martin Ucayali 2006 Ancash Arequipa Ica La Libertad Piura 2007 (Lima Rural) 2008 (Rural) Ancash Ayacucho Cajamarca 2009 (Rural) Loreto Ucayali 2010 Apurimac Huancavelica 2012 (Lima Callao) 2013 Huánuco Pasco Total sample 5616 42 80 97.8 3910 11.80 9360 108 39 98.4 6555 19.79 Planned adults Rejection Absence /other Response rate (%) Final sample Percent 2628 3196 0.5% 1.4% 3% 3.7% 96.5 94.8 2536 3031 7.65 9.15 2400 - - 97.1 2331 7.04 1824 1807 5332 2.6% 3.0% 10.4% 1.6% 2.0% 6.2% 1786 1797 8.9% 13.2% 7.3% 5.1% 95.7 95.0 83.4 83.8 81.7 3462 1746 1716 4445 2965 1496 1469 33130 10.45 13.42 8.95 100.00 Some districts were selected for sampling in two different years; in these cases, a decision was made considering the sample size (i.e., participants from the year with the smallest number were excluded) and the year of sampling (i.e., if equal sample sizes, 29 participants from the last year were excluded). In addition, 49 participants had missing responses to key study variables and were excluded from the estimations. For these reasons, the effective sample size for this research is 31729 (95.8% of the total initial sample). The unweighted numbers of survey participants, district by district, are reported in the first Table of this paper’s Results section. IRB Considerations The Peruvian National Institute of Mental Health Institutional Review Board for Protection of Human Subjects in Research reviewed and approved all the protocols utilized for the "Noguchi Surveys." Before beginning the interviews, every participant in the study provided their signature on an informed consent form. Participation in the study was entirely voluntary, and participants had the right to decline to answer any questions they found uncomfortable or terminate the interview at any point. In addition, the protocol for this research was reviewed by the Michigan State University Institutional Review Board for the Protection of Human Subjects in Research (IRB). As this research study was carried out with de-identified data, the IRB considered that this research does not involve human subjects as defined by the US National Institutes of Health. Instruments and measurement The survey questionnaire for adults has five sections (i.e., sociodemographic, general adult health, clinical syndromes part A, clinical syndromes part B, and access to health services) and contains 233 questions. Appendix A of this dissertation includes an English translation of the standardized survey items used in this study to assess each participant's mood state during the last four weeks. Questions were applied to participants in their native “mother tongue.” The 30 altitude of households was identified at the district level and retrieved from the Peruvian National Center for Strategic Planning (CEPLAN) (79). Actively depressed mood (i.e., depressed mood during the last four weeks at the moment of the survey interview) was assessed in two different parts of the questionnaires. The first included a Likert question directly asking about the mood of the participant: “How often do you feel Sad?” (i.e., Question 6, from the “prevalent mood states” part of the questionnaire). This question was applied to all participants. The second question about depressed mood was included as one of the items for the screening tool SRQ-17 (80). This abbreviated version of the SRQ-20 has 17 questions compared to the original version. These questions were applied to all participants inquiring about depressive and anxiety symptoms in the last four weeks. One of these questions about the previous four weeks was: “Have you felt sad frequently?” (i.e., Question 26). In addition to frequent sadness during the last four weeks, the SRQ screening tool asks about physical symptoms, appetite, anxiety, and other than depressed mood questions. Actively depressed mood was finally identified in persons who recognized feeling sadness “always” or “almost always” in question #6 and also answered “Yes” to the question “Have you felt sad frequently in the last four weeks?”. The concept of mood in the psychopathology of affect is conceptualized as the subject’s baseline affective state and is more enduring and stable over time (41,45). For DSM 5, a depressed mood refers to a “sustained feeling of sadness or hopelessness” (44). Thus, the subjective feeling of sadness is present most of the time. It is a symptom of Major Depressive Disorder (MDD) but can also be a symptom of other mental health disorders. 31 The Mini-International Neuropsychiatric Interview - MINI (81), which includes questions that match the criteria of ICD-10 for MDD, was also included in the surveys. However, in MINI, the question about depressed mood was applied over the participant's lifetime. Prevalence proportions of Depressed Mood (DM) at the district level Prevalence proportions of DM at the district level (i.e., 223 districts) with the aggregated data were estimated in three different ways. First, I obtained survey proportions estimates using Stata © commands, considering the respective strata, clusters, and weights. These are called the “direct estimates”. Then, I calculated the standardized prevalence proportions using the direct standardization method, which was adjusted for sex and age categories. Three age categories were considered: 18-44, 45-64, and >=65 years. These categories are meaningful regarding depression because of the possible relationship between depressive symptoms and postmenopausal ages in women. The reference population for the standardization was the Peruvian census from 2005 (accessible from: censos.inei.gob.pe/Censos2005/redatam/) and their openly available numbers. Finally, I utilized the Small Area Estimation framework to compute the Empirical Best Linear Unbiased Prediction (EBLUP) estimators based on the Fay- Herriot (FH) model (73,82). The Fay-Herriot (FH) model The FH model was introduced in 1979 (73) and opened the development of the Small Area Estimation (SAE) methodologies. The general objective of these models is to obtain better estimates from survey data, which was not initially designed for having reliable estimates for more small areas or “domains.” Nowadays, the United Nations stimulates the development of SAE models in the Sustainable Development Goal (SDG) agenda (83). 32 The FH is a two-part model and uses auxiliary data of the small areas or domains. These auxiliary data come from a different data source. The Fay-Herriot model is designed to increase accuracy by using direct estimates from the survey data and the auxiliary data obtained from any official registers or administrative sources at the domain level. This auxiliary data should be measured with greater accuracy and is associated with the outcome of interest, so it contains additional information that is utilized to correct the direct estimates. The FH model uses this additional information to forecast the outcome of interest with a linear regression model. The FH model is described for “m” small areas (84): The Sampling Model: 𝑌̂𝑖 = 𝜃𝑖  +  𝑒𝑖               𝑖 = 1,  2, … , 𝑚 The Linking Model: θ𝑖 = 𝑥′ 𝑖β + 𝑢𝑖 Where : θ𝑖 is the population characteristic of interest for the area i. 𝑌̂𝑖 is the direct survey estimate of θ𝑖. 𝑒𝑖 is the sampling error in 𝑌̂𝑖, generally assumed to be 𝑁(0, 𝑣𝑖) with 𝑣𝑖 known. 𝑢𝑖 is the area i random effect, usually assumed to be i.i.d 𝑁(0, σ𝑢 independent of the 𝑒𝑖. “Synthetic estimator” 𝑥′ 2 ) and 𝑖β. The best linear predictor is a linear combination of the “direct estimator” and the “synthetic estimator”: The best linear predictor of θ𝑖 (β and σ𝑢 2 known): Where: 𝜃̂𝑖 = (1 − γ𝑖)𝑌̂𝑖 + γ𝑖𝑥′ 𝑖β γ𝑖 = 𝑣𝑖 2 𝑣𝑖 + σ𝑢 The FH models for obtaining EBLUPs estimates were executed in Stata© with the fayherriot command (85). As described by the Stata command authors, the “arcsine of the 33 estimate square root” transformation is a good option when using a proportion variable (i.e., in our case, prevalence proportion) direct estimates. Then, the estimates were back-transformed using the arcsine square root transformation to ensure that EBLUPs are restricted to the interval zero to one (see footnote)3. When standard errors of the direct estimates were invalid (i.e., 40 districts, 13 without cases, and 27 with infinitesimal values), they were inputted with the minimum value obtained in the other domains. On the other hand, the FH model without including these 40 districts was also computed for comparison reasons. The auxiliary data used three domain-level variables with a more reliable estimation provided by the Peruvian National Institute of Statistics and Informatics (INEI) and Peruvian National Center for Strategic Planning (CEPLAN). The other variables for calculating the synthetic estimator were population density, poverty rates, and human development index (HDI). HDI is built based on three indicators: life expectancy at birth, the proportion of the population over 18 with secondary education, and per capita family income. Values of the HDI close to 1 will indicate a better position of human development in the territory. After obtaining the three estimators for the prevalence proportion of depressed mood in the districts (i.e., direct weighted estimates, standardized estimates adjusting for age and sex, and EBLUPs), I used linear regressions models to examine the relationship between altitude and depressed mood after transforming the estimates with the “arcsine of the estimate square root” function. Then, models were compared using these three different estimates. The relationship between altitude and DM prevalence proportions was tested using altitude as a 3 asin(x)= the radian value of the arcsine of x. arcsine of x is the angle which their sine is x. 34 continuous variable and then as categorical with terciles and deciles of altitude. Finally, the prediction command was used to evaluate how the predicted values of the prevalence proportion of DM go along altitude levels. 3.3 Results Table 3.4 shows the names of all 223 sampled districts sorted by altitude (in meters, m) above sea level. Column four shows the year of the survey. The fifth column shows the unweighted size of the “Noguchi Survey” sample in that district that year, and the last column indicates the number of people with depressed mood (DM). The methods section described that if a district was selected for sampling in two different years, a decision was made considering the sample size and the year of the study. The estimates in this project, based on samples, are either for the initial year the district was sampled or for the most significant sample size obtained for that district, without considering the weights of the samples. The analysis-weighted district population sizes are reported in Table 3.10 (Appendix B). Also, the weighted proportions for depressed mood (DM) and the results from direct standardization methods for adjusting for sex and age categories (i.e., 18-44, 45-64, >=65 years). Finally, in Table 3.10, we can also see the results from the Fay Herriot approach for calculating the called empirical best linear unbiased prediction (EBLUP), which incorporates information from the survey and auxiliary data (i.e., population density, poverty rates, and human development index for each district). 35 Table 3.4. Districts, samples, and cases. N° District 1 Lurin 2 Bellavista 3 Ancon 4 Victor Larco Herrera 5 Callao 6 La Punta 7 Coishco 8 Castilla 9 La Perla 10 Nuevo Chimbote 11 Ventanilla 12 Chimbote 13 Piura 14 Los Olivos 15 Chorrillos 16 Pachacamac 17 Santa Rosa 18 Trujillo 19 Carmen de la Legua Re 20 San Miguel 21 Magdalena del Mar 22 Magdalena Vieja 23 El Porvenir 24 Florencia de Mora 25 Barranco 26 Mazan 27 Comas 28 Santiago de Surco 29 30 31 Belen 32 San Juan Bautista Maynas 33 Punchana 34 Miraflores Lima 35 Surquillo 36 Nauta 37 Yarinacocha 38 San Juan de Miraflore 39 La Esperanza Iquitos Independencia Lima Year 2012 2012 2012 2006 2012 2012 2006 2006 2012 2006 2012 2006 2006 2012 2012 2012 2012 2006 2012 2012 2012 2012 2006 2006 2012 2009 2012 2012 2004 2012 2004 2009 2004 2012 2012 2009 2004 2012 2006 Total sample 33 34 21 93 225 5 56 457 46 477 201 801 854 112 185 45 5 508 25 49 33 26 255 60 28 146 160 156 668 60 178 413 220 55 33 234 233 197 306 Altitude (m) 12 13 14 24 27 29 31 35 37 40 43 52 57 67 68 68 72 74 82 84 90 90 92 92 97 106 107 107 107 111 116 120 124 125 125 127 131 133 137 36 Depressed mood 6 1 3 9 12 0 4 55 1 73 10 97 109 5 11 4 1 54 2 1 2 1 28 5 1 13 5 3 73 8 21 43 19 6 1 21 29 9 44 Table 3.4 (cont’d) 40 San Martin de Porres 41 Jesus Maria 42 La Victoria 43 Lince 44 Masisea 45 Breña 46 Rimac 47 Lima 48 Calleria 49 San Borja 50 Curimana 51 Nueva Requena 52 Puente Piedra 53 San Isidro 54 El Agustino 55 Campoverde 56 Villa El Salvador 57 Villa Maria del Triun 58 San Luis 59 San Juan de Luriganch 60 Irazola 61 Carabayllo 62 La Molina 63 Padre Abad 64 Santa Anita 65 Coayllo 66 Cieneguilla 67 Morales 68 Tarapoto 69 Ate 70 La Banda de Shilcayo 71 72 La Tinguina 73 Parcona 74 Llochegua 75 Sivia 76 Chaclacayo 77 Zuniga 78 Lurigancho 79 Santa Rosa de Quives Ica 12 0 4 1 9 0 4 8 110 1 6 2 12 1 7 24 25 26 0 48 24 11 4 8 15 2 0 29 83 20 14 96 30 46 11 7 3 3 14 6 2012 2012 2012 2012 2009 2012 2012 2012 2004 2012 2009 2009 2012 2012 2012 2009 2012 2012 2012 2012 2009 2012 2012 2009 2012 2007 2012 2004 2004 2012 2004 2006 2006 2006 2008 2008 2012 2007 2012 2007 198 20 85 32 107 44 75 124 1028 21 86 29 97 37 86 204 235 196 21 596 235 77 61 203 131 41 9 249 877 386 212 849 189 317 62 46 26 40 142 51 138 142 142 150 150 153 153 162 162 170 181 183 187 195 200 203 204 210 214 222 228 238 262 275 285 285 287 290 342 378 418 432 463 472 540 561 685 827 879 936 37 Table 3.4 (cont’d) Jacobo Hunter 80 Huanchay 81 Catahuasi 82 Pariacoto 83 Magdalena 84 Antioquia 85 San Bartolome 86 Huanuco 87 Amarilis 88 Putinza 89 Pillco Marca 90 Surco 91 San Mateo de Otao 92 Ambar 93 Cochabamba 94 Chumuch 95 Tiabaya 96 Asuncion 97 Sachaca 98 99 San Juan 100 Socabaya 101 Cortegana 102 103 Yanahuara 104 Arequipa 105 Cerro Colorado 106 Miraflores Arequipa 107 Paucarpata 108 Manas 109 Mariano Melgar 110 Lampian 111 Cospan 112 Abancay 113 Alto Selva Alegre 114 Cayma 115 Arahuay 116 Huasmin 117 Bambamarca Jesus 118 119 Pacaycasa 120 Luricocha Jose Luis Bustamante 30 64 29 16 39 63 729 529 25 235 54 37 55 16 15 23 30 35 69 15 119 14 138 34 124 145 78 206 73 90 45 14 1565 128 143 39 61 221 45 12 36 2008 2007 2008 2008 2007 2007 2013 2013 2007 2013 2007 2007 2007 2008 2008 2006 2008 2006 2006 2008 2006 2008 2006 2006 2006 2006 2006 2006 2007 2006 2007 2008 2010 2006 2006 2007 2008 2008 2008 2008 2008 1067 1203 1264 1298 1573 1644 1921 1950 1985 1996 2049 2084 2084 2135 2202 2218 2254 2300 2309 2336 2352 2352 2389 2402 2429 2441 2450 2453 2457 2459 2467 2471 2500 2510 2531 2533 2543 2556 2568 2571 2598 38 5 8 3 1 4 8 38 39 2 16 11 3 7 1 2 5 5 7 13 2 14 2 11 2 6 27 17 40 7 15 13 0 183 17 25 4 6 25 7 1 7 Jose Galvez Table 3.4 (cont’d) 121 Tinco 122 Marca 123 124 Tamburco 125 Llacanora 126 Celendin 127 Sucre 128 Carhuaz 129 Sorochuco 130 San Pedro de Pilas 131 San Lorenzo de Quinti 132 Huanta 133 Los Baños del Inca 134 Acopampa 135 Cajamarca 136 Namora 137 Chugur 138 Marcara 139 Sangallaya 140 Sanjuan Bautista Huamanga 141 Ayacucho 142 Anta 143 Miguel Iglesias Jesus Nazareno 144 145 Jangas 146 Pariahuanca 147 Tarica 148 Matara 149 Yungar 150 Oxamarca 151 Ihuari 152 Santo Domingo de Los 153 Acos Vinchos 154 Langa 155 Yauyos 156 Amashca 157 Huayllapampa 158 Carmen Alto 159 Huachupampa 160 La Libertad de Pallan 161 San Miguel de Aco 2606 2615 2618 2620 2621 2629 2632 2663 2663 2678 2682 2685 2685 2692 2731 2765 2765 2767 2779 2786 2797 2800 2813 2817 2824 2830 2832 2834 2836 2836 2850 2861 2874 2889 2895 2905 2908 2921 2938 2952 2956 39 2008 2008 2008 2010 2008 2008 2008 2008 2008 2007 2007 2008 2008 2008 2003 2008 2008 2008 2007 2003 2003 2008 2008 2003 2008 2008 2008 2008 2008 2008 2007 2007 2008 2007 2007 2008 2008 2003 2007 2008 2008 40 15 16 181 15 31 28 64 41 41 62 78 74 28 1197 44 13 65 39 255 792 30 15 127 59 15 42 16 26 14 100 43 47 53 84 14 12 125 21 27 32 4 3 0 11 0 2 4 15 4 5 10 18 5 5 123 1 0 17 5 46 162 4 2 18 9 5 6 4 4 1 19 7 4 6 19 1 2 28 2 5 5 Table 3.4 (cont’d) Independencia Huaraz 162 Carhuanca 163 Llacllin 164 Shilla 165 166 Gorgor 167 Concepcion 168 Iguain 169 Huaraz 170 Encañada 171 Tambillo 172 Vischongo 173 Ayauca 174 Ocros 175 Vinchos 176 Huarochiri 177 Caujul 178 Huancapon 179 Santiago de Pischa 180 Acocro 181 San Damian 182 Santillana 183 Paccho 184 San Jose de Ticllas 185 Madean 186 Huambalpa 187 Leoncio Prado 188 Huamanguilla 189 Quinua 190 Huantan 191 Viðac 192 San Andres de Tupicoc 193 Socos 194 Accomarca 195 Colonia 196 Pararin 197 Recuay 198 Azangaro 199 Olleros 200 Carampoma 201 Ticapampa 202 Vilcas Huaman 16 10 42 673 117 16 26 640 91 29 45 76 44 87 38 40 74 16 61 79 45 126 30 65 31 60 39 27 46 111 38 45 15 101 13 15 68 31 25 32 60 3 0 10 101 24 2 5 85 7 10 8 22 5 19 4 9 9 1 8 9 5 19 5 18 5 8 6 3 12 17 3 5 0 22 1 2 16 2 4 3 16 2008 2008 2008 2003 2007 2008 2008 2003 2008 2008 2008 2007 2008 2008 2007 2007 2007 2008 2008 2007 2008 2007 2008 2007 2008 2007 2008 2008 2007 2007 2007 2008 2008 2007 2008 2008 2007 2008 2007 2008 2008 2980 3020 3036 3047 3049 3061 3063 3073 3087 3111 3150 3151 3153 3155 3170 3185 3187 3210 3251 3252 3265 3275 3282 3292 3294 3299 3300 3301 3315 3315 3321 3368 3387 3399 3402 3428 3435 3443 3459 3485 3494 40 Independencia Vilcas H. Table 3.4 (cont’d) 203 Lincha 204 Hualgayoc 205 Chiara 206 Santa Cruz de Andamar 207 Pampas Chico 208 Saurama 209 Catac 210 Huancaya 211 Pira 212 213 Huaros 214 Vitis 215 Miraflores Yauyos 216 Laraos 217 Lachaqui 218 Pampas 219 Ascension 220 Huancavelica 221 Simon Bolivar 222 Chaupimarca 223 Yanacancha Total 3516 3530 3540 3550 3552 3574 3579 3591 3602 3606 3614 3625 3677 3683 3686 3698 3711 3746 4234 4373 4394 2007 2008 2008 2007 2008 2008 2008 2007 2008 2008 2007 2007 2007 2007 2007 2008 2010 2010 2013 2013 2013 36 46 45 109 14 16 14 40 45 31 88 23 37 22 43 16 366 1350 257 611 601 31729 6 10 5 24 0 8 1 6 5 5 15 4 4 2 9 1 40 132 18 32 32 3588 41 Different models were tested with the auxiliary data, including another two variables (i.e., alimentary insecurity vulnerability and extreme poverty). Finally, the model with better adjusted R2 and Fay Herriot R2 (i.e., 0.83) was selected for using their estimates. Although the SAE framework is used for having more reliable estimates in small areas for governmental public reasons, I am here using it as a novel previous step to have better estimates that will allow us to study the relationship with altitude. The screenshot of the final proposed FH model for the EBLUPs estimates for the prevalence proportion of DM is shown in Figure 3.7 (Appendix B). Although the assumptions of normality for residuals and random effects were not found, I think it is valuable to use the FH estimates to have another way to compare the relationship between altitude and depressed mood. Notably, among all the FH models evaluated with these different auxiliary variables, population density had consistently more significant coefficients (see also Figure 3.8 in Appendix B). The comparisons of the DM estimate’s distribution across the altitude in meters are shown in Figure 3.1. We can see a slight tendency for increasing estimates of DM across the altitude for the three estimates. The direct estimates (i.e., weighted prevalence proportions) and standardized estimates (i.e., using the direct standardized adjustment for sex and age) have very similar patterns (plots A and B). With EBLUPs, the pattern is less dispersed (plot C in Figure 3.1). The association between EBLUPs and the direct estimates is also shown in Figure 3.1 (plot D); after regressing these two estimates, I plotted their relationship. 42 Figure 3.1. Depressed Mood (DM) Estimates distribution across Altitude. Data from the “Noguchi Surveys,”1 Peru, 2003-2013. 1 “Noguchi Surveys”: epidemiological surveys from the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health. I used the “arcsine of the estimate square root” transformation for the three different estimates to test linear regressions between the estimates of DM and altitude as a continuous variable. We can see in Table 3.5. a summary of results from the three models. The model with EBLUPs estimates (Model C) performs better. The coefficient for altitude in Model C has a value in the middle of the two other models, but the confidence interval is smaller. All three regressions show a significant association between altitude and DM prevalence proportions in 43 223 districts of Peru. I include more details on the results of these three models in Table 3.11 (Appendix B). Table 3.5. Comparing linear regressions between altitude and depressed mood estimates: Direct, Standardized, and EBLUP estimates. Model Model A Model B Model C β Altitude(m) p-value <0.00001 0.090496 0.086381 0.116845 0.00002676 0.00002172 0.058556 0.054296 0.119972 0.00002583 <0.0000001 0.130615 0.126682 0.091790 Log-likelihood 1.63e+02 1.57e+02 2.17e+02 <0.001 Adj.R2 RMSE R2 Model A: with weighted proportion estimates (direct estimates); Model B: with standardized prevalence proportions (direct standardization by sex and age); Model C: with EBLUP estimates. All models use the “arcsine of the estimate square root” transformation. In addition, the FH model without the forty districts with null standard errors was also computed, and the regression between altitude and DM with this exclusion also showed a significant coefficient (see Figure 3.8 and Table 3.12 in Appendix B). The residual plots of these three models across the fitted values show less dispersion in Model C. But, in all three models, the range increases at higher values in relationship with higher values of altitude (Figure 3.2). 44 Figure 3.2. Residual Plots of linear regressions models A, B, and C. 45 When weighted mean age and the proportion of women in the districts are included as covariates in Models A and C, the parameter for altitude did not change appreciably, and model fit increased modestly. Only the weighted mean age for districts was associated with DM prevalence (Tables 3.6 and 3.7). Table 3.6. Adjusted Model A, age and sex covariates in the regression with weighted proportion estimates (direct estimates). Model A Model A2 Model A3 0.00002676 (<0.01) Altitude (m) Age (mean) Women (%) Intercept 0.25952414 (<0.01) 223 0.09049599 Observations R-squared P-values are in parentheses. 0.00002853 (<0.01) -0.00354297 (<0.01) 0.40854368 (<0.01) 223 0.11890332 0.00002803 (<0.01) -0.00341295 (<0.05) 0.09542722 (<0.1) 0.35551325 (<0.01) 223 0.12654629 Table 3.7. Adjusted Model C, age and sex covariates in the regression with EBLUP1 estimates. Model C Model C3 Model C2 0.00002583 (<0.01) 0.00002708 (<0.01) -0.00248675 (<0.05) Altitude (m) Age (mean) Women (%) Intercept 0.22240446 (<0.01) 223 0.13061499 0.32699872 (<0.01) 223 0.15229181 Observations R-squared P-values are in parentheses. 1 EBLUP – Empirical best linear unbiased predictor. 0.00002683 (<0.01) -0.00242422 (<0.05) 0.04589315 (≥0.01) 0.30149517 (<0.01) 223 0.15502991 46 Depressed Mood across Altitude Quantiles Figure 3.3 and Figure 3.4 show density distributions of the prevalence proportions of DM in the total population and across quantiles (i.e., terciles and deciles) using the direct and EBLUP estimates. The exact altitude values that divide the respective tertiles were 204 and 2682 meters. The altitude values for quintiles were 120, 342, 2500, and 3047 meters. Finally, the altitude values that divide the deciles are 67, 120,162, 342, 1921, 2500, 2731, 3047, and 3625 meters. Figure 3.3. Estimated kernel density distributions of DM prevalence across quantiles. Direct estimates (weighted proportions of DM). Figure 3.4. Estimated kernel density distributions of DM prevalence across quantiles. EBLUP estimates. Finally, the regression estimates across altitude quantile levels were estimated using the direct and EBLUP estimates without transformations. It was found consistently that estimates of depressed mood (DM) are increasingly higher when participants are living at higher altitude levels. Tables 3.8 and Table 3.9. show linear regressions using the direct estimates (i.e., 47 weighted proportions) and EBLUPs for three and ten quantiles, respectively. The ‘age mean’ and the proportion of women in the districts were not associated with DM at this level of analysis. Interestingly, the association between altitude and DM switches to a significance level at the fifth quantile when looking across ten quantiles. Similar results are obtained when I compare with the FH model excluding the forty districts with null values, confirming this association. With this last model, the significance level starts at the fourth quantile (see Figure 3.9 in Appendix B). 48 49 Model Four from Table 3.9 examines the regression of altitude deciles on the EBLUP estimates without covariates, and the residual distribution from this model is shown in Figure 3.5. Figure 3.5. Residual distribution from Model 4 in Table 3.9. In a post-estimation procedure, Figure 3.6 shows the predicted values for DM prevalence across deciles of altitude using model four in Table 3.9. The predicted prevalence proportion of DM increases with altitude levels above the sea. 50 Figure 3.6. Predicted prevalence proportion of Depressed Mood across altitude deciles, using Model 4 in Table 3.9. 3.4. Discussion The main finding of this study is that districts located at higher altitude levels tend to have larger estimated Depressed Mood (DM) prevalence proportions, based upon three different inter-related estimators. Whereas accurate estimation of DM prevalence at the district level might be the case for government agencies, for this project I used the Fay Herriot model to improve the precision of slope estimates linking altitude-DM and to improve the fit of the subsequent regression models. Before discussing the results in detail, it is important to acknowledge several study limitations. The surveys were conducted over ten years, but in this study, I did not thoroughly evaluate whether there is a significant variation in the likelihood of living at higher or lower altitudes over the years. Nor did I assess whether there is substantial variability in depression over time. There could be a variation in the incidence or in the duration of depressive symptoms 51 over the years, and if there are differences in residential altitude levels over time, it could affect the study's conclusions. However, as can be seen in Figure 3.10. (see Appendix B), there is no verification of a trend or variation in altitude levels over the years. Another relevant limitation of this study is that the analyzed surveys were not designed to enable statistical inference at the district level. Although a Small Area Estimation (SAE) methodology was used to improve estimates with the Fay-Herriot model, this does not guarantee that estimates are as precise as if we were to survey at the district level to determine prevalence more accurately. There are currently new SAE methodologies to ensure better precision of BLUPs. Still, these new methodologies require more resources and are usually reserved for situations where the primary goal is to determine precise estimators that guide government agencies' decision-making (e.g., determining extreme poverty rates to allocate government funds). Regarding the Fay-Herriot model used in this study, one limitation was that 40 districts were found to have no usable standard error values for variance calculation. One possible option was to combine districts with very little sample or no cases of depression with neighboring communities with similar altitudes. This approach could allow the use of this data without imputing values in standard errors. This possibility could have the disadvantage of artificially combining populations and introducing noise into the analyses. It would also require greater depth in the geographical recognition of the districts. It is also important to highlight a limitation in this study related to assessing depressed mood. Although there are validated 2-item questionnaires for depression screening and mood state evaluation (86,87), I combined two questions that do not yet have psychometric validations for depressed mood evaluation. Also, the “prevalent mood states” part of the 52 questionnaire is not validated in other languages than Spanish. However, I believe that when a person says that they experience sadness "always or almost always" and that sadness is experienced "frequently during the last four weeks," it captures the essential concept of “depressed mood,” according to the European psychopathological tradition (41,45,46) and the definition of the American classification for depressed mood (44). On the other hand, this study did not evaluate other variables that may be relevant to the altitude-depression relationship. In this sense, it is interesting to note that populations living at higher altitudes tend to be made up of predominantly native people. The implications of this relationship are intriguing. Other variables such as poverty or educational level, among others, are also relevant to study, but considering that they could be rather mediators between the altitude-depression relationship. This type of analysis goes beyond the objective of this specific dissertation research project aim. In addition to the mentioned limitations, a novel idea was unveiled during the dissertation committee presentation that would enable me to assess the relationship between altitude and depressed mood in a dose-response manner using other models with inverse variance weighting estimation. I want to acknowledge this limitation in my dissertation, and I hope to complete the analyses before publication of articles based on my dissertation findings. Despite the mentioned limitations, the study findings are of interest. To my knowledge, this is the second published ecologic study addressing the relationship between altitude and depression, and it qualifies as one of a few such studies completed outside of the United States. Studies at the community level of analysis are developed less frequently and usually require useful information from a broader range of populations. Compared to the DelMastro study (5), 53 altitude levels are higher in this research, and the relationship appears to increase correspondently. As such the results might have implications for public policies concerning the mental health care of high-altitude populations. Regardless of other potential causal factors associated with the higher likelihood of depressive symptoms at altitude, healthcare services need to consider these findings. Many high-altitude populations may also have limited access to health services, especially in Peru and other Andean countries. Much of this population may require mental health care that is underdiagnosed and undertreated. Although it is still very preliminary, the relationship between altitude and DM prevalence found in this study could lead us to consider the possibility of an asymptotic relationship between both. The regression coefficient values and the curve of predicted values across altitude deciles show a steeper increase in intermediate deciles and a less pronounced increase after that. While this could have many other explanations (e.g., cultural aspects related to the highest altitudes, or genetic differences of native populations, etc.) that I currently do not expect to discuss deeper, it is also possible that it follows the curve of “effective oxygen pressure” across altitude levels, which is related to the atmospheric pressure and is asymptotically described. Furthermore, this study is relevant because, in an exploratory manner, it introduces the SAE methodology in evaluating factors associated with depression. To our knowledge, there are very few studies that use SAE methodology to estimate mental health prevalence (88–92). Still, none of them conducted a more analytical approach to factors associated with depression. The findings of this study raise relevant topics that need further investigation. Native populations can develop changes in their physiology and build adaptation processes to high 54 altitudes, which may, in turn, influence the likelihood of experiencing depressive symptoms. In this first study for my dissertation project, variations across ‘diversity’ subgroups could not be addressed. As mentioned earlier, in Peru, there is a relationship between belonging to native ethnic groups such as Quechuas or Aymaras and a greater likelihood of living at high altitudes. Although reports have found higher levels of depressive symptoms at higher altitudes within the same native group (93), there is also evidence that native populations have developed adaptive mechanisms over thousands of years (37) that could somehow mitigate the negative effect of hypoxia on mood. Peru is a good scenario to study this aspect since, in addition to the native population at high altitudes, there is also a varied diversity of migrants to the high Andean regions with different types of ancestry. Related to the above is the topic of migration and the length of time spent or residing in a high-altitude area. Although some reports have followed migrants for up to a year (57) and found increases in depressive levels, further studies are needed that evaluate more extended periods of residency or staying at high altitudes. It is critical to elucidate whether depressive symptoms increase over time as with Chronic Mountain Sickness or whether they remain stable or decrease as the years pass. No studies have evaluated this variation. Individual-level studies are also necessary to estimate the role of other potentially intervening variables such as poverty, education, religiosity, social support, etc. However, these assessments must be cautious, as some of these variables could mediate the effect of altitude on depression. Some studies demonstrate neurocognitive effects due to exposure to hypoxia (52,55,58,59,94). 55 Another relevant topic to study is the effect of traditional coca leaf consumption on the mood of people living at high altitudes. In Peru, coca leaf consumption is related to cultural factors of the native population and living in the high Andean regions. The energizing effect of the coca leaf is well recognized because it contains up to 14 different alkaloids, including cocaine. Finally, it is essential to mention the need to explore the possible association between altitude and other mental disorders, such as anxiety disorders or post-traumatic stress disorder. Living at high altitudes is already a stressor and may confer vulnerability to these problems. 56 CHAPTER 4: ESTIMATING THE ALTITUDE-DEPRESSED MOOD ASSOCIATION AT THE INDIVIDUAL 4.1. Introduction LEVEL In the previous report from this series, I discussed the hypoxia theory and its potential connection to depression in higher altitudes. Under Study 1, I estimated the degree to which the odds of active depressed mood (persistent sadness) might vary across altitude levels in Peru, including the Peruvian Andes highlands, with the community district as the unit of analysis. I estimated prevalence proportions for each community district and studied the relationship in what might be described as an "ecological" analysis. As explained below, this investigation extends the evidence base beyond what can be learned in the multiple prior United States studies (4–6,8,9). It differs from studies in Nepal (65), Tibet (72), and recently published Peru studies (66,67) because the case definition for this study is focused more tightly on persistent sadness or depressed mood (DM) during the four-week period before the survey assessment. In this new report, with the individual as the unit of analysis (in contrast with the Aim1 focus on estimates with the district as the unit of analysis), I present estimates of the association between altitude and DM at the individual level. To be clear, my first paper in this series focused on community-level prevalence proportions with the district as the unit of analysis. Instead, in this paper, I estimate the relative odds of DM being present always or almost always within that four-week interval before assessment, as the odds might vary across altitude levels of the community district in which the participants reside, across the wide range of altitude levels in Peru. 57 Apart from the estimates for the United States, which are limited regarding altitude levels, as explained in my previous report, only three large epidemiological studies found in the published literature examine the relationship between altitude and depression with the inclusion of especially high-altitude communities. These prior studies were conducted in Nepal (65) and Peru (66,67). The Peruvian studies were published in 2022. However, the case definitions for all these prior studies differ from the one I specified in my dissertation research. The study in Nepal investigated a score based on a multi-item assessment of depressive symptoms in the week interval before assessment. The studies in Peru used the Demographic and Family Health Survey (ENDES in Spanish) data, with a focus on (a) the score of depressive symptoms during the two-week interval before assessment, as elicited in a multi-item scale and (b) the imposition of a threshold cut point on that summary score. In the present study, the outcome variable is depressed mood, as sadness persists "always" or "almost always" during the four-week interval before the assessment. This case definition of depressed mood (DM) considers a longer and more relevant period. This dissertation research project's approach is more in line with the concept of "depressed mood" in terms of affective psychopathology (41,45). DM is one of the core criteria in current psychiatric diagnostic classifications (43,44). As such, the estimates based on individual unit-of-analysis estimates for altitude-DM association presented in this study represent a step forward beyond previous research. As noted in the final sections of this paper, as the field continues to gain insight into the relationship between altitude and depression, it should become possible to derive more definitive evidence 58 from longitudinal studies as well as randomized experimental trials that can be devised to eliminate some of the constraints described in the Discussion section. 4.2. Materials and Methods Nine consecutive annual epidemiological surveys on mental health conducted in Peru between 2003 and 2013 by the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health were included in this study. The surveys (i.e., "Noguchi Surveys") employed nearly identical methodology and questionnaires, covering various factors such as sociodemographic variables, mental health problems, substance use, and protective factors for mental health, among others (95,75,78). The surveys were conducted across 33 provinces within 15 regions of Peru, spanning a diverse range of altitudes above sea level. These household surveys aimed to assess four distinctive groups of non-institutionalized community residents: adolescents, adults (aged 18 years or older, including elders), elderly individuals, and married or cohabitant women. Only household residents were considered for this series of surveys. For this research, I analyzed data only from the adult population modules. The Peruvian National Institute of Statistics and Informatics (i.e., INEI) provided the census clusters for each surveyed city or rural area. INEI used the latest available census survey data of the Peruvian residents to make the census clusters. It is important to note that this study focused on the population residing in Peru, precisely their place of residence. In the "Noguchi Surveys," the survey sampling procedures were carried out in a total of 22 urban (i.e., cities) or rural areas (usually located near previously surveyed cities) across 15 regions of the country. The total number of people considered for the sampling procedures in all these areas was over 9.8 million. 59 Figure 4.1. Regions Surveyed (n=15). From the “Noguchi Surveys,”1 Peru, 2003-2013. 1 “Noguchi Surveys”: epidemiological surveys from the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health. 60 The "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health researchers employed a homogeneous sample selection process for each surveyed area, whether urban or rural. They used a multistage random sampling process consisting of three stages. In the first stage, a group of census clusters was randomly selected as the primary sampling units within each stratum, which referred to the surveyed city or rural area. Each census cluster comprised approximately 80 dwelling units. Then, field workers elaborated a sample frame from the entire cluster. The dwelling units within each census cluster were randomly chosen in the second stage. In the final stage, the Kish table approach randomly selected individuals from the target population within each dwelling unit. For this, the field staff collected a list of residents in each dwelling unit (77). Participation levels (‘response rates’) for each of the surveys were reported by the "Honorio Delgado – Hideyo Noguchi" National Institute of Mental Health of Peru." Values varied from 81.7% to 98.4%, most above 90%. The number of participants included in each yearly survey ranged from 2331 to 6555. In some years, more than one region was surveyed. The total resulting sample consists of 33130 participants. However, 50 participants had missing responses to crucial study variables and were excluded from this report. The total sample included in this study was 33,080 participants. IRB Considerations The "Noguchi Surveys" underwent review and approval by the Peruvian National Institute of Mental Health Institutional Review Board to ensure the protection of human subjects in research. Before conducting the interviews, all participants provided informed 61 consent by signing a form. Participation in the study was voluntary, and participants could decline to answer questions or end the interview at any time if they felt uncomfortable. Furthermore, the Michigan State University Institutional Review Board for Protection of Human Subjects in Research also reviewed the research protocol. Given no human subjects contact and de-identified data in this project, the IRB determined that my dissertation project did not involve human subjects according to the US National Institutes of Health definition. Instruments and measurement The survey administered to adult participants consisted of 233 questions divided into five sections: sociodemographic, general adult health, clinical syndromes part A, clinical syndromes part B, and access to health services. Appendix A of this dissertation provides an English translation of the standardized survey questions used to evaluate the mood state of each participant during the previous four weeks. The questions were posed to participants in their native language. The altitude of households was determined at the district level using data from the Peruvian National Center for Strategic Planning (CEPLAN) (79). The survey used two questions to assess actively depressed mood within the last four weeks. The first question involved a Likert question that asked participants directly about their mood: "How often do you feel Sad?" This question was included in the "prevalent mood states" section of the questionnaire as Question #6. The second question involved a screening tool called SRQ-17, a shorter version of the SRQ-20 (80). This tool was used to inquire about depressive and anxiety symptoms experienced by participants in the past four weeks. One of the questions in the SRQ-17 asked if the participant had frequently felt sad during the past four weeks (Question 26). In addition to 62 questions about sadness, the SRQ-17 also asked about physical symptoms, appetite, anxiety, and other non-depressed mood-related issues. In terms of affective psychopathology, the concept of "mood" refers to an individual's baseline affective state, which tends to be enduring and stable over time(41,45). DSM-5 states a depressed mood is "a sustained feeling of sadness or hopelessness" (44). This definition implies that the subjective feeling of sadness is present most of the time rather than just for a few days or a week. In this study, the actively depressed mood was identified in individuals who reported feeling sadness "always" or "almost always" in response to Question #6 and also answered "Yes" to the question "Have you felt sad frequently in the last four weeks?" (Appendix A). Depressed mood is the most essential symptom of Major Depressive Disorder (MDD) but can also be a symptom of other mental health disorders like dysthymia. There was another set of depression questions. The Mini-International Neuropsychiatric Interview - MINI (81), which includes questions that match the criteria of ICD-10 for MDD, was also applied in the surveys. However, these questions about MDD were asked over the participant's lifetime and not directly about the last weeks before the interview assessment. Data analysis I used Stata 17© software (96) for the analyses. The models were managed to have individual-level estimates, considering the complex survey data characteristics described in Paper 1 (i.e., strata, census clusters, and analysis weights). Stata© survey commands were used to conduct bivariate and covariate-adjusted analyses based on the generalized linear model with a logit link function ('logistic regression') for complex survey data. Taylor series linearization (TSL) methods have been used to yield approximations of variances, from which standard 63 errors, 95% Fisherian confidence intervals, and p-values have been derived. I relied on a two- tailed alpha equal to 0.05, with p<0.05, when I drew attention to contrasts between subgroup estimates in the Results section. To be clear, when I estimated the relationship between the outcome variable (i.e., DM as manifest in persistent sadness) and the covariates, I used logistic regression models with 95% confidence intervals for crude and adjusted models. Collinearity diagnostics on the covariates were performed using the "Collin" command in Stata ©. My initial plan for hypothesis testing for nested models employed the "Nestreg" command that reports comparative goodness of fit (GoF) statistics. These GoF statistics provide evidence in favor of the null hypothesis of no improvement of fit when new terms are added, or against that null hypothesis. During my dissertation work, I learned that the ‘nestreg’ approach is no longer deemed appropriate when the context is complex survey samples and the response variable is binary as is the case for studying the odds of being in a state of active depressed mood. For this reason, I am providing the ‘nestreg' output for this chapter of my dissertation, but for inference I rely more heavily upon the ratio of the estimated slopes and their respective standard errors derived from Taylor series linearization. In this way, the ratios provide a way of learning whether the residual variation in the response variable depends upon the model’s newly added terms. The altitude variable was categorized in terms of tertiles for the primary analysis, in conformity with the approved specific aims for the dissertation research protocol. However, I show estimates in Appendix C based on other pertinent altitude quantiles (e.g., quintiles and 64 deciles). Appendix C also shows estimates using different altitude categories defined by other conventions. In the covariate-adjustment process, other predefined demographic covariates were introduced in the models. These covariates were age, sex, mother tongue (an indicator of ethnicity), and civil status. Finally, the time of residence was introduced in the models (i.e., more than two years). Then, a stratified analysis was performed to estimate the degree to which odds of depressed mood (DM) might vary across subgroups defined by duration of residence in the community district. This exploratory analysis addresses an issue not previously covered – namely, whether the period of residence in a high-altitude community district might have something to do with the experience of depressive symptoms. For this analysis, I pre-specified the cutout point based on whether the community participant had lived in the community district for at least two years before the assessment date. While it might seem that the estimates presented in the previous study and those in this second study are duplicative because they are conducted in the same populations, I believe they are actually complementary studies and not necessarily repetitive. This second study intends to obtain adjusted models while assessing the effects of other individual variables that could not be incorporated into a community-level approach. In other words, in this second study, we can delve deeper into individual-level variables that are not adequately addressed with aggregated data. For example, in this second study and in the upcoming one in this series, which is also designed for individual-level analysis, we can evaluate the relationship between marital status, ancestry, or coca leaf consumption and the presence of depressed mood. This could not have been optimally assessed with aggregated data. In this final version of the 65 dissertation, I present these two approaches (i.e., aggregated or community vs. individual level of analysis) so that readers can evaluate and compare both and decide which one might be preferable under what circumstances. 4.3. Results Based on unweighted sample data, 3729 participants had experienced persistent sadness (DM) during the four-week interval before assessment. The analysis-weighted prevalence proportion indicates that roughly 8% of the population met the DM case definition. Table 4.1 shows how the analysis-weighted prevalence proportions vary across relevant subgroups defined by the measured characteristics under study. We can see that weighted DM prevalence proportions increase from seven to 11.5% across the three altitude levels. The lowest altitude level has a significantly lower prevalence of DM compared to the other two higher levels. The age mean is higher in the group of people who experienced DM. The female population has much higher prevalence proportions of DM than males (i.e., around 12% vs. 4%). About ancestry, Quechua/Aymara participants have much higher levels of depressed mood DM. Regarding civil status, divorced and widowed participants also have a much higher proportion of DM. 66 Table 4.1. Depressed Mood (DM) relative to specified population subgroup membership of individual participants in the study sample. Data from the “Noguchi Surveys,”1 Peru, 2013- 2023. Depressed Mood (DM) NO Weighted Proportion (%) 91.9 CI YES n Weighted Proportion (%) 8.1 CI [7.5,8.8] [91.2,92.4] 3729 92.8 90.6 88.5 40.8 [91.9,93.5] 1111 [89.4,91.6] 1167 [87.6,89.3] 1451 [40.3, 41.3] 7.2 9.5 11.5 45.5 [6.5,8.0] [8.4,10.7] [10.6,12.4] [44.0,46.9] n 29351 10001 9821 9529 Total Altitude Tertile 1st 2nd 3rd Age (mean) Sex Male 13585 Female 15766 96.0 87.9 739 [95.3,96.6] [86.9,88.9] 2990 3.9 12.1 [3.4,4.7] [11.1,13.1] Mother tongue Spanish 24308 4702 92.4 86.3 [91.7,92.9] 2765 930 [83.9,88.4] 7.6 13.7 [7.0,8.3] [11.6,16.1] 292 95.2 [81.1,98.9] 29 4.8 [1.1,18.9] Married 17938 4567 91.8 86.0 [90.9,92.6] 2103 [84.1,87.6] 1139 8.1 14.0 [7.4,9.1] [12.3,15.1] 6835 95.1 [94.1,95.8] 486 4.9 [4.1,5.9] Quechua /Aymara Other Civil status Divorced/ Widowed Single Living more >2 years2 p<0.0001 p<0.0001 p<0.0001 p<0.0001 p= 0.056 No 1083 Yes 28268 87.5 91.9 [80.9,92.0] 122 [91.3,92.5] 3607 12.5 8.0 [7.9,19.1] [7.4,8.7] 1 “Noguchi Surveys”: epidemiological surveys from the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health. 2: Two or more years of residence here 67 Table 4.2. Estimated associations from bivariate analyses with depressed mood (DM) expressed as a function of each covariate, one by one. Data from the “Noguchi Surveys,”1 Peru, 2013-2023. Variable Altitude Tertiles 2nd 3rd Intercept Age Age Intercept Sex Female Intercept Mother Tongue (Ref. Spanish) Quechua/Aymara Other (Amazonia) Intercept Civil status (Ref. Married/Cohab) divorce/widow single Intercept Coef. St.Err. t-value p-value [95% Conf. Interval] 0.294 0.509 -2.553 0.0898 0.0734 0.0584 3.27 6.94 -43.70 0.001 <0.001 <0.001 0.118 0.365 -2.668 0.467 0.653 -2.439 .0145 -3.049 0.0023 0.1081 6.25 -28.21 <0.001 <0.001 0.010 -3.261 0.019 -2.837 1.201 -3.187 0.0973 0.0853 12.35 -37.39 <0.001 <0.001 1.010 -3.35 1.391 -3.020 0.652 -0.484 -2.495 0.1072 0.7781 0.0457 6.08 -0.62 -54.59 <0.001 .53401 <0.001 0.442 -2.009 -2.585 0.862 1.042 -2.406 0.607 -0.536 -2.420 0.0932 0.1152 0.0576 6.52 -4.65 -42.06 <0.001 <0.001 <0.001 0.424 -0.762 -2.533 0.789 -0.31 -2.308 Two or more years living in the area 0.018 Yes Intercept -1.445 1 “Noguchi Surveys”: epidemiological surveys from the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health. 0.2599 0.2561 .05871 <0.001 -1.001 -2.449 -0.491 -1.947 -1.89 -7.60 Table 4.2 estimates are for bivariate associations estimated with the GLM and the logit link function using Stata© "Svy" commands that yield Taylor series linearization (TSL) approximations of variances. With no covariate adjustments, they are presented to provide the reader with information about these associations before any additional covariate adjustments. 68 In these bivariate analyses, the association between all the previously described demographic variables and DM was confirmed in the logistic regression framework for survey data. Table 4.3 shows estimates from covariate-adjusted models. The key finding, relative to the study aim, is that in this analysis with individuals as the unit of analysis, there is a statistically robust association that links the altitude of an individual resident of a community district in Peru with the odds of experiencing a recent depressed mood (DM), even with covariate adjustments as shown in Table 4.3. In this depiction of the results, different columns show the estimates for other models, adjusting consecutively the covariates. Estimated odds ratios and estimated p-values based on the TSL approximation for the variances and standard errors are shown. The strength of this association tends to be modest but is consistently confirmed across all the covariate adjustment models. 69 Table 4.3. Estimated associations from covariate-adjusted analyses with depressed mood (DM) expressed as a function of each covariate. Data from the "Noguchi Surveys,"1 Peru, 2003-2013. Model 3 Model 5 Model 6 Model 4 Model 2 Model 1 Un-adjusted Altitude Tertiles 2nd altitude tertile 3rd altitude tertile Age Female OR p-value OR p- value OR p- value OR p-value OR p- value OR p-value 1.34 1.66 0.001 1.39 <0.001 1.76 <0.001 <0.001 1.41 1.78 <0.001 <0.001 1.36 1.60 0.001 <0.001 1.36 1.61 0.001 <0.001 1.36 1.61 0.001 <0.001 1.02 <0.001 1.02 <0.001 1.01 <0.001 1.01 0.001 1.01 0.001 3.37 <0.001 3.36 <0.001 3.14 <0.001 3.15 <0.001 1.47 0.71 Mother tongue (reference: Spanish) Quechua/Aymara Other (Amazonia) Civil status (reference: married) divorce/widow single Two or more years living here Long residence Observations 1 "Noguchi Surveys": epidemiological surveys from the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health. 0.011 33013 0.001 0.026 0.002 0.024 0.002 0.626 0.001 0.633 0.002 0.59 1.38 0.76 1.37 0.75 1.44 0.67 1.44 0.70 33013 33080 33078 33024 33078 0.54 70 As promised in my dissertation proposal, I completed goodness of fit (GoF) comparisons. However, as described in the Methods section, these results are presented here but cannot be regarded as definitive evidence due to uncertainty about GoF evaluations when the context is that of a complex sample survey. Results are shown in Figure 4.2 (i.e., in Appendix C). Questions about multicollinearity might be of interest to some readers. My evaluation of multicollinearity suggested that my estimates were not distorted due to violations of assumptions about multicollinearity. Figure 4.3 in Appendix C shows the Stata© output of Model 6 from Table 4.3. Finally, when the relationship between altitude and depressed mood (DM) was estimated using stratification for the elapsed residence time in the community district, the association between altitude and DM was found only for those living there for two or more years. Table 4.4. Stratified analyses of beta estimates of altitude’s effect on depressed mood by length of residence in the district. Data from the “Noguchi Surveys,”1 Peru, 2003- 2013. 4.4. Discussion As for the main finding of this study, with the individual as the unit of analysis, I postulated a tangible association between altitude and the prevalence of active and persistent 71 depressed mood (DM). My resulting estimates show this association when altitude is sorted by tertiles and other quantiles, as shown in Table 4.5. (see Appendix C). Before discussing the results in detail, it is important to acknowledge some of the most relevant limitations of the study. Concerning the assessment of depressed mood, the recently active experience of persistent sadness (DM) was measured via two combined questions that have not yet been validated with psychometric studies. However, it is entirely plausible that when a person says that they experience sadness "always or almost always" and that sadness is experienced "frequently during the last four weeks," it captures the essential concept of the "persistent sadness" facet of a "depressive mood," according to the European psychopathological tradition (41,45,46) and the definitions used in the American classification for depressed mood (44). Other key limitations involve the cross-sectional character of the data with no specification of the age of onset of DM relative to the time of residence at the current altitude. In addition, this study did not evaluate other community characteristics that might be covarying with the altitude-depression relationship (e.g., exposure to trauma). However, careful consideration is required to sort these potential influences about a Directed Acyclic Graphing logic so that "ancestor" variables are included as covariates, but "mediators" and "descendants" are not included in the analyses (97). The association between altitude and depression I found in this study goes in line with the previous paper in this series with the district as the unit of analysis and with the other epidemiological studies that have evaluated shorter periods of depressive symptoms (i.e., one or two weeks). In this case, I examined a more extended period of persistent sadness or 72 depressed mood (DM) that has been regarded as a crucial criterion on the path to a Major Depressive Disorder (MDD) (44). As for prior studies, it might be of interest that the estimated prevalence of depression in the Tibetian Plateau was considerably higher (72) (i.e., 28% versus 8% in this study). However, this variation might be traced to different case definitions and time intervals. Please note that in the Tibetan research, the time interval evaluated was one week, and for this dissertation, the time interval was four weeks. In Nepal, another nationwide study found around 10% prevalence of depression. They also found that living at high altitudes (i.e., 2000 m) was associated with depression (65). As for the other two studies in Peru from the ENDES surveys (66,67), these studies also showed an altitude-depression association. However, the ENDES studies did not disclose any association with the mother tongue (Quechua or Aymara), and the result is that the mother tongue association deserves greater scrutiny, as described in a later section on future directions for research. To the best of my knowledge, this work represents the first study evaluating the time of residence concerning the association between altitude and depressive symptoms (DM). Estimates from the duration-of-residence stratified analyses prompt a speculative hypothesis that the effect of altitude on people's mood might be more salient when at least two years of exposure have passed. It has been reported that adverse cognitive effects at high altitude levels fluctuate over time. Initially marked deterioration, a period of acclimatization follows, and finally, in some cases, worsening can be seen again (59). This pattern of deterioration- acclimatization-deterioration was described previously and played a central role in Chronic 73 Mountain Sickness (30). Maybe something similar might happen with persistent sadness, with initial acute sadness forming and then fluctuating but persisting as residence in higher altitudes increases in duration. Potential implications for public health deserve mention. With over 11% of the population, particularly those in higher altitudes, potentially experiencing a persistent state of depressed mood, this could have repercussions on both physical and mental health (98). Furthermore, it could lead to adverse economic and social consequences in highland populations (99,100). Health policymakers should remain vigilant in developing more preventive, early diagnostic, and intervention measures related to depression, especially for high-altitude residents. These findings are also interesting because recent data from a new Peruvian National Institute of Mental Health survey will soon be available to evaluate and replicate what was presented here. The relationship between altitude and depression should be studied more in- depth with more and larger data. Studies on migration, their different patterns, and their relationship with mood changes are also needed. Furthermore, if these findings are confirmed, we must develop and design future preventive interventions and incorporate the altitude variable in managing patients with depression in the highlands. If feasible, developing these epidemiological, preventive, and treatment-related studies should enhance the public health significance of this dissertation research project. In the final chapter of my dissertation research report, I will provide some additional ideas about future directions for research, with coverage of both treatment intervention studies 74 and preventive trials. Suppose hypoxia is one of the determining influences in this relationship. In that case, there are some motivating opportunities for randomized controlled trials to evaluate how oxygen supplementation might help alleviate depression once it starts. In some instances, the relationship might guide preventive interventions. This topic will be covered in the final chapter of my dissertation research report. 75 CHAPTER 5: THE ESTIMATED EFFECT OF COCA-LEAF USE IN THE ASSOCIATION BETWEEN ALTITUDE AND DEPRESSED MOOD 5.1. Introduction I evaluated the relationship between altitude and mood in the previous two studies of this series. In the first study, with ecological units of analysis, districts located at higher altitudes had more residents with depressed mood (i.e., larger prevalence proportions). The second study found this association at the individual level while controlling for relevant covariates. This third investigation will focus on the possible role of coca leaf consumption in the relationship between residence altitude and mood state. As we will see, in the Peruvian Andes, not only is the issue of environmental stress caused by hypobaric hypoxia relevant (25,31,32,37,101), but traditional coca leaf consumption is also a prevalent ancestral practice (10,11,102,103) that must be considered and studied. In the South American Andes, coca leaf has been consumed for millennia. Traces of evidence dating back 3000 years have been found to support this assertion (104). There is also evidence that coca has been cultivated in the central region of Peru for almost 2000 years (105). Its use seems quite widespread in the Andean South American region from that time. Even before the Inca era, its use was ceremonial and religious. During the Inca period, the coca leaf was used in special ceremonies, was associated with the highest royalty (105), and was considered the "most sacred plant of all" (106). With the arrival of the Spanish (i.e., in the 16th century) and after their unsuccessful attempts to eradicate its use, coca leaf chewing apparently became more widespread. A custom of providing daily coca leaves to workers in the fields and mines emerged (102). Since then, the 76 use of coca leaves has been sustained (107), and the practice of taking coca leaf rations during work tasks remains in the high Andean areas. Since colonial times, it was recognized that coca leaves could better withstand mining and farming tasks. Due to this, coca leaves were highly valued, and even taxes and payments to farmers were made using them (105). Elsewhere in the world, interest in the coca leaf surfaced in the 19th century. In 1863, in Paris, Angelo Mariani developed the "coca wine tonic" based on coca leaf, which became very popular in Europe. Mariani even received recognition from the Pope, and Thomas Edison was an enthusiastic consumer of this tonic (108). Research interest in coca leaves surfaced at about the same time (109). In 1859, it was already described in Europe that coca leaf had properties such as "reducing fatigue and improving mood" (102). Currently, the consumption of coca leaves is still quite widespread in the Andean region (10,11). Consumption mainly involves coca leaf chewing or coca tea preparations. Typically, using coca leaves involves more of a sucking action than actual chewing. Initially, a bundle of dried coca leaves is placed inside the mouth, between the gums and cheek, then slowly soaked with saliva. Then, to assist in separating the alkaloids of the leaf, including cocaine, a small amount of lime or a similar substance is inserted into the center of the coca bundle. Additionally, there can be a ceremonial aspect to this process (110). In some Andean communities, the consumption of coca leaves is seen as part of the cultural identity of Quechua and Aymara native populations (110). Advances in medicinal chemistry and pharmacognosy since the mid-19th century have disclosed at least 18 alkaloids in coca leaves (102). The 'cocaine' extraction has received most 77 attention. However, it is believed that these alkaloids can work synergistically and that their effects should not be solely attributed to the effect of cocaine. More studies are needed in this regard (111). Even though chewing coca leaves cannot be equated to consuming purified and crystallized cocaine (111,112), research has demonstrated that consuming coca leaves via chewing or as tea (111) can result in the rapid detection of cocaine in the body. For example, a study comparing traditional coca leaf users to non-users found that after one hour, levels of 70 ng/ml of cocaine were detected in the blood (113). Because coca-leaf consumption is more frequent in the Andean highlands, several decades ago, it was hypothesized that the consumption of coca leaves may play a role in adapting to high altitudes (114). For example, it has been suggested that coca products might better regulate the increase in red blood cell production (i.e., erythropoiesis) (112). A mild beneficial effect on body temperature management has also been described, which could be helpful in the adaptive process to high altitudes (115). Additionally, it has been proposed that coca leaf consumption could help maintain stable blood glucose levels and improve nutrient availability in the body in environments with scarce food (116). However, studies have not provided conclusive evidence in these areas. Some even argue that the association between altitude and coca leaf consumption is due to cultural factors rather than high altitude adaptation (117). In a study conducted with a limited number of participants, it was observed that regular consumers of coca leaves had comparable levels of adrenergic activity (i.e., epinephrine and norepinephrine in the bloodstream) relative to non-consumers. However, coca users exhibited 78 higher levels of fatty acids in their blood. Furthermore, during strenuous physical activity, coca users could tolerate lower oxygen saturation levels while maintaining similar ventilatory capacities (113). Another experiment involving non-regular coca users demonstrated that consuming coca leaves did not elicit significant changes at rest but did enhance oxygenation, cardiac function, and other related factors during exercise compared to those who did not consume coca leaves (118). I have not found any reports of controlled studies assessing the potential mood-altering effects of coca leaf consumption. Nevertheless, given that cocaine does have mood-altering effects (13–15,119), the consumption of coca leaves may have some potential impact on how consumers experience their moods. These background details help motivate the present study and its investigation of the possibility that coca product consumption, especially coca leaf chewing, might modulate the altitude-affectivity association described in earlier chapters of this dissertation. That is, the primary objective of this study is to evaluate whether coca leaf consumption has a modifying effect and might modulate the altitude-affectivity association described in this dissertation research project’s initial inquiries. A starting expectation might be that the recent use of coca leaf might attenuate an observed greater likelihood of an active depressed mood at the higher altitudes of residences in the Andean communities of Peru. Studied across tertiles (or other quantiles) of altitude, the recent users of coca products might experience a mood-elevating effect of the alkaloids in the coca leaf such that these coca product users would have lower prevalence of active depressed mood as compared with non-users of these products who are residing at the same altitude levels. 79 5.2. Materials and Methods For this study, as described in prior chapters of this dissertation report, I turned to data from the nine consecutive annual epidemiological surveys on mental health conducted in Peru between 2003 and 2013 by the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health. The surveys (i.e., "Noguchi Surveys”) were implemented across 15 regions and 33 provinces of Peru, utilizing almost identical methodology and questionnaires for each year. Several mental health-related factors were assessed, such as sociodemographic variables, mental health problems, substance use, etc. These household surveys were conducted to evaluate four distinct groups of non- institutionalized individuals within the community: adolescents, adults (18 years of age or older, including seniors), elderly individuals, and women who were married or cohabitating (75). Only household residents were included. Only data from the adult population modules were analyzed for this study series. The Peruvian National Institute of Statistics and Informatics provided the census clusters for the surveys using the latest available survey data. The surveys were focused on the population residing in Peru. Field workers from "Noguchi Institute" created sampling frames within each census cluster, from which sampled dwelling units were derived. In the "Noguchi Surveys,” the survey sampling procedures were carried out in 22 urban or rural areas across the country. The total number of people in the populations considered for the sampling procedures in all these areas was over 9.8 million. The "Noguchi Institute" researchers employed a multistage random sampling method that consisted of three stages. In the first stage, a set of census clusters was randomly chosen as the primary sampling units within each stratum, which referred to either the surveyed city or 80 rural area. Each census cluster was comprised of approximately 80 dwelling units. Next, field workers created a sample frame from the entire cluster. In the second stage, dwelling units were randomly selected from each census cluster. In the final stage, individuals from the target population were randomly chosen within each dwelling unit using the Kish table approach. For this, the field staff obtained a list of residents in each dwelling unit (75,77,78). The "Noguchi Institute" survey participation levels (‘response rates’) varied from 81.7% to 98.4% across the surveys. Most values exceeded 90%. The number of participants in each annual survey ranged from 2,331 to 6,555. In some years, more than one region was surveyed. The resulting overall sample comprised 33,130 participants. However, due to missing responses on crucial study variables, 101 participants were excluded from this report. Thus, the total sample included in this study was 33,029 participants. IRB Considerations The "Noguchi Surveys” were subjected to review and approval by the Peruvian National Institute of Mental Health Institutional Review Board to safeguard the rights and welfare of research participants. Before conducting interviews, all participants were required to provide informed consent by signing a form. Participation in the study was voluntary, and participants could decline to answer questions or terminate the interview at any point. In addition, the research protocol for dissertation research project with de-identified datasets was reviewed by the Michigan State University Institutional Review Board for the Protection of Human Subjects in Research. Because no human contact was required for this de-identified data analysis project, the Institutional Review Board (IRB) determined that the study did not fall under the category of human subjects research as outlined by the US National Institutes of Health. 81 Instruments and measurement The survey questionnaires given to adult participants comprised 233 questions, organized into five sections. The five sections included items on sociodemographic characteristics, general adult health, clinical syndromes part A, clinical syndromes part B, and access to health services. The dissertation's Appendix A contains an English translation of the standardized survey questions utilized to assess the mood of each participant over the previous four weeks. In this study, the actively depressed mood was identified in individuals who reported feeling sadness "always" or "almost always" in response to Question 6 and also answered "Yes" to the question "Have you felt sad frequently in the last four weeks?". A depressed mood (DM) is a manifestation of the ‘affectivity’ facet of the mental life of individuals as described in earlier chapters of this dissertation. Depressed mood qualifies as the essential symptom of Major Depressive Disorder (MDD) but can also be a symptom of other mental health disorders. Appendix A also contains an English translation of questions about alcohol and other drug use, including coca-leaf consumption. Whereas these questions assessed presence of abuse or dependence on each substance, for this study, the covariate was specified as any coca leaf use during the month prior to the survey assessment. The questions were posed to participants in their native language. Both groups of questions about mood and coca-leaf use were asked to all participants without gated or branching questions as might be embedded within questionnaire skip patterns. 82 The altitude of households was determined at the district level using data from the Peruvian National Center for Strategic Planning (CEPLAN) (79). The specific details have been described in prior chapters of this dissertation. Data analysis I used two complementary approaches to evaluate the effect of coca leaf use on the previously found altitude-mood relationship. As in prior chapters, the altitude variable was categorized in terms of tertiles for the primary analysis, in conformity with the approved specific aims for the dissertation research protocol. First, I conducted a stratified analysis, grouping individuals based on their coca leaf consumption. Then, after stratified analyses, I studied coca leaf users and non-users as separate subpopulations, searching for evidence of non-overlapping 95% confidence intervals for the estimated slopes that convey the altitude→depressed mood association. I then evaluated the possibility of subgroup variation or an interaction between altitude and coca leaf consumption using product terms. In this dissertation, subgroup variation in slope estimates does not count as an example of ‘interaction’ because the underlying mechanisms remain hidden from view. To be confident about any ‘interaction,’ the researcher must be able to specify the nature of the variation as can be true when studying a neurotransmitter agonist with effects blocked by an antagonist, or when a partial agonist’s effects are boosted by co-administration of some other drug or alteration of an extraneous variable under the control of the researcher (e.g., ambient temperature). For this reason, the term ‘interaction’ is used sparingly, and the pattern of evidence generally is described as one of ‘subgroup variation’ in the slope estimates. 83 For the just-described stratified analysis and assessment of the product terms, I used the final model from the second study in this series of papers (i.e., Chapter four of this dissertation), which includes altitude divided into tertiles. I then considered covariates mentioned previously: age, sex, marital status, mother tongue, and length of residence in the current location (i.e., whether they have lived in the place for two or more years) as covariates. The project’s estimation models were fit using Stata 17© software to account for the complex survey data characteristics described previously (i.e., strata, census clusters, and analysis weights). The initial analyses were from stratifications with estimation of point prevalence proportions and their standard errors via fairly simple cross-tabulations of altitude, coca consumption, and depressed mood for coca users versus non-users. Corresponding estimates from a general linear model for complex survey data were derived using an ‘identity’ link function and the standard regression model. Then, covariates were accommodated via my use of the generalized linear model with a logit link function. In this work, I utilized Stata survey commands to carry out bivariate and covariate-adjusted analyses on complex survey data. To obtain approximations of variances, I utilized the Taylor series linearization method and derived standard errors, 95% Fisherian confidence intervals, and p-values. My statistical analysis used a two-tailed alpha level of 0.05. When comparing subgroup variations, I specified alpha and p- values less than 0.05 as indicative of what traditionally has been termed “statistical significance.” For hypothesis testing on nested models, I used the "Nestreg" command, which provides F statistics for the comparison tests between the nested models. Collinearity between coca-leaf use, altitude, mother tongue, and other variables was evaluated using the “Collin” command. For example, the collinearity analysis produced variance inflation factor coefficients 84 below 1.1 and tolerance coefficients all were above 0.9, indicating that no further investigation was required. 5.3. Results Table 5.1 presents each covariate under study and its association with coca leaf consumption in Peru. Approximately two percent of the population consumed coca leaves the month before the surveys. However, this percentage increases from 0.5% in the lowest level to almost 12% in the group living at higher altitudes above the sea. Table 5.2 shows the estimates from stratified analyses with three tertiles of altitude crossed with the prevalence of active depressed mood for users of coca versus non-users. Table 5.2 shows clearly an unexpected finding about coca leaf consumption and being a case of actively depressed mood. Contrary to expectations, altitude level by level and gauged by the point prevalence estimates, it is the coca leaf users who are more likely to experience active depressed mood. The 95% confidence intervals (CI) overlap considerably and suggest that we should not make too much of this subgroup variation. Nevertheless, these simple stratified analyses provide an initial challenge to the idea that coca leaf use might be protecting the users from otherwise mood-affecting circumstances involved with living at the highest altitudes in Peruvian communities. 85 Table 5.1. Unweighted sample counts and weighted estimates that contrast coca leaf users versus non-users. Data from the “Noguchi surveys,”1 Peru, 2003-2013. 86 Table 5.2. Estimated prevalence proportions for active depressed mood, stratified by coca leaf use and by altitude. Data from the “Noguchi surveys,”1 Peru, 2003-2013. Tertiles of Altitude in Peru Middle 22.0% Recent Use of Coca Leaf Recent Use of Coca Leaf Lowest 9.4% 3.4%, 23.5% 7.2% 6.5%, 8.0% 1 “Noguchi Surveys”: epidemiological surveys from the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health. YES 95% CI NO 95% CI 7.9%, 10.1% 10.7% 9.8%, 11.7% 16.0%, 29.5% 14.5%, 20.1% Highest 17.1% 8.9% With ‘svy’ commands, and before covariates are introduced, it is possible to quantify the prevalence differences illustrated in these tables via a general linear model and ‘identity’ link function, and after converting the tertiles to the median elevation above sea level in kilometers. Based on this analysis with median km altitude as the covariate of interest, the occurrence of actively depressed mood among non-users of the coca leaf shows an increase in estimated DM prevalence of one percentage points for each km increase in altitude, with p-value less than 0.001. The corresponding slope estimate for coca users is not appreciably different from the null slope (p = 0.637). Evaluated via a product term for the total population (coca=yes * tertile as product term), the Wald statistic has a p-value of 0.133, well above the conventional alpha specifications for product-terms in epidemiological analyses. That is, the evidence does not suggest the anticipated modulation of the altitude-DM association as evaluated across the subgroups of recent coca users versus non-users. The results of the stratified analysis with covariate adjustment, in relation to coca leaf consumption, are presented in Tables 5.3 and 5.4. The first row of Table 5.3 illustrates the altitude-depressed mood association irrespective of coca leaf use, and the next two rows show corresponding estimates of the association for the subgroups of coca leaf users at each level of 87 attitude. If we were to study only non-users, the altitude-DM association would qualify as not appreciably distant from what is seen for all residents of these Peruvian communities. If we were to study coca leaf users only, there is an association with altitude (tertile levels) but there is no clear or robust association of altitude and depressed mood in the third row of estimates in Table 5.3. In Table 5.4, the stratified analysis shows the ORs using the group at the lowest altitude level who do not consume coca leaf as the reference group, following a format recommended by some authors (120). Please note that there are only 59 participants who consumed coca leaves in the last month and live in the lowest altitude level. Table 5.3. Stratified analyses of the estimated association between altitude (by tertiles) and mood, by coca-leaf use in the last month. Data from the “Noguchi Surveys”1, Peru, 2003-2013. 88 Table 5.4. Stratified analysis of the estimated association between altitude (by tertiles) and mood, by coca-leaf use in the last month, with one group as reference (lower altitude and coca nonusers). Data from the “Noguchi Surveys,”1 Peru 2003-2013. Altitude Tercile 1 Altitude Tercile 2 Altitude Tercile 3 DM Cases/ No cases 1,100/ 9,927 OR (p-value) 1 Coca-leaf Nonusers 95% CI DM Cases/ No cases 1,061/ 9,275 OR (p-value) 1.29 (p=0.006) 95% CI [1.08, 1.56] DM Cases/ No cases 1,158/ 8,142 OR (p-value) 1.52 (p<0.001) 95% CI [1.29, 1.80] 11/48 Coca-leaf Users Survey data analysis using generalized linear model with a logit link function, with covariate adjustment for: age, sex, maternal tongue, civil status, and time of residence (>2 years). DM: depressed mood. OR: odds ratio. 1 “Noguchi Surveys”: epidemiological surveys from “Honorio Delgado – Hideyo Noguchi” Peruvian National Institute of Mental Health. 2.10 (p=0.182) 3.31 (p<0.001) [0.71,6.27] 105/536 293/ 1373 2.43 [1.83, 3.23] [2.14, 5.13] (p<0.001) 89 Table 5.5 displays the stratified analysis using only two altitude levels divided by the median (i.e., 1921 meters). In this case, the sample size in the group of coca users at the lowest altitude level has increased by over 200 individuals (n=277). However, the stratified analysis still shows that the increase in depressed mood as altitude increases is tangible for the population that does not consume coca leaves (p-value<0.001). As in the prior analysis, among coca leaf consumers, the altitude-DM association is not robust. Table 5.5. Stratified analysis of the estimated association between altitude (two groups) and mood, by coca-leaf use in the last month, with one group as reference (lower altitude and coca nonusers). Data from the “Noguchi Surveys,”1 Peru 2003-2013. Table 5.5 adds covariate-adjusted evidence about coca users and depressed mood. If we specify the low altitude non-users as a reference group, the odds of active DM among high altitude non-users is 1.54 times greater. Note, however, that the corresponding odds of active DM among coca users at the high altitude is 3.24 times greater than the odds in that reference group (p = 0.001). Here again, the idea of a protective effect of coca use at higher altitudes is not supported by the evidence from this analysis. 5.4. Discussion The main finding of this part of my dissertation research project on the occurrence of active depressed mood (DM) in Peru is the absence of evidence that consumption of coca leaf 90 actually modifies the altitude-DM association. If anything, the coca users are more likely than non-users to experience active depressed mood, even at the higher altitudes. This discovery was an unexpected finding in this study but is in line with similar associations found with alcohol and other drugs as well (14,121–125). However, due to the study's cross-sectional nature, it is impossible to evaluate this association's temporality. To my knowledge, this association between coca leaf use and depressive symptoms was not previously reported. It is plausible that people, when experiencing depressive symptoms, may seek to consume coca leaf more frequently to alleviate symptoms such as fatigue or lethargy. This behavior could be due to the coca leaf’s energizing effects. However, a psychostimulant like coca leaf would only have a temporary impact and would not reduce the duration of depressive symptoms or affect the prevalence of depression. (Appendix D provides some additional results from post-estimation exploratory data analyses that help to confirm the results presented in this chapter.) The initial stratified analysis disclosed that in the subgroup of coca leaf consumers, the estimated effect of altitude on people's mood is not very substantial (Table 5.3), although a relationship was seen for non-coca-users. Regrettably, the few people who consume coca leaves and live in low-altitude areas limit the certainty of these results. As I consider limitations in my work, I must mention a measurement issue. It is possible that the measurement of depressed mood is not entirely equivalent across subgroups in Peruvian communities, possibly traced to genetic traits or to cultural variations. Measurement equivalence (126,127) studies are needed in this regard with the instruments used in this study but also with other questionnaires. 91 Notwithstanding my evidence, I wonder whether the potential role of the coca leaf as a mitigator of altitude effects (e.g., loss of energy or depressive symptoms) might remain worthy of consideration. This would align with ideas proposed for decades that suggest that the coca leaf is not only related to cultural factors but also could play a role in counteracting the effects of high altitudes (112,114,116). Let’s imagine that the prevalence of active depressed mood would be even higher among coca users in the higher altitudes than it now is found to be, and that the coca use dampens the DM prevalence from what otherwise would be a remarkably high prevalence proportion. This idea prompts consideration of a randomized trial to encourage highlanders to stop using coca and to see whether those who stop their use have lower levels of depression after a suitable washout interval. This is the kind of evidence that might be needed in order to clarify the altitude-DM association that might be modified by coca use. Some other study limitations deserve mention. Several have already been mentioned in this series's discussion sections of previous studies. For example, there is a possible measurement error of depressed mood with the questions used, whether this can vary among the population according to their ethnic origin or the chance that other relevant confounding variables were not included in the models. However, this time, I want to emphasize the cross- sectional nature of the present research, which does not allow for a more detailed evaluation of the temporality of the variables studied, which is of great importance in this case. Lastly, the findings of this study suggest lines of research that need to be developed to elucidate the role of coca leaf in depressive symptoms in the Andean population. Particular attention should be paid to evaluating these conditions in residents not belonging to the 92 Quechua and Aymara ethnic groups. They may more frequently show coca leaf consumption unrelated to cultural factors but rather to physical or mental symptoms. From a public health perspective, the management of patients in high-altitude areas may be influenced by the findings of this study. It will be necessary to evaluate whether it is required to implement preventive, diagnostic, and treatment measures for cases of depression at high altitudes in a different way than at sea level. For example, developing criteria to determine in which cases it is convenient to indicate a change of residence to lower places. Assessing whether the currently universally used criteria for depression need to be refined for the people living at high altitudes, etc. These are just a few examples of what could be proposed as a research agenda for the coming years. 93 CHAPTER 6: CONCLUSIONS AND FUTURE DIRECTIONS FOR RESEARCH Based on the results obtained in this dissertation, it can be stated that evidence has been found to support the association between altitude and depressed mood (DM). In this chapter, I will first summarize the main conclusions from the first two studies, which address this relationship. Then, I will discuss the findings related to the estimated effect of coca leaf use on the association between altitude and DM. Finally, I will propose suggestions for relevant lines of research based on these findings. 6.1. At the community level The first article of this series of studies showed that at the community level (i.e., district level), those localities located at higher altitudes also had higher prevalence proportions of depressed mood. It was found by evaluating the direct estimates of survey data analysis, using standardized estimates (i.e., with direct adjustment for age and sex), and finally using estimators corresponding to the Small Area Estimation methodology (i.e., Fay-Herriot model). These findings could have implications for public policies regarding the mental health care of highlanders. Given that many high-altitude populations, especially in Peru and other Andean countries, have limited access to health services, there may be a significant need for mental health care that is currently being underdiagnosed and undertreated. To my knowledge, the study presented in the chapter three of this dissertation is one of the very few epidemiological surveys conducted outside the United States to investigate the relationship between altitude and depression. Compared to the DelMastro study, this research examines higher altitude levels and finds that the relationship between altitude and depression 94 also increases at higher altitudes. It also is the first to consider coca leaf chewing as a potential modifier of the altitude-DM relationship. The broader range of exposure to hypobaric hypoxia suggests that the relationship between altitude and the prevalence of DM may follow an asymptotic pattern. The regression coefficient values and the predicted value curve across altitude deciles indicate a steeper increase in middle deciles, followed by a less pronounced increase. While this pattern might be attributed to various factors, it is also possible that it mirrors the curve of "effective oxygen pressure" across altitude levels, which is related to atmospheric pressure and can be asymptotically described. 6.2. At the individual level Upon analyzing the relationship between altitude and DM at the individual level, I added new evidence about the previously described association at the district level. That is, for individuals residing in higher altitude tertiles there is an association with a higher probability of experiencing DM. The prevalence proportion using weighted analysis indicates that approximately 8% of the population met the DM case definition. However, this proportion ranges from 7% to 11%, depending on the altitude tertile. The association between altitude and DM remains after covariate adjustment (i.e., age, sex, mother tongue, marital status, and length of residence). It also remains stable using different altitude categories, such as other quantiles (i.e., quintiles, deciles) or other altitude divisions. Another novelty in this study is examining the association between altitude and mood (DM) regarding the duration of residence. By stratifying the analysis according to the length of stay, the findings indicate a tentative hypothesis that the estimated effect of high altitude on a 95 person's mood may be more pronounced after a minimum of two years of residing at such heights. 6.3. Coca leaf and the Altitude-mood Association The third study investigated the consumption of coca leaf products. The results showed, for the first time, that active users of coca leaf are more likely to be experiencing active depressed mood (DM) states, even after adjusting for relevant covariates. However, the study's cross-sectional design makes it impossible to establish the direction of causal sequences in this relationship. That is, a prior mood state might have prompted the use of coca leaf products rather than vice versa. The role of coca-leaf use in the association between altitude and DM was evaluated via stratification and interaction terms (i.e., product terms). The stratified analysis suggests that among the coca leaf users, the estimated effect of altitude on DM is not as relevant compared to the non-users of coca. However, this preliminary evidence is not conclusive because the sample size is much smaller in the coca users' group. This finding could suggest a possible effect modifier role of coca use in the estimated effect of altitude on mood. The product terms analysis between altitude and coca leaf does not show evidence of subgroup variation between people who use coca and live at higher altitudes and the basal non-users of coca leaf. 6.4. Future directions for research Further investigation is necessary to explore the relationship between altitude and mood thoroughly. This research could have significant implications for the well-being of roughly 500 million people living in high-altitude areas(69). It may also shed light on the mechanisms that 96 underlie depression and related disorders. If some are viable, these ideas for future research should enhance the public health significance of this dissertation research project. I will categorize the research proposals outlined here into two main groups of studies. First, I will review the intervention and longitudinal studies necessary to complement the existing evidence on the possible causal effect of altitude on depressive symptoms. Second, I will address the research that can be done with available data from cross-sectional surveys carried out in Peru and other countries while longitudinal studies are being implemented. By combining these approaches, we can obtain a complete picture of the impact of altitude on mental health and develop effective strategies to address these problems. In the history of epidemiology, as with other suspected causes, cessation-of-effect studies could be seen as the first option concerning intervention studies. It is highly plausible to conduct randomized studies where a group of participants living at high altitudes is assigned to receive supplemental oxygen at different levels and modalities. Alternately, coca users might be randomly encouraged to stop such use, for an estimate of the effects on current depressed mood. Some studies suggest poor sleep quality is associated with hypobaric hypoxia and could trigger altitude-related cognitive and mood problems (55–58). I can envision the following groups in a randomized clinical study conducted in a high-altitude city, ideally with volunteers experiencing depressive symptoms and working sedentary jobs for most of the day: Group A: No supplemental oxygen; they continue their daily routine. Group B: Supplemental oxygen only during sleeping hours at night. 97 Group C: Supplemental oxygen during sleeping hours at night and intermittently during the day. Group D: Continuous supplemental oxygen during both daytime and nighttime. Longitudinal follow-up studies could also be performed on people who migrate from high-altitude areas to lower-altitude locations. These studies could also be done through randomized controlled trials (RCTs). For example, to facilitate their relocation to lower-altitude cities, economic or other incentives could be provided to participants. Cohort studies without intervention by the researchers could also be achieved. These longitudinal studies on migrants would ideally involve participants with depressive symptoms and close monitoring by healthcare personnel so that the possibility of medical treatment can be evaluated closely if necessary. Clinical treatment guidelines for Major Depressive Disorder (MDD) allow for delaying the initiation of medication when a patient has mild depression and is closely monitored by the therapeutic team (128). Cohort studies evaluating migration in the opposite direction are also necessary. These studies would follow individuals who migrate from lower-altitude cities to higher-altitude locations, usually for work-related reasons. Many questions surround this topic, and only small and short-term studies have been conducted so far(57). The finding reported in the second study of this series shows that the estimated effect of altitude on mood would be significant after at least two years. This suggests that follow-up studies of more than two years in migrants going to higher altitudes are necessary to measure a more substantial effect on mood. Overall, these new longitudinal studies will help to resolve uncertainties related to the association between altitude and depression, as well as their temporality, whether with 98 participants migrating from high to low altitude or vice versa. These studies will also be crucial because they will contribute to understanding the possible mechanisms involved in the association found in this series of studies. The association between coca leaf consumption and depressive symptoms is particularly intriguing. It is plausible that people with depressive symptoms such as fatigue or low mood seek the energizing effect of coca leaf to alleviate those symptoms. Cohort studies can adequately evaluate the temporal relationship by examining these variables' onset. The combined use of alcohol and coca leaf in the Andean population has been reported for decades (129,130). This combination leads to the appearance of Cocaethylene in the body (131,132). Cocaethylene is described as a substance that generates higher levels of dependency than alcohol or cocaine alone. Therefore, evaluating the interaction of these substances and whether it causes greater cognitive impairment or more significant mood disturbances is essential. To prevent depression, it is worth assessing the potential advantages of including the hypobaric hypoxia variable in preventive interventions. For example, research should evaluate the impact of moving individuals with genetic predispositions (e.g., those with genetic variants linked to severe depressive episodes or those with a family history of severe or refractory depression) to lower-altitude cities or using supplemental oxygen. Furthermore, these measures could be initiated as soon as mild depressive symptoms appear, such as at the start of a brief depression spell (e.g., 1-3 days of depressive symptoms) (133). The implementation of such measures could have a significant impact on public health. 99 Regarding depression treatment, research must explore methods to identify more specific criteria for incorporating hypobaric hypoxia in patient management. For example, a particular patient profile may require supplementary oxygen or a recommendation to descend to lower-altitude cities for better depression management, like how Chronic Mountain Sickness is currently managed. Defining a subtype of Altitude-related Depression is relevant. The specific study of what could be referred to as Altitude-related Depression could lead to the determination of the biochemical mechanisms that are specifically involved in this association. One published paper suggests that changes in tryptophan metabolism may be related to changes in mood at high altitudes (50). This highlights the need for studies that evaluate the effects of specific interventions with substances linked to these mechanisms, such as tryptophan supplementation in the diet. On the other hand, before or in parallel with conducting longitudinal or experimental studies, there is also a need for further research using population-based cross-sectional survey data. Numerous databases still require more analysis in this line of research. The "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health periodically runs its surveys. A recent study has been carried out throughout Peru that will provide more data for replicating and evaluating the results described in this dissertation. Extending the study to other populations will also be possible. New studies can be achieved using the current and recently collected data, for example, concerning the age of onset of depressive symptoms and the length of residence in high- altitude cities. Further research regarding migration and its effects on depressive symptoms is also needed and plausible with existing data. 100 There is also a need for a more in-depth study of the various forms of depression manifestations, such as Major Depressive Disorder (MDD) or dysthymia, and intermediate syndromes, such as a sustained depression spell (without MDD). By doing so, it may be possible to determine at what stage in the pathogenesis of MDD living at a higher altitude appears to have the most significant influence. It is also worth noting other surveys with data still to be analyzed. For example, the Demographic and Health Surveys (DHS) Program has been conducted in up to 90 different countries worldwide, many of which include questions addressing depressive symptoms and are easily accessible. For example, Peru (i.e., ENDES survey), Bolivia, and Colombia have DHS surveys in the Andean region. These and other national surveys from various countries that are currently freely accessible can be integrated to conduct analyses that cover multiple countries simultaneously. As such, the DHS surveys might make it possible to study residents in a range of high-altitude countries around the world. Another relevant aspect to explore with the available or soon-to-be-accessible data is the evaluation of models with latent variables. For example, models that allow for the assessment of Differential Item Functioning (DIF) between the native population and the rest of the people. Figure 6.1 shows a MIMIC (Multiple Indicators Multiple Causes) model (134–136) that evaluates the possibility that the Quechua or Aymara population may endorse the presence of depressive mood differently from the Major Depressive Episode (MDE) construct or latent variable. 101 Figure 6.1. Path diagram for MIMIC model. Differential Item Functioning on Depressed Mood. Note: The right column represents Major Depressive Episode criteria. Latent variable models can also aid in evaluating whether there is a greater prevalence of suicidal ideation or behavior at high altitudes, above what would be expected based on depressive symptoms alone. This would involve testing the estimated direct effect of altitude level on suicidal ideation and assessing whether this association varies across different altitude subgroups. Finally, it will be important to explore the potential link between altitude and other mental disorders, such as anxiety disorders or post-traumatic stress disorder. Living at high altitudes is already a stressor, which may increase susceptibility to these conditions. Therefore, further research is necessary not only on depression but also on other mental health problems. Elaborated in this fashion, epidemiological research can help guide public health tactics that ultimately should improve the health of our populations. 102 REFERENCES 1. Wear A. Place, Health, and Disease: The Airs, Waters, Places Tradition in Early Modern England and North America. Journal of Medieval and Early Modern Studies. 2008 Sep 1;38(3):443–65. 2. Unanue, Hipolito. Observaciones sobre el Clima de Lima y sus influencias en los Seres Organizados en Especial el Hombre. Imprenta Real de los Huerfanos; 1806. 136 p. 3. Twining W. Some Account of Cretinism, and the Institution for Its Cure on the Abendberg, near Interlachen, in Switzerland. Br Foreign Med Rev. 1844 Apr;17(34):514–6. 4. Kious BM, Bakian A, Zhao J, Mickey B, Guille C, Renshaw P, et al. Altitude and risk of depression and anxiety: findings from the intern health study. International Review of Psychiatry. 2019 Nov;31(7–8):637–45. 5. DelMastro K, Hellem T, Kim N, Kondo D, Sung YH, Renshaw PF. Incidence of major depressive episode correlates with elevation of substate region of residence. Journal of Affective Disorders. 2011 Mar;129(1–3):376–9. 6. Brenner B, Cheng D, Clark S, Camargo CA. Positive Association between Altitude and Suicide in 2584 U.S. Counties. High Altitude Medicine & Biology. 2011 Apr;12(1):31–5. 7. Reno E, Brown TL, Betz ME, Allen MH, Hoffecker L, Reitinger J, et al. Suicide and High Altitude: An Integrative Review. High Altitude Medicine & Biology. 2018 Jun;19(2):99–108. 8. Kious BM, Kondo DG, Renshaw PF. Living High and Feeling Low: Altitude, Suicide, and Depression. Harvard Review of Psychiatry. 2018 Mar;26(2):43–56. 9. Sabic H, Kious B, Boxer D, Fitzgerald C, Riley C, Scholl L, et al. Effect of Altitude on Veteran Suicide Rates. High Altitude Medicine & Biology. 2019 Jun;20(2):171–7. 10. Brutsaert T, Milotich M, Frisancho AR, Spielvogel H. Coca Chewing Among High-Altitude Natives - Work and Muscular Efficiencies of Nonhabitual Chewers. American Journal of Human Biology. 1995;7(5):607–16. 11. Biondich AS, Joslin JD. Coca: High Altitude Remedy of the Ancient Incas. Wilderness & Environmental Medicine. 2015 Dec;26(4):567–71. 12. Novák M, Salemink CA, Khan I. Biological activity of the alkaloids of Erythroxylum coca and Erythroxylum novogranatense. J Ethnopharmacol. 1984 May;10(3):261–74. 13. Helmus TC, Downey KK, Wang LM, Rhodes GL, Schuster CR. The relationship between self- reported cocaine withdrawal symptoms and history of depression. Addictive Behaviors. 2001 May;26(3):461–7. 103 14. Fonseca F, Mestre-Pinto JI, Gomez-Gomez A, Martinez-Sanvisens D, Rodriguez-Minguela R, Papaseit E, et al. The Tryptophan System in Cocaine-Induced Depression. Journal of Clinical Medicine. 2020 Dec;9(12). 15. Hammond ER, Lai SH, Wright CM, Treisman GJ. Cocaine Use May be Associated with Increased Depression in Persons Infected with HIV. Aids and Behavior. 2016 Feb;20(2):345– 52. 16. Buttimer A. Chapter 12: Airs, waters, places: perennial puzzles of health and environment. Med Hist. 2000;44(S20):211–6. 17. Jouanna J, Allies N. Air, Miasma and Contagion in the Time of Hippocrates and the Survival of Miasmas in Post-Hippocratic Medicine (rufus of Ephesus, Galen and Palladius). In: van der Eijk P, editor. Greek Medicine from Hippocrates to Galen [Internet]. Brill; 2012 [cited 2023 Apr 3]. p. 119–36. (Selected Papers). Available from: http://www.jstor.org.unmsm.lookproxy.com/stable/10.1163/j.ctt1w76vxr.12 18. Berryman JW. Motion and rest: Galen on exercise and health. The Lancet. 2012 Jul;380(9838):210–1. 19. Cheeseman EA, Martin WJ, Russell WT. Disease and Environment. Biometrika. 1939;30(3):341–62. 20. Bind MA. Causal Modeling in Environmental Health. In: Fielding JE, editor. Annual Review of Public Health, Vol 40 [Internet]. Palo Alto: Annual Reviews; 2019. p. 23–43. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6445691/ 21. Celentano D. Gordis Epidemiology. 6th Edition. Amsterdam: Elsevier; 2018. 22. Hartig T, Mitchell R, de Vries S, Frumkin H. Nature and Health. Annu Rev Public Health. 2014 Mar 18;35(1):207–28. 23. Lossio J. Nación, ciencia y salud: investigación médica en los Andes y la construcción de una «patología de las alturas». Histórica. 2009 Mar 7;33(1):65–86. 24. Peacock AJ. ABC of oxygen: Oxygen at high altitude. BMJ. 1998 Oct 17;317(7165):1063–6. 25. Penaloza D, Sime F, Ruiz L. Pulmonary Hemodynamics in children living at high altitudes. High Alt Med Biol. 2008 FAL;9(3):199–207. 26. Morla LL. Soroche agudo: Edema agudo del pulmón. Anales de la Facultad de Medicina. 1955 Jun 20;38(2):244–74. 27. Luks AM, Auerbach PS, Freer L, Grissom CK, Keyes LE, McIntosh SE, et al. Wilderness Medical Society Clinical Practice Guidelines for the Prevention and Treatment of Acute 104 Altitude Illness: 2019 Update. Wilderness & Environmental Medicine. 2019 Dec;30(4):S3– 18. 28. Baertsch P, Swenson ER. Acute High-Altitude Illnesses. N Engl J Med. 2013 Jun 13;368(24):2294–302. 29. Ren Y, Fu Z, Shen W, Jiang P, He Y, Peng S, et al. Incidence of High Altitude Illnesses among Unacclimatized Persons Who Acutely Ascended to Tibet. High Altitude Medicine & Biology. 2010 Apr;11(1):39–42. 30. Penaloza D, Arias-Stella J. The heart and pulmonary circulation at high altitudes: healthy highlanders and chronic mountain sickness. Circulation. 2007 Mar 6;115(9):1132–46. 31. Beall CM. Two routes to functional adaptation: Tibetan and Andean high-altitude natives. Proc Natl Acad Sci U S A. 2007 May 15;104:8655–60. 32. Monge C. Chronic mountain sickness. Physiological Reviews. 1943 Apr;23(2):166–84. 33. Monge C, Arregui C, Leonvelarde F. Pathophysiology and Epidemiology of Chronic Mountain-Sickness. Int J Sports Med. 1992 Oct;13:S79–81. 34. De Ferrari A, Miranda JJ, Gilman RH, Dávila-Román VG, León-Velarde F, Rivera-Ch M, et al. Prevalence, Clinical Profile, Iron Status, and Subject-Specific Traits for Excessive Erythrocytosis in Andean Adults Living Permanently at 3,825 Meters Above Sea Level. Chest. 2014 Nov;146(5):1327–36. 35. Spielvogel H, Vargas E, Paz Zamora M, Haas J, Beard JL, Tufts G, et al. Poliglobulia y ejercicio muscular. Gaceta Torax. 1983;21:6–12. 36. Wu T. The Qinghai–Tibetan Plateau: How High Do Tibetans Live? High Altitude Medicine & Biology. 2001 Dec;2(4):489–99. 37. Beall CM. Adaptations to Altitude: A Current Assessment. Annu Rev Anthropol. 2001 Oct;30(1):423–56. 38. Bible Gateway [Internet]. [cited 2023 Apr 4]. Bible Gateway passage: 1 Samuel 28 - Reina- Valera 1995. Available from: https://www.biblegateway.com/passage/?search=1%20Samuel%2028&version=RVR1995 39. Horwitz AV, Wakefield JC, Lorenzo-Luaces L. History of Depression. In: DeRubeis RJ, Strunk DR, editors. The Oxford Handbook of Mood Disorders [Internet]. Oxford University Press; 2017 [cited 2023 Mar 4]. p. 10–23. Available from: https://academic.oup.com/edited- volume/41353/chapter/352517403 105 40. Telles-Correia D, Marques JG. Melancholia before the twentieth century: fear and sorrow or partial insanity? Front Psychol [Internet]. 2015 Feb 3 [cited 2023 Apr 4];6. Available from: http://journal.frontiersin.org/Article/10.3389/fpsyg.2015.00081/abstract 41. Ruiloba JV, editor. Introducción a la psicopatología y la psiquiatría. 8 edition. Elsevier Masson; 2015. 712 p. 42. APA dictionary of psychology, 2nd ed. Washington, DC, US: American Psychological Association; 2015. xv, 1204 p. (VandenBos GR, editor. APA dictionary of psychology, 2nd ed). 43. World Health Organization. International Classification of Diseases, Tenth Revision (ICD-10). 2015. 2020 p. 44. American Psychiatric Association. Diagnostic and statistical manual of mental disorders: DSM-5TM, 5th ed. Arlington, VA, US: American Psychiatric Publishing, Inc.; 2013. xliv, 947 p. (Diagnostic and statistical manual of mental disorders: DSM-5TM, 5th ed). 45. Berrios GE. The psychopathology of affectivity: conceptual and historical aspects. Psychological Medicine. 1985 Nov;15(4):745–58. 46. Bleuler E. Affectivity, Suggestibility, Paranoia. 2018. 129 p. 47. Rottenberg J, Johnson SL. Emotion and Psychopathology: Bridging Affective and Clinical Science. 2007. 816 p. 48. Park S, Jang EY, Xiang Y, Kanba S, Kato TA, Chong M, et al. Network analysis of the depressive symptom profiles in Asian patients with depressive disorders: Findings from the Research on Asian Psychotropic Prescription Patterns for Antidepressants (REAP-AD). Psychiatry Clin Neurosci. 2020 Jun;74(6):344–53. 49. Kumar GK. Hypoxia. 3. Hypoxia and neurotransmitter synthesis. American Journal of Physiology-Cell Physiology. 2011 Apr;300(4):C743–51. 50. Alcantara-Zapata DE, Lucero N, De Gregorio N, Astudillo Cornejo P, Ibarra Villanueva C, Baltodano-Calle MJ, et al. Women’s mood at high altitude. sexual dimorphism in hypoxic stress modulation by the tryptophan–melatonin axis. Frontiers in Physiology [Internet]. 2023 [cited 2023 Apr 5];13. Available from: https://www.frontiersin.org/articles/10.3389/fphys.2022.1099276 51. Nguyen K, Gates C, Hassell J, Foxx C, Salazar S, Luthens A, et al. Evaluation of the effects of altitude on biological signatures of inflammation and anxiety- and depressive-like behavioral responses. Progress in Neuro-Psychopharmacology and Biological Psychiatry. 2021 Apr 1;111:110331. 106 52. Zhu D, Zhang M, He B, Wan Y, Wang L, Gao F. The role of sex and ovarian hormones in hippocampal damage and cognitive deficits induced by chronic exposure to hypobaric hypoxia. Front Neurosci. 2022;16:953417. 53. Murray AJ. Energy metabolism and the high-altitude environment: Energy metabolism at high altitude. Exp Physiol. 2016 Jan 1;101(1):23–7. 54. Andres RH, Ducray AD, Schlattner U, Wallimann T, Widmer HR. Functions and effects of creatine in the central nervous system. Brain Res Bull. 2008 Jul 1;76(4):329–43. 55. de Aquino Lemos V, Antunes HKM, dos Santos RVT, Lira FS, Tufik S, de Mello MT. High altitude exposure impairs sleep patterns, mood, and cognitive functions. Psychophysiology. 2012;49(9):1298–306. 56. Stadelmann K, Latshang TD, Lo Cascio CM, Tesler N, Stoewhas AC, Kohler M, et al. Quantitative changes in the sleep EEG at moderate altitude (1630 m and 2590 m). PLoS One. 2013;8(10):e76945. 57. Wang H, Li X, Li J, Gao Y, Li W, Zhao X, et al. Sleep, short-term memory, and mood states of volunteers with increasing altitude. Front Psychiatry. 2022 Oct 12;13:952399. 58. Zhang J, Tang S, Chen C, Jiang H, Liao H, Liu H, et al. A bibliometric analysis of the studies in high-altitude induced sleep disturbances and cognitive impairment research. Front Physiol. 2023 Jan 13;14:1133059. 59. Li Y, Wang Y. Effects of Long-Term Exposure to High Altitude Hypoxia on Cognitive Function and Its Mechanism: A Narrative Review. Brain Sciences. 2022 Jun 20;12(6):808. 60. Cheng D, Yakobi R, Dobbins W, Neuman K, Brenner B. Moderate altitude increases suicide deaths. Ann Emerg Med. 2002;40:s55. 61. Brenner B. Suicide Rates Strongly Correlate with Altitude: A Study of 3,060 U.S. Counties. Academic Emergency Medicine - ACAD EMERG MED. 2006 May 1;13. 62. Kim N, Mickelson JB, Brenner BE, Haws CA, Yurgelun-Todd DA, Renshaw PF. Altitude, Gun Ownership, Rural Areas, and Suicide. AJP. 2011 Jan;168(1):49–54. 63. Kim J, Choi N, Lee YJ, An H, Kim N, Yoon HK, et al. High Altitude Remains Associated with Elevated Suicide Rates after Adjusting for Socioeconomic Status: A Study from South Korea. Psychiatry Investig. 2014;11(4):492. 64. Alameda-Palacios J, Ruiz-Ramos M, Garcia-Robredo B. Suicide Mortality in Andalusia, Spain: Geographical Distribution and Relationship with Antidepressants, Altitude and Socioeconomic Inequalities. Revista Espanola De Salud Publica. 2015 May;89(3):283–93. 107 65. Risal A, Manandhar K, Linde M, Steiner TJ, Holen A. Anxiety and depression in Nepal: prevalence, comorbidity and associations. BMC Psychiatry. 2016 Apr 14;16(1):102. 66. Hernández-Vásquez A, Vargas-Fernández R, Rojas-Roque C, Gamboa-Unsihuay JE. Association between altitude and depression in Peru: An 8-year pooled analysis of population-based surveys. Journal of Affective Disorders. 2022 Feb;299:536–44. 67. Zegarra-Rodríguez CA, Plasencia-Dueñas NR, Failoc-Rojas VE. Disparities in the prevalence of screened depression at different altitudes in Peru: A retrospective analysis of the ENDES 2019. Silva MT, editor. PLoS ONE. 2022 Dec 21;17(12):e0278947. 68. World Health Organization. Depression and other common mental disorders: global health estimates [Internet]. World Health Organization; 2017 [cited 2023 Apr 8]. Report No.: WHO/MSD/MER/2017.2. Available from: https://apps.who.int/iris/handle/10665/254610 69. Tremblay JC, Ainslie PN. Global and country-level estimates of human population at high altitude. Proc Natl Acad Sci USA. 2021 May 4;118(18):e2102463118. 70. Cheng DC, Mendenhall TI, Brenner BE. Suicide rates strongly correlate with altitude. Academic Emergency Medicine. 2005;12(5 Supplement 1):141. 71. Wang F, Liu S, Zhang Q, Ng CH, Cui X, Zhang D, et al. Prevalence of Depression in Older Nursing Home Residents in High and Low Altitude Regions: A Comparative Study. Frontiers in Psychiatry [Internet]. 2021 [cited 2023 Apr 8];12. Available from: https://www.frontiersin.org/articles/10.3389/fpsyt.2021.669234 72. Wang J, Zhou Y, Liang Y, Liu Z. A Large Sample Survey of Tibetan People on the Qinghai– Tibet Plateau: Current Situation of Depression and Risk Factors. International Journal of Environmental Research and Public Health. 2020 Jan;17(1):289. 73. Fay RE, Herriot RA. Estimates of Income for Small Places: An Application of James-Stein Procedures to Census Data. Journal of the American Statistical Association. 1979;74(366):269–77. 74. Ouma J, Jeffery C, Awor CA, Muruta A, Musinguzi J, Wanyenze RK, et al. Model-based small area estimation methods and precise district-level HIV prevalence estimates in Uganda. Ortega JA, editor. PLoS ONE. 2021 Aug 6;16(8):e0253375. 75. Institute SMH. 2003 Epidemiological Study on Mental Health in the Peruvian Andes. General Report. Anales de Salud Mental [Internet]. 2003;19(1 and 2). Available from: https://www.insm.gob.pe/investigacion/estudios.html 76. Lohr SL. Sampling: design and analysis. 2nd ed., international ed. Boston: Brooks/Cole; 2010. 596 p. 108 77. Cruz, Victor Orlando. Epidemiology of coca leaf chewing in the 21st century peru [Internet]. Michigan State University; Available from: https://d.lib.msu.edu/etd/49410 78. Instituto Nacional de Salud Mental. Estudio Epidemiológico de Salud Mental en Lima Metropolitana y Callao Replicación 2012. Informe General. Anales de Salud Mental [Internet]. 2013;XXIX(Suplemento 1). Available from: http://www.insm.gob.pe/investigacion/archivos/estudios/2012%20asm%20-eesm%20- lm.pdf 79. CEPLAN. Información para el planeamiento a nivel departamental, provincial y distrital [Internet]. CEPLAN. Available from: https://www.ceplan.gob.pe/informacion-sobre-zonas-y- departamentos-del-peru/ 80. Harding TW, de Arango MV, Baltazar J, Climent CE, Ibrahim HH, Ladrido-Ignacio L, et al. Mental disorders in primary health care: a study of their frequency and diagnosis in four developing countries. Psychological medicine. 1980 May;10(2):231–41. 81. Sheehan DV, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E, et al. The Mini- International Neuropsychiatric Interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin Psychiatry. 1998;59 Suppl 20:22-33;quiz 34-57. 82. Hainbach N, Halbmeier C, Schmid T, Schröder C. A practical guide for the computation of domain-level estimates with the Socio-Economic Panel (and other household surveys) [Internet]. SOEPpapers on Multidisciplinary Panel Data Research; 2019 [cited 2024 Sep 25]. Report No.: 1055. Available from: https://www.econstor.eu/handle/10419/206641 83. United Nations. Toolkit on Using Small Area Estimation for SDGs [Internet]. [cited 2023 Apr 9]. Available from: https://www.unsdglearn.org/microlearning/toolkit-on-using-small-area- estimation-for-sdgs/ 84. Franco C. An Introduction to Small Area Estimation [Internet]. Available from: https://community.amstat.org/committeeoninternationalrelationsinstatistics/events2/webi nars 85. Halbmeier C, Kreutzmann AK, Schmid T, Schröder C. The fayherriot command for estimating small-area indicators. The Stata Journal. 2019 Sep;19(3):626–44. 86. Kroenke K, Spitzer RL, Williams JBW. The Patient Health Questionnaire-2: validity of a two- item depression screener. Med Care. 2003 Nov;41(11):1284–92. 87. Löwe B, Kroenke K, Gräfe K. Detecting and monitoring depression with a two-item questionnaire (PHQ-2). J Psychosom Res. 2005 Feb;58(2):163–71. 88. Morales D, Krause J, Burgard JP. On the Use of Aggregate Survey Data for Estimating Regional Major Depressive Disorder Prevalence. Psychometrika. 2022 Mar;87(1):344–68. 109 89. Moradpour F, Hajebi A, Salehi M, Solaymani-Dodaran M, Rahimi-Movaghar A, Sharifi V, et al. Province-Level Prevalence of Psychiatric Disorders: Application of Small-Area Methodology to the Iranian Mental Health Survey (IranMHS). Iran J Psychiatry. 2019 Jan;14(1):16–32. 90. Lin YH, McLain AC, Probst JC, Bennett KJ, Qureshi ZP, Eberth JM. Health-related quality of life among adults 65 years and older in the United States, 2011-2012: a multilevel small area estimation approach. Ann Epidemiol. 2017 Jan;27(1):52–8. 91. Hudson CG. Validation of a model for estimating state and local prevalence of serious mental illness. Int J Methods Psychiatr Res. 2009 Dec;18(4):251–64. 92. Hudson CG, Abbott MW. Modeling the geographic distribution of serious mental illness in New Zealand. Soc Psychiatry Psychiatr Epidemiol. 2013 Jan;48(1):25–36. 93. Ortiz-Prado E, Simbaña-Rivera K, Duta D, Ochoa I, Izquierdo-Condoy JS, Vasconez E, et al. Optimism and Health Self-Perception-Related Differences in Indigenous Kiwchas of Ecuador at Low and High Altitude: A Cross-Sectional Analysis. High Alt Med Biol. 2022 Mar;23(1):26– 36. 94. Zhang D, Ma H, Huang J, Xinjuan Z, Ma H, Liu M. Exploring the impact of chronic high- altitude exposure on visual spatial attention using the ERP approach. Brain and Behavior. 2018 Mar 25;8:e00944. 95. Instituto Especializado en Salud Mental. Estudio Epidemiologico Metropolitano en Salud Mental 2002 [Internet]. 2002. Available from: http://www.insm.gob.pe/investigacion/estudios.html 96. StataCorp. Stata Statistical Software: Release 17. College Station, TX: StataCorp LLC; 2021. 97. Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic research. Epidemiology. 1999 Jan;10(1):37–48. 98. Penninx BW, Milaneschi Y, Lamers F, Vogelzangs N. Understanding the somatic consequences of depression: biological mechanisms and the role of depression symptom profile. BMC Med. 2013 Dec;11(1):129. 99. Pincus HA, Pettit AR. The societal costs of chronic major depression. J Clin Psychiatry. 2001;62 Suppl 6:5–9. 100. Linder A, Gerdtham UG, Trygg N, Fritzell S, Saha S. Inequalities in the economic consequences of depression and anxiety in Europe: a systematic scoping review. European Journal of Public Health. 2020 Aug 1;30(4):767–77. 101. Basnyat B, Murdoch DR. High-altitude illness. Lancet. 2003 Jun 7;361(9373):1967–74. 110 102. Biondich AS, Joslin JD. Coca: The History and Medical Significance of an Ancient Andean Tradition. Emergency Medicine International [Internet]. 2016;2016. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4838786/pdf/EMI2016-4048764.pdf 103. Burchard RE. Coca Chewing and Diet. Current Anthropology. 1992 Feb;33(1):1–24. 104. Rivera MA, Aufderheide AC, Cartmell LW, Torres CM, Langsjoen O. Antiquity of Coca-Leaf Chewing in the South Central Andes: A 3,000 Year Archaeological Record of Coca-Leaf Chewing from Northern Chile. Journal of Psychoactive Drugs. 2005 Dec;37(4):455–8. 105. Stolberg VictorB. The Use of Coca: Prehistory, History, and Ethnography. Journal of Ethnicity in Substance Abuse. 2011 Apr;10(2):126–46. 106. Martin RT. The Role of Coca in the History, Religion, and Medicine of South American Indians. Economic Botany. 1970;24(4):422–38. 107. Rospigliosi F, Blondet C, Lloréns Amico JA. El consumo tradicional de la hoja de coca en el Perú. 1. ed. Lima: Instituto de Estudios Peruanos; 2004. 146 p. (Estudios de la sociedad rural). 108. Haas LF. Coca shrub (Erythroxylum coca). J Neurol Neurosurg Psychiatry. 1995 Jul;59(1):25. 109. Bennett A. The Physiological Action Of Coca. The British Medical Journal. 1874;1(694):510– 510. 110. Allen CJ. To Be Quechua: The Symbolism of Coca Chewing in Highland Peru. American Ethnologist. 1981;8(1):157–71. 111. Bauer I. Travel medicine, coca and cocaine: demystifying and rehabilitating Erythroxylum - a comprehensive review. Tropical Diseases Travel Medicine and Vaccines [Internet]. 2019 Nov;5(1). Available from: https://tdtmvjournal.biomedcentral.com/articles/10.1186/s40794-019-0095-7 112. Fuchs A, Burchard RE, Curtain CC, De Azeredo PR, Frisancho AR, Gagliano JA, et al. Coca Chewing and High-Altitude Stress: Possible Effects of Coca Alkaloids on Erythropoiesis [and Comments and Reply]. Current Anthropology. 1978;19(2):277–91. 113. Spielvogel H, Caceres E, Koubi H, Sempore B, Sauvain M, Favier R. Effects of coca chewing on metabolic and hormonal changes during graded incremental exercise to maximum. J Appl Physiol (1985). 1996 Feb;80(2):643–9. 114. Monge C. El problema de la coca en el Perú. Anales de la Facultad de Medicina. 1946 Dec 30;29(4):311–5. 115. Hanna JM. Responses of Quechua Indians to coca ingestion during cold exposure. American Journal of Physical Anthropology. 1971;34(2):273–7. 111 116. Torchetti T. Coca Chewing and High Altitude Adaptation. The University of Western Ontario Journal of Anthropology [Internet]. 1994 [cited 2023 Apr 14];1(1). Available from: https://ojs.lib.uwo.ca/index.php/uwoja/article/view/8731 117. Bray W, Dollery C, Barnett G, Bolton R, Deltgen F, Dufour D, et al. Coca Chewing and High- Altitude Stress: A Spurious Correlation [and Comments and Reply]. Current Anthropology. 1983;24(3):269–82. 118. Favier R, Caceres E, Guillon L, Sempore B, Sauvain M, Koubi H, et al. Coca chewing for exercise: hormonal and metabolic responses of nonhabitual chewers. J Appl Physiol (1985). 1996 Nov;81(5):1901–7. 119. Siqueland L, Horn A, Moras K, Woody G, Weiss R, Blaine J, et al. Cocaine-induced mood disorder: prevalence rates and psychiatric symptoms in an outpatient cocaine-dependent sample. Am J Addict. 1999;8(2):165–9. 120. Knol MJ, VanderWeele TJ. Recommendations for presenting analyses of effect modification and interaction. Int J Epidemiol. 2012 Apr;41(2):514–20. 121. Martini S, Wagner FA, Anthony JC. The Association of Tobacco Smoking and Depression in Adolescence: Evidence from the United States. Substance Use & Misuse. 2002 Jan 1;37(14):1853–67. 122. Boden JM, Fergusson DM. Alcohol and depression. Addiction. 2011;106(5):906–14. 123. Conway KP, Compton W, Stinson FS, Grant BF. Lifetime comorbidity of DSM-IV mood and anxiety disorders and specific drug use disorders: results from the National Epidemiologic Survey on Alcohol and Related Conditions. J Clin Psychiatry. 2006 Feb;67(2):247–57. 124. Chahua M, Sordo L, Molist G, Domingo-Salvany A, Brugal MT, de la Fuente L, et al. Depression in young regular cocaine users recruited in the community. Gaceta Sanitaria. 2014 Mar;28(2):155–9. 125. Meredith SE, Juliano LM, Hughes JR, Griffiths RR. Caffeine Use Disorder: A Comprehensive Review and Research Agenda. J Caffeine Res. 2013 Sep;3(3):114–30. 126. Putnick DL, Bornstein MH. Measurement invariance conventions and reporting: The state of the art and future directions for psychological research. Developmental Review. 2016 Sep;41:71–90. 127. Davidov E, Meuleman B, Cieciuch J, Schmidt P, Billiet J. Measurement Equivalence in Cross- National Research. Annu Rev Sociol. 2014 Jul 30;40(1):55–75. 128. Practice Guideline for the Treatment of Patients with Major Depressive Disorder: Approved in May 2010 and Published in October 2010 [Internet]. American Psychiatric Association; 2010. 118 p. Available from: 112 https://psychiatryonline.org/pb/assets/raw/sitewide/practice_guidelines/guidelines/mdd.p df 129. Gutiérrez-Noriega C. Acción de la coca sobre la actividad mental de sujetos habituados. Revista Peruana de Medicina Experimental y Salud Publica. 1944 Mar;3(1):1–18. 130. Schinder EOM, Ruder AM. Epidemiology of coca and alcohol use among high-altitude miners in argentina. American Journal of Industrial Medicine. 1989;15(5):579–87. 131. Pergolizzi J, Breve F, Magnusson P, LeQuang JAK, Varrassi G. Cocaethylene: When Cocaine and Alcohol Are Taken Together. Cureus. 2022 Feb;14(2):e22498. 132. Andrews P. Cocaethylene Toxicity. Journal of Addictive Diseases. 1997 Jul 1;16(3):75–84. 133. Quinlan P, Vanderziel A, Gregoire V, Anthony JC. The negative growth of a brief depressive spell among U.S. adolescents. Int J Methods Psychiatr Res. 2019 Dec;28(4):e1798. 134. Chang C, Gardiner J, Houang R, Yu YL. Comparing multiple statistical software for multiple- indicator, multiple-cause modeling: an application of gender disparity in adult cognitive functioning using MIDUS II dataset. BMC Medical Research Methodology. 2020 Nov 12;20(1):275. 135. Gallo JJ, Anthony JC, Muthén BO. Age differences in the symptoms of depression: a latent trait analysis. J Gerontol. 1994 Nov;49(6):P251-264. 136. Gallo JJ, Rabins PV, Anthony JC. Sadness in older persons: 13-year follow-up of a community sample in Baltimore, Maryland. Psychol Med. 1999 Mar;29(2):341–50. 113 APPENDIX A: DEPRESSED MOOD AND COCA USE QUESTIONS FROM EPIDEMIOLOGICAL SURVEYS OF THE PERUVIAN NATIONAL INSTITUTE OF MENTAL HEALTH “HONORIO DELGADO – HIDEYO NOGUCHI” (“NOGUCHI SURVEYS”) Epidemiological Surveys of The Peruvian National Institute of Mental Health “Honorio Delgado – Hideyo Noguchi” (“Noguchi Surveys”) employed a survey questionnaire for adults that has five sections (i.e., sociodemographic, general adult health, clinical syndromes part A, clinical syndromes part B, and access to health services). Here, we include an English language translation of the standardized survey items used in this dissertation to assess the mood state during the last four weeks and to determine participants who were coca-leaf users in the last month before the interview. All questions were applied to participants in their native “mother tongue.” Depressed mood during the last four weeks before the survey interview was evaluated using two different parts of the questionnaires. The first included a Likert question directly asking about the participant's mood (i.e., Question 6, from the “prevalent mood states” part of the questionnaire). The second question was included as one of the items for the screening tool SRQ-17, which asks about depressive and anxiety symptoms in the last four weeks. One of these questions concerns frequent sadness (i.e., Question 26). Actively depressed mood was finally identified in persons who recognized feeling sadness “always” or “almost always” in question #6 and also answered “Yes” to the question “Have you felt sad frequently in the last four weeks?”. Coca-leaf use during the last month was assessed in the Clinical Syndromes part A module, which assesses substance use. 114 ADULT GENERAL HEALTH SECTION QUESTION 6: Prevalent Mood States (“estados anímicos prevalentes”) 6. HOW OFTEN DO YOU FEEL…? CARD 2 Codes 1. Never 4. Almost always 2. Rarely 5. Always 3. Occasionally 6. Doesn't respond a. Sad? b. Tense? c. anguished? d. Irritable (or rabid)? e. Worried? f. Calm? g. Happy? h. Bored? z. Other?(specify) 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 115 ADULT GENERAL HEALTH SECTION QUESTION 26: SRQ 26. ABOUT YOUR HEALTH IN THE PAST 4 WEEKS: (YES / NO) a) Have you had headaches, neck, back, or pain in other parts of the body? b) Has your appetite decreased? or c) Has your appetite increased? d) Have you had heartburn? f) Have you slept badly? g) Are you easily scared? h) Have you had a hand tremor? i) Have you felt nervous or tense? or j) Have you felt bored? k) Have you had poor digestion? n) Have you been able to think clearly? o) Have you felt sad frequently? p) Have you cried frequently? q) Do you enjoy your daily activities less? r) Has your ability to make decisions decreased? u) Have you lost interest in things? v) Has a person felt useless? w) Have you often felt tired? 24) Have you ever wanted to die in your life? and 25) In the last month? 116 CLINICAL SYNDROMES - PART B QUESTIONS ABOUT SUBSTANCE USE (INCLUDING COCA LEAF) INTERVIEWER: CIRCLE THE LETTERS CORRESPONDING TO THE POSITIVE RESPONSES 51. Which of these substances have you tried in your lifetime, either for curiosity, pleasure or because of peer pressure (not for medical indication)? INTERVIEWER: IF YOU HAVE NOT CODED ANY ANSWER, GO TO Q76 52. At what age did you consume for the first time? 53. When was the last time you consumed? a. In the last 30 days b. More than one month but less than a year c. More than one year 54. Have you ever though, or someone said to you, that you consume too much? 55. Have you ever wanted (or do you want) to stop consuming? YES = 1 NO = 0 i ) e t t e r a g C ( o c a b a T s r e z i l i u q n a r T s l l i i P g n p e e l S r o F s l l i P o T ( s t n a u m l i t S p u r y S h g u o C s e h c a d a e H i ) t h g e W e s o L s i s a t c E a n a u h i r a M . l d y h r o h C e n a c o C i ) C B P ( a c i s a B a t s a P s n e g o n i c u l l a H i e n o r e H p i r T ) D S L ( f a e L a c o C s t n a a h n l I A B C D E F G H I J K L M N A B C D E F G H A B C D E F G H A B C D E F G H A B C D E F G H I I I I J K L M N J K L M N J K L M N J K L M N 1 0 1 1 1 1 1 0 0 0 0 0 1 1 1 1 1 1 0 0 0 0 0 0 1 1 0 0 117 Table 3.10.A. Direct estimates for depressed mood prevalence proportions at the district level. APPENDIX B: CHAPTER 3 District Lurin Bellavista Ancon Victor Larco Herrera Callao La Punta Coishco Castilla La Perla Nuevo Chimbote Ventanilla Chimbote Piura Los Olivos Chorrillos Pachacama Santa Rosa Trujillo Carmen de la Legua Reynoso San Miguel Magdalena Del Mar Magdalena Vieja altitude (m) 12 13 14 24 27 29 31 35 37 40 43 52 57 67 68 68 72 74 total sample 33 34 21 93 225 5 56 457 46 477 201 801 854 112 185 45 5 508 Weighted population 30968 44671 28659 68445 316857 5078 17515 135512 47858 186049 186835 197853 274312 208440 227417 45724 5472 496572 depressed mood cases 6 1 3 9 12 0 4 55 1 73 10 97 109 5 11 4 1 54 82 84 90 90 25 49 33 26 34045 126159 85095 40621 2 1 2 1 118 weighted proportion of depressed1 0.1212542 0.0391081 0.1292439 0.0696910 0.0553846 0.0000000 0.0581787 0.0916007 0.0422082 0.1084768 0.0485669 0.1086564 0.1066528 0.0408607 0.0726931 0.0529263 0.1312135 0.1036768 Linearized Std Error 0.045376 0.035472 0.05119 0.035359 0.015956 Lower Lim 95% 0.05649 0.00635 0.05735 0.02507 0.03123 . . . 0.019229 0.014472 0.036471 0.016138 0.020422 0.014754 0.012012 0.017744 0.02795 0.029389 0.03011 0.0669 0.00746 0.08065 0.02101 0.08293 0.08529 0.01723 0.0336 0.01739 Upper Lim 95% 0.241298 0.205764 0.265853 0.179158 0.09635 0.109467 0.124202 0.205374 0.144404 0.108279 0.141143 0.132597 0.093799 0.150202 0.149972 . . . 0.017654 0.07381 0.143747 0.0722279 0.0110495 0.032166 0.010805 0.0547858 0.0327417 0.036757 0.031894 0.02948 0.00161 0.01421 0.00468 0.166346 0.07206 0.189056 0.19606 Table 3.10.A. (cont’d) El Porvenir Florencia de Mora Barranco Mazan Comas Santiago de Surco Iquitos Independen Belen San Juan Bautista Punchana Miraflores Surquillo Nauta Yarinacocha San Juan de Miraflores La Esperanza San Martin de Porres Jesus Maria La Victoria Lince Masisea Breña Rimac Lima Calleria San Borja 92 92 97 106 107 107 107 111 116 120 124 125 125 127 131 133 137 138 142 142 150 150 153 153 162 162 170 255 60 28 146 160 156 668 60 178 413 220 55 33 234 233 197 306 198 20 85 32 107 44 75 124 1028 21 130446 55497 21329 1577 361708 185770 127568 98000 34678 3957 47952 78344 52517 2658 22726 241112 165756 389709 42375 139829 41571 1410 80822 159995 209166 106693 43525 0.1051776 0.0697695 0.0663885 0.1037132 0.0305209 0.0247187 0.1152850 0.1089592 0.1388504 0.1071972 0.0903451 0.0608853 0.0386351 0.0880361 0.1495833 0.026135 0.023813 0.050509 0.024725 0.013927 0.015543 0.016173 0.031586 0.041917 0.021865 0.021277 0.034382 0.032473 0.016899 0.037689 0.0483178 0.1308429 0.019702 0.022492 0.06384 0.03524 0.01418 0.06427 0.01235 0.00711 0.08716 0.06069 0.075 0.07125 0.05641 0.01955 0.00719 0.06005 0.08958 0.02145 0.09268 0.168475 0.133448 0.260062 0.163134 0.073413 0.082339 0.150981 0.187933 0.242782 0.158194 0.141634 0.174094 0.182443 0.127291 0.239227 0.105243 0.181575 . . 0.0673708 0.0000000 0.0409254 0.0160208 0.1290780 0.0000000 0.0406763 0.0851238 0.1014977 0.0482022 0.021385 0.03574 0.123419 0.022625 0.013716 0.042348 0.018648 0.027095 0.012324 0.049379 0.01359 0.00295 0.06613 0.01634 0.04492 0.07976 0.0061 0.116713 0.08229 0.236766 0.097659 0.155452 0.128339 0.294725 28 5 1 13 5 3 73 8 21 43 19 6 1 21 29 9 44 12 0 4 1 9 0 4 8 110 1 119 Table 3.10.A. (cont’d) Curimana Nueva Requena Puente Piedra San Isidro El Agustino Campoverde Villa El Salvador Villa Maria Del Triunfo San Luis San Juan de Lurigancho Irazola Carabayllo La Molina Padre Abad Santa Anita Coayllo Cieneguilla Morales Tarapoto Ate La Banda De Shilcayo Ica La Tinguina Parcona Llochegua Sivia 181 183 187 195 200 203 204 210 214 222 228 238 262 275 285 285 287 290 342 378 418 432 463 472 540 561 86 29 97 37 86 204 235 196 21 596 235 77 61 203 131 41 9 249 877 386 212 849 189 317 62 46 1332 471 179820 49038 105562 2265 263244 206200 31229 633358 2066 141874 65850 3040 124334 693 7737 13378 43535 352291 10468 144321 27137 51844 4958 3226 0.0720721 0.0360934 0.020237 4.34E-19 0.0977422 0.0168033 0.0590894 0.0966887 0.0908987 0.023793 0.016745 0.016058 0.02323 0.025181 0.04115 0.03609 0.06 0.00234 0.03442 0.05974 0.05214 0.123257 0.036093 0.155309 0.110872 0.099619 0.152769 0.153785 0.1157565 0.0000000 . 0.022516 0.07838 0.167713 . . . 0.0754879 0.1040658 0.1494284 0.0507517 0.0434211 0.0856941 0.0655632 0.0000000 0.1154281 0.0906306 0.0487892 0.0797129 0.1025769 0.1084866 0.1254533 0.1644763 0.1247241 0.012654 0.026096 0.05615 0.023404 0.016333 0.0235 0.00272 0.029327 0.01137 0.010876 0.025437 0.015165 0.02842 0.02097 0.023965 0.02892 0.05412 0.06288 0.0688 0.02021 0.02056 0.04948 0.06043 0.06916 0.07067 0.03138 0.04204 0.07643 0.06402 0.08976 0.12269 0.07814 0.104353 0.167419 0.294663 0.12173 0.089368 0.144394 0.071103 0.186447 0.115531 0.075111 0.145998 0.136355 0.177976 0.172645 0.216979 0.193256 6 2 12 1 7 24 25 26 0 48 24 11 4 8 15 2 0 29 83 20 14 96 30 46 11 7 120 Table 3.10.A. (cont’d) Chaclacayo Zuðiga Lurigancho Santa Rosa De Quives Huanchay Catahuasi Pariacoto Magdalena Antioquia San Bartolome Huanuco Amarilis Putinza Pillco Marca Surco San Mateo De Otao Ambar Cochabamb Chumuch Tiabaya Asuncion Sachaca Jacob Hunter San Juan Socabaya Cortegana 685 827 879 936 1067 1203 1264 1298 1573 1644 1921 1950 1985 1996 2049 2084 2084 2135 2202 2218 2254 2300 2309 2336 2352 2352 26 40 142 51 30 64 29 16 39 63 729 529 25 235 54 37 55 16 15 23 30 35 69 15 119 14 24163 3649 134734 3895 1044 1208 1208 1530 1139 1080 45009 38985 1065 20172 1194 2740 694 1018 1930 15931 4342 10700 41133 1651 55143 2803 0.1064851 0.0581796 0.0791263 0.0347893 0.1531222 0.1059964 0.0667477 0.0292257 0.1867312 0.0978018 0.0431107 0.0579364 0.0421715 0.0483964 0.1287751 0.0726108 0.0802309 0.0233441 0.1182837 0.2459984 0.1489309 0.2018692 0.1509007 0.0535630 0.1376603 0.1729343 0.054012 0.027063 0.018058 0.028682 0.035502 0.066812 0.018196 7.37E-18 0.019025 0.03957 0.007843 0.010826 1.08E-17 0.018715 0.033662 0.000728 0.021807 1.30E-18 6.07E-18 0.002372 0.034771 0.090855 0.048337 1.04E-17 0.048971 1.04E-17 0.03768 0.02292 0.0502 0.00671 0.0956 0.02889 0.03877 0.02923 0.15225 0.04306 0.0301 0.04002 0.04217 0.02241 0.07585 0.0712 0.04659 0.02334 0.11828 0.24138 0.09272 0.07725 0.07818 0.05356 0.06638 0.17293 0.266189 0.139933 0.122568 0.161356 0.236224 0.32086 0.112547 0.029226 0.226928 0.207089 0.061391 0.083187 0.042172 0.101388 0.210235 0.074051 0.134729 0.023344 0.118284 0.25068 0.230567 0.433168 0.271341 0.053563 0.263866 0.172934 3 3 14 6 5 8 3 1 4 8 38 39 2 16 11 3 7 1 2 5 5 7 13 2 14 2 121 Table 3.10.A. (cont’d) Jose Luis Bustamante Yanahuara Arequipa Cerro Colorado Miraflores Paucarpata Manas Mariano Melgar Lampian Cospan Abancay Alto Selva Alegre Cayma Arahuay Huasmin Bambamarca Jesus Pacaycasa Luricocha Tinco Marca Jose Galvez Tamburco Llacanora Celendin Sucre Carhuaz Sorochuco 2389 2402 2429 2441 2450 2453 2457 2459 2467 2471 2500 2510 2531 2533 2543 2556 2568 2571 2598 2606 2615 2618 2620 2621 2629 2632 2663 2663 138 34 124 145 78 206 73 90 45 14 1565 128 143 39 61 221 45 12 36 40 15 16 181 15 31 28 64 41 113604 24942 83377 97706 48656 115064 482 67484 672 1242 29779 52231 76338 601 8365 26708 5601 742 2043 3119 245 1874 3663 1809 3063 2907 4489 4178 0.0579293 0.0582552 0.0505295 0.1534501 0.2143826 0.2019485 0.0737177 0.1018464 0.2501439 0.0000000 0.1084948 0.0812544 0.1280620 0.0796215 0.0753670 0.0959410 0.0840636 0.0443948 0.2430948 0.1331871 0.1617828 0.0000000 0.0522157 0.0000000 0.0827215 0.1311303 0.2417230 0.0927269 . . . 0.014695 0.053965 0.023117 0.032133 0.053676 0.048081 0.012517 0.02285 0.015313 0.009653 0.022479 0.039376 0.012494 0.016102 0.02618 0.038452 6.07E-18 0.020738 0.05477 1.13E-17 0.035 0.00891 0.02027 0.1004 0.12746 0.12356 0.05263 0.06497 0.22133 0.09097 0.04671 0.06852 0.05832 0.04926 0.05546 0.03332 0.04439 0.20476 0.05715 0.16178 0.094417 0.298564 0.120409 0.227449 0.337648 0.312343 0.10234 0.156161 0.281352 0.128922 0.137645 0.226752 0.10782 0.113654 0.160931 0.19637 0.044395 0.286022 0.280327 0.161783 0.012926 0.03196 0.084201 0.066618 0.041332 0.076981 0.0251 0.01587 0.06899 0.12274 0.05387 0.335277 0.235109 0.420741 0.155019 11 2 6 27 17 40 7 15 13 0 183 17 25 4 6 25 7 1 7 4 3 0 11 0 2 4 15 4 122 Table 3.10.A. (cont’d) San Pedro De Pilas San Lorenzo De Quinti Huanta Los Baños Del Inca Acopampa Cajamarca Namora Chugur Marcara Sangallaya San Juan Bautista Ayacucho Anta Miguel Iglesias Jesus Nazareno Jangas Pariahuanc Tarica Matara Yungar Oxamarca Ihuari Santo Domingo De Los Acos Vinchos Langa 2678 2682 2685 2685 2692 2731 2765 2765 2767 2779 2786 2797 2800 2813 2817 2824 2830 2832 2834 2836 2836 2850 2861 2874 2889 41 62 78 74 28 1197 44 13 65 39 255 792 30 15 127 59 15 42 16 26 14 100 43 47 53 1291 657 3812 12008 1150 97516 5943 2390 2857 743 15801 46842 1513 1437 7652 2097 1532 2315 2524 1233 1847 1516 841 1861 572 5 0.0973511 0.016383 0.06962 0.134534 0.1301827 0.1596776 0.040807 0.0578 . 0.0549839 0.1423890 0.0885711 0.0362618 0.0000000 0.2016173 0.1113378 0.1459350 0.1670890 0.0827145 0.1448737 0.0876457 0.1388774 0.3066157 0.1783122 0.2442460 0.1530339 0.0507303 0.1167042 0.02496 0.028617 0.009481 0.030445 0.036593 0.029532 0.017593 0.015437 0.02427 6.94E-18 0.019189 0.028115 2.26E-17 0.035404 2.08E-17 0.0408 9.11E-18 0.021176 0.06876 0.07549 0.02218 0.09491 0.07166 0.00677 0.1392 0.06523 0.1147 0.13897 0.04595 0.14487 0.05661 0.09232 0.30662 0.11903 0.24425 0.08881 0.05073 0.08115 0.232773 0.306619 0.129854 0.208156 0.109008 0.171956 0.282831 0.183639 0.18391 0.199584 0.14445 0.144874 0.133287 0.203645 0.306616 0.258456 0.244246 0.250907 0.05073 0.165037 0.1050430 0.007331 0.09151 0.12031 0.0587863 0.0910606 0.053492 0.010656 0.0093 0.07221 0.293676 0.114223 10 18 5 5 123 1 0 17 5 46 162 4 2 18 9 5 6 4 4 1 19 7 4 6 123 Table 3.10.A. (cont’d) Yauyos Amashca Huayllapamp Carmen Alto Huachupamp La Libertad De Pallan San Miguel De Aco Carhuanca Llacllin Shilla Independenc Gorgor Concepcion Iguain Huaraz Encañada Tambillo Vischongo Ayauca Ocros Vinchos Huarochiri Caujul Huancapon Santiago De Pischa Acocro San Damian 2895 2905 2908 2921 2938 84 14 12 125 21 2424 547 344 6260 501 19 1 2 28 2 0.1727309 0.0580088 0.1496042 0.2200759 0.0657959 0.027885 1.73E-18 5.20E-18 0.034994 1.13E-17 0.12466 0.05801 0.1496 0.15909 0.0658 0.234377 0.058009 0.149604 0.2962 0.065796 2952 27 2910 5 0.1441236 0.043031 0.07831 0.250223 2956 2980 3020 3036 3047 3049 3061 3063 3073 3087 3111 3150 3151 3153 3155 3170 3185 3187 3210 3251 3252 32 16 10 42 673 117 16 26 640 91 29 45 76 44 87 38 40 74 16 61 79 1217 399 537 2212 26550 195 1068 1322 26530 14818 2004 1825 552 1809 3803 1364 1040 834 403 3842 1061 5 3 0 10 101 24 2 5 85 7 10 8 22 5 19 4 9 9 1 8 9 124 . 0.1932970 0.1020116 0.0000000 0.1968203 0.1200595 0.1632999 0.0527453 0.2425729 0.1159611 0.1217696 0.3604221 0.1836713 0.2477332 0.1017419 0.1744624 0.1178047 0.1882301 0.0994341 0.0477326 0.1010243 0.0798327 0.013395 4.34E-19 0.028794 0.012149 0.026579 5.64E-18 0.021359 0.014063 0.036566 0.0349 0.072999 0.049672 0.035525 0.035418 0.003923 0.017394 0.046792 9.11E-18 0.028523 0.035528 0.16838 0.10201 0.14636 0.0982 0.11761 0.05275 0.20317 0.0911 0.06622 0.29517 0.07971 0.16337 0.0502 0.11542 0.11033 0.15647 0.03812 0.04773 0.05723 0.03252 0.220924 0.102012 0.259393 0.145992 0.222277 0.052745 0.286861 0.146508 0.21327 0.43127 0.368882 0.35707 0.195319 0.254999 0.125717 0.224725 0.235248 0.047733 0.172205 0.182981 Table 3.10.A. (cont’d) Santillana Paccho San Jose De Ticllas Madean Huambalpa Leoncio Prado Huamanguilla Quinua Huantan Viðac San Andres De Tupicoc Socos Accomarca Colonia Pararin Recuay Azangaro Olleros Carampoma Ticapampa Vilcas Huaman Lincha Hualgayoc Chiara Santa Cruz De Andamar Pampas Chico 3265 3275 3282 3292 3294 3299 3300 3301 3315 3315 3321 3368 3387 3399 3402 3428 3435 3443 3459 3485 3494 3516 3530 3540 45 126 30 65 31 60 39 27 46 111 38 45 15 101 13 15 68 31 25 32 60 36 46 45 3089 1236 2244 201 1286 3116 2141 1230 784 666 1161 2218 509 1013 349 451 136 997 278 1073 3066 935 6966 2815 3550 109 3836 5 19 5 18 5 8 6 3 12 17 3 5 0 22 1 2 16 2 4 3 16 6 10 5 24 0.0659849 0.1414185 0.1709370 0.2631673 0.1406767 0.1329162 0.1671351 0.0897240 0.2662035 0.1163343 0.0993926 0.1126035 0.0000000 0.2182210 0.0918476 0.1242236 0.2411326 0.0269720 0.1267429 0.0692478 0.2845998 0.0977272 0.2614837 0.0678484 . 0.003 0.048415 0.027525 0.044587 0.08989 0.008867 0.056974 0.002148 0.022989 0.024159 0.023551 0.064175 0.026285 1.91E-17 4.68E-17 0.021445 0.019403 2.08E-17 0.027964 0.065383 0.030025 0.039487 0.041691 0.06034 0.07008 0.12349 0.18536 0.0367 0.11647 0.08252 0.0856 0.22361 0.07667 0.06181 0.03477 0.17106 0.09185 0.12422 0.20161 0.00646 0.12674 0.0308 0.17488 0.05263 0.19167 0.01959 0.072117 0.264698 0.231801 0.359231 0.412984 0.151286 0.309272 0.094026 0.313628 0.172673 0.156025 0.308933 0.274092 0.091848 0.124224 0.285633 0.105652 0.126743 0.148355 0.427495 0.174356 0.34585 0.20955 0.2006279 0.012345 0.17751 0.22593 3552 14 417 0 0.0000000 . 125 Table 3.10.A. (cont’d) Saurama Catac Huancaya Pira Independen Huaros Vitis Miraflores Laraos Lachaqui Pampas Ascension Huancavelica Simon Bolivar Chaupimarca Yanacancha 3574 3579 3591 3602 3606 3614 3625 3677 3683 3686 3698 3711 3746 4234 4373 4394 16 14 40 45 31 88 23 37 22 43 16 366 1350 257 611 601 876 657 925 1826 1463 511 893 1072 578 664 587 6172 19679 6778 14364 18474 8 1 6 5 5 15 4 4 2 9 1 40 132 18 32 32 0.4255405 0.0486437 0.1374064 0.0802828 0.1293494 0.1417445 0.1370189 0.0755937 0.0847354 0.1443052 0.0597973 0.0845755 0.0894354 0.0586937 0.0412328 0.0413926 9.02E-17 8.67E-18 0.014493 0.015192 0.017644 0.037946 3.12E-17 0.001053 4.34E-18 0.066232 1.04E-17 0.015013 0.010353 0.017445 0.010706 0.010068 0.42554 0.04864 0.11138 0.0551 0.09851 0.08223 0.13702 0.07355 0.08474 0.05563 0.0598 0.05941 0.07111 0.03248 0.02466 0.02558 0.425541 0.048644 0.168367 0.115568 0.16805 0.233391 0.137019 0.077684 0.084735 0.325579 0.059797 0.119044 0.11192 0.103791 0.068154 0.066308 1 Direct estimate using survey set settings and Taylor Series Linearization. 126 Table 3.10.B. Standardized, and Fay-Herriot (EBLUP) estimates for depressed mood prevalence proportions at the district level. District Lurin Bellavista Ancon Victor Larco Herrera Callao La Punta Coishco Castilla La Perla Nuevo Chimbote Ventanilla Chimbote Piura Los Olivos Chorrillos Pachacama Santa Rosa Trujillo Carmen de la Legua Reynoso San Miguel Magdalena Del Mar Magdalena Vieja altitude (m) 12 13 14 24 27 29 31 35 37 40 43 52 57 67 68 68 72 74 82 84 90 90 total sample 33 34 21 93 225 5 56 457 46 477 201 801 854 112 185 45 5 508 25 49 33 26 depress mood cases 6 1 3 9 12 0 4 55 1 73 10 97 109 5 11 4 1 54 Standardized prevalence proportion2 linzd Std Error Input3 0.079947 0.045376 0.080987 0.035472 0.05119 0.055609 0.069257 0.035359 0.044958 0.015956 0.000000 0.000728 0.078877 0.019229 0.093621 0.014472 0.006968 0.036471 0.125021 0.016138 0.046284 0.020422 0.104098 0.014754 0.102355 0.012012 0.041981 0.017744 0.078213 0.02795 0.053190 0.029389 0.117366 0.000728 0.100086 0.017654 variance input 0.067946 0.04278 0.055028 0.116274 0.057283 2.65E-06 0.020705 0.095707 0.061187 0.124232 0.083825 0.174369 0.12322 0.035262 0.144521 0.038867 2.65E-06 0.158331 EBLUP64 0.0696348 0.0353903 0.0753254 0.0564733 0.0475560 0.0000000 0.0647712 0.0734788 0.0288702 0.0768325 0.0545014 0.0752259 0.0739734 0.0308278 0.0448812 0.0597304 0.1312001 0.0548859 0.124677 0.032166 0.004874 0.010805 0.025866 0.00572 0.0409197 0.0152636 0.033989 0.036757 0.029405 0.031894 0.044587 0.026448 0.0290729 0.0238478 2 1 2 1 127 Table 3.10.B. (cont’d) El Porvenir Florencia De Mora Barranco Mazan Comas Santiago De Surco Iquitos Independen Belen San Juan Bautista Punchana Miraflores Surquillo Nauta Yarinacocha San Juan de Miraflores La Esperanza San Martin de Porres Jesus Maria La Victoria Lince Masisea Breña Rimac Lima Calleria San Borja 92 92 97 106 107 107 107 111 116 120 124 125 125 127 131 133 137 138 142 142 150 150 153 153 162 162 170 255 60 28 146 160 156 668 60 178 413 220 55 33 234 233 197 306 198 20 85 32 107 44 75 124 1028 21 28 5 1 13 5 3 73 8 21 43 19 6 1 21 29 9 44 12 0 4 1 9 0 4 8 110 1 128 0.094772 0.026135 0.174173 0.0658213 0.071790 0.023813 0.009280 0.050509 0.122491 0.024725 0.031699 0.013927 0.025930 0.015543 0.112090 0.016173 0.095623 0.031586 0.140553 0.041917 0.116561 0.021865 0.075820 0.021277 0.065032 0.034382 0.023490 0.032473 0.101794 0.016899 0.118808 0.037689 0.044313 0.019702 0.145227 0.022492 0.057664 0.021385 0.000000 0.000728 0.038740 0.022625 0.038754 0.013716 0.122819 0.042348 0.000000 0.000728 0.043309 0.018648 0.066914 0.027095 0.100221 0.012324 0.057992 0.049379 0.034022 0.071432 0.08925 0.031036 0.037686 0.174735 0.059859 0.312748 0.197446 0.099592 0.065016 0.034798 0.066825 0.330974 0.076472 0.154808 0.090546 1.06E-05 0.043512 0.00602 0.191886 2.33E-05 0.026082 0.091036 0.156121 0.051203 0.0330000 0.0366740 0.0875277 0.0352315 0.0354732 0.0754414 0.0456306 0.0768533 0.0748603 0.0771518 0.0365880 0.0183966 0.0777898 0.0927122 0.0310653 0.0559470 0.0349525 0.0000000 0.0246574 0.0146491 0.1088924 0.0000000 0.0347675 0.0378552 0.0909357 0.0374694 Table 3.10.B. (cont’d) Curimana Nueva Requena Puente Piedra San Isidro El Agustino Campoverde Villa El Salvador Villa Maria del Triunfo San Luis San Juan de Lurigancho Irazola Carabayllo La Molina Padre Abad Santa Anita Coayllo Cieneguilla Morales Tarapoto Ate La Banda de Shilcayo Ica La Tinguina Parcona Llochegua Sivia Chaclacayo Zuðiga 181 183 187 195 200 203 204 210 214 222 228 238 262 275 285 285 287 290 342 378 418 432 463 472 540 561 685 827 86 29 97 37 86 204 235 196 21 596 235 77 61 203 131 41 9 249 877 386 212 849 189 317 62 46 26 40 0.113677 0.020237 0.048774 0.000728 0.035222 1.54E-05 0.1016307 0.0361466 0.104760 0.023793 0.007468 0.016745 0.053699 0.016058 0.097876 0.02323 0.094072 0.025181 0.096697 0.022516 0.000000 0.000728 0.077806 0.012654 0.103495 0.026096 0.122686 0.05615 0.038612 0.023404 0.037559 0.016333 0.074712 0.0235 0.125841 0.00272 0.000000 0.000728 0.112885 0.029327 0.089918 0.01137 0.048069 0.010876 0.078471 0.025437 0.097347 0.015165 0.02842 0.102733 0.124654 0.02097 0.181293 0.023965 0.141068 0.02892 0.054737 0.054012 0.077173 0.027063 0.054911 0.010374 0.022176 0.110085 0.149005 0.099363 1.11E-05 0.095435 0.160033 0.242766 0.033413 0.054153 0.072345 0.000303 4.77E-06 0.214154 0.113372 0.045659 0.137176 0.19524 0.152651 0.139401 0.035606 0.038472 0.075851 0.029296 0.0607501 0.0270461 0.0422409 0.1034860 0.0405010 0.0572620 0.0000000 0.0488270 0.1106470 0.0670984 0.0498861 0.0850321 0.0331790 0.0657050 0.0000000 0.0767559 0.0726437 0.0472055 0.0761008 0.0801667 0.0841299 0.0830169 0.1089162 0.0896776 0.0661602 0.0624724 6 2 12 1 7 24 25 26 0 48 24 11 4 8 15 2 0 29 83 20 14 96 30 46 11 7 3 3 129 Table 3.10.B. (cont’d) Lurigancho Santa Rosa de Quives Huanchay Catahuasi Pariacoto Magdalena Antioquia San Bartolome Huanuco Amarilis Putinza Pillco Marca Surco San Mateo de Otao Ambar Cochabamba Chumuch Tiabaya Asuncion Sachaca Jacob Hunter San Juan Socabaya Cortegana Jose Luis Bustamante Yanahuara Arequipa Cerro Colorado Miraflores 879 936 1067 1203 1264 1298 1573 1644 1921 1950 1985 1996 2049 2084 2084 2135 2202 2218 2254 2300 2309 2336 2352 2352 2389 2402 2429 2441 2450 142 51 30 64 29 16 39 63 729 529 25 235 54 37 55 16 15 23 30 35 69 15 119 14 138 34 124 145 78 14 6 5 8 3 1 4 8 38 39 2 16 11 3 7 1 2 5 5 7 13 2 14 2 11 2 6 27 17 130 0.080537 0.018058 0.040213 0.028682 0.103450 0.035502 0.088139 0.066812 0.072704 0.018196 0.045639 0.000728 0.250594 0.019025 0.03957 0.096362 0.042887 0.007843 0.058293 0.010826 0.049331 0.000728 0.047872 0.018715 0.107871 0.033662 0.055145 0.000728 0.067798 0.021807 0.006640 0.000728 0.098513 0.000728 0.200014 0.002372 0.102316 0.034771 0.194364 0.090855 0.130347 0.048337 0.020893 0.000728 0.140321 0.048971 0.092203 0.000728 0.049894 0.014695 0.085099 0.053965 0.038322 0.023117 0.143074 0.032133 0.223277 0.053676 0.046304 0.041956 0.037812 0.285684 0.009602 8.48E-06 0.014117 0.098643 0.04484 0.061997 1.32E-05 0.082311 0.061189 1.96E-05 0.026154 8.48E-06 7.95E-06 0.000129 0.03627 0.288914 0.161217 7.95E-06 0.28538 7.42E-06 0.029801 0.099014 0.066262 0.149713 0.224729 0.0643181 0.0648836 0.1148265 0.0890247 0.0773745 0.0292428 0.1445636 0.0898827 0.0669651 0.0692121 0.0422053 0.0705608 0.0924860 0.0726392 0.0886489 0.0233664 0.1182717 0.2444341 0.1043668 0.0745249 0.0776495 0.0535749 0.0714195 0.1728861 0.0550670 0.0408212 0.0232999 0.0799007 0.0764839 Table 3.10.B. (cont’d) Paucarpata Manas Mariano Melgar Lampian Cospan Abancay Alto Selva Alegre Cayma Arahuay Huasmin Bambamarca Jesus Pacaycasa Luricocha Tinco Marca Jose Galvez Tamburco Llacanora Celendin Sucre Carhuaz Sorochuco San Pedro de Pilas San Lorenzo De Quinti Huanta Los Baños del Inca Acopampa Cajamarca 2453 2457 2459 2467 2471 2500 2510 2531 2533 2543 2556 2568 2571 2598 2606 2615 2618 2620 2621 2629 2632 2663 2663 2678 2682 2685 2685 2692 2731 206 73 90 45 14 1565 128 143 39 61 221 45 12 36 40 15 16 181 15 31 28 64 41 41 62 78 74 28 1197 40 7 15 13 0 183 17 25 4 6 25 7 1 7 4 3 0 11 0 2 4 15 4 5 10 18 5 5 123 131 0.189008 0.048081 0.098725 0.012517 0.087979 0.02285 0.183524 0.015313 0.000000 0.000728 0.108660 0.009653 0.080756 0.022479 0.145982 0.039376 0.080101 0.012494 0.087633 0.016102 0.02618 0.087440 0.090665 0.038452 0.040790 0.000728 0.195352 0.020738 0.111455 0.05477 0.098188 0.000728 0.000000 0.000728 0.059515 0.012926 0.000000 0.000728 0.086481 0.066618 0.079366 0.041332 0.206867 0.076981 0.111423 0.0251 0.198081 0.016383 0.143245 0.040807 0.103737 0.0578 0.060533 0.02496 0.112468 0.028617 0.092653 0.009481 0.476219 0.011437 0.046989 0.010552 7.42E-06 0.145824 0.06468 0.22172 0.006088 0.015816 0.151473 0.066536 6.36E-06 0.015483 0.119989 7.95E-06 8.48E-06 0.030243 7.95E-06 0.137578 0.047833 0.379272 0.025831 0.011004 0.103242 0.260589 0.046102 0.02293 0.107602 0.0704244 0.0797613 0.0774522 0.1861476 0.0000000 0.0737195 0.0566242 0.0757226 0.0845180 0.0744675 0.0770769 0.0668614 0.0444103 0.1580106 0.0823937 0.1617417 0.0000000 0.0667397 0.0000000 0.0760677 0.0870447 0.0936068 0.0827519 0.1055394 0.1063442 0.0801333 0.0658663 0.1067876 0.0710803 Table 3.10.B. (cont’d) Namora Chugur Marcara Sangallaya San Juan Bautista Ayacucho Anta Miguel Iglesias Jesus Nazareno Jangas Pariahuanc Tarica Matara Yungar Oxamarca Ihuari Santo Domingo de Los Acos Vinchos Langa Yauyos Amashca Huayllapamp Carmen Alto Huachupamp La Libertad de Pallan San Miguel de Aco Carhuanca Llacllin Shilla 2765 2765 2767 2779 2786 2797 2800 2813 2817 2824 2830 2832 2834 2836 2836 2850 2861 2874 2889 2895 2905 2908 2921 2938 2952 2956 2980 3020 3036 44 13 65 39 255 792 30 15 127 59 15 42 16 26 14 100 43 47 53 84 14 12 125 21 27 32 16 10 42 1 0 17 5 46 162 4 2 18 9 5 6 4 4 1 19 7 4 6 19 1 2 28 2 5 5 3 0 10 132 0.020492 0.030445 0.000000 0.000728 0.224114 0.036593 0.155614 0.029532 0.164363 0.017593 0.172509 0.015437 0.068917 0.02427 0.077686 0.000728 0.108223 0.019189 0.095242 0.028115 0.172609 0.000728 0.165560 0.035404 0.320784 0.000728 0.0408 0.133534 0.033264 0.000728 0.120405 0.021176 0.103388 0.007331 0.039507 0.053492 0.103196 0.010656 0.154428 0.027885 0.059496 0.000728 0.060904 0.000728 0.242069 0.034994 0.019559 0.000728 0.143127 0.043031 0.129659 0.013395 0.197380 0.000728 0.000000 0.000728 0.250955 0.028794 0.040784 6.89E-06 0.087037 0.034013 0.078923 0.188742 0.017671 7.95E-06 0.046761 0.046637 7.95E-06 0.052646 8.48E-06 0.04328 7.42E-06 0.044842 0.002311 0.134486 0.006018 0.065314 7.42E-06 6.36E-06 0.153072 1.11E-05 0.049995 0.005741 8.48E-06 5.3E-06 0.034823 0.0595534 0.0000000 0.1085737 0.1053622 0.0840430 0.0858912 0.0795653 0.1448389 0.0782104 0.1036216 0.3064933 0.1048648 0.2441453 0.1027020 0.0507383 0.1084179 0.1045488 0.0920432 0.0952761 0.0863232 0.0580248 0.1495862 0.0823295 0.0658186 0.0956578 0.1612086 0.1020017 0.0000000 0.1436752 Table 3.10.B. (cont’d) Independenc Gorgor Concepcion Iguain Huaraz Encañada Tambillo Vischongo Ayauca Ocros Vinchos Huarochiri Caujul Huancapon Santiago De Pischa Acocro San Damian Santillana Paccho San Jose De Ticllas Madean Huambalpa Leoncio Prado Huamanguilla Quinua Huantan Viðac San Andres De Tupicoc Socos 3047 3049 3061 3063 3073 3087 3111 3150 3151 3153 3155 3170 3185 3187 3210 3251 3252 3265 3275 3282 3292 3294 3299 3300 3301 3315 3315 3321 3368 673 117 16 26 640 91 29 45 76 44 87 38 40 74 16 61 79 45 126 30 65 31 60 39 27 46 111 38 45 0.121834 0.012149 0.166361 0.026579 0.045639 0.000728 0.244365 0.021359 0.119571 0.014063 0.095605 0.036566 0.367439 0.0349 0.184673 0.072999 0.225058 0.049672 0.059239 0.035525 0.156922 0.035418 0.050611 0.003923 0.135313 0.017394 0.087316 0.046792 0.018726 0.000728 0.119461 0.028523 0.068111 0.035528 0.003 0.083911 0.131507 0.048415 0.135804 0.027525 0.245299 0.044587 0.153852 0.08989 0.059456 0.008867 0.186132 0.056974 0.061112 0.002148 0.237040 0.022989 0.099947 0.024159 0.098601 0.023551 0.099122 0.064175 0.099325 0.082651 8.48E-06 0.011861 0.126572 0.121674 0.035321 0.239796 0.187518 0.05553 0.109138 0.000585 0.012102 0.162024 8.48E-06 0.049626 0.099719 0.000405 0.295348 0.022728 0.129222 0.250485 0.004718 0.126593 0.000125 0.02431 0.064786 0.021076 0.185327 0.0858664 0.1117200 0.0527629 0.1630653 0.0832834 0.0694840 0.1747842 0.0939462 0.0852323 0.0948775 0.1034588 0.1170430 0.1471109 0.0918449 0.0477489 0.0931438 0.1008816 0.0663510 0.1371170 0.1158605 0.1336052 0.1069616 0.1255297 0.0908418 0.0897736 0.1425700 0.0995365 0.1080302 0.0849915 101 24 2 5 85 7 10 8 22 5 19 4 9 9 1 8 9 5 19 5 18 5 8 6 3 12 17 3 5 133 Table 3.10.B. (cont’d) Accomarca Colonia Pararin Recuay Azangaro Olleros Carampoma Ticapampa Vilcas Huaman Lincha Hualgayoc Chiara Santa Cruz de Andamar Pampas Chico Saurama Catac Huancaya Pira Independen Huaros Vitis Miraflores Laraos Lachaqui Pampas Ascension Huancavelica Simon Bolivar Chaupimarca 3387 3399 3402 3428 3435 3443 3459 3485 3494 3516 3530 3540 3550 3552 3574 3579 3591 3602 3606 3614 3625 3677 3683 3686 3698 3711 3746 4234 4373 15 101 13 15 68 31 25 32 60 36 46 45 109 14 16 14 40 45 31 88 23 37 22 43 16 366 1350 257 611 0.000000 0.000728 0.165006 0.026285 0.015644 0.000728 0.096058 0.000728 0.225303 0.021445 0.022988 0.019403 0.109409 0.000728 0.039959 0.027964 0.228409 0.065383 0.102883 0.030025 0.205148 0.039487 0.085554 0.041691 0.153664 0.012345 0.000000 0.000728 0.334581 0.000728 0.008768 0.000728 0.142419 0.014493 0.066806 0.015192 0.146897 0.017644 0.127416 0.037946 0.146478 0.000728 0.042581 0.001053 0.235806 0.000728 0.122780 0.066232 0.023194 0.000728 0.081539 0.015013 0.086282 0.010353 0.053885 0.017445 0.039955 0.010706 7.95E-06 0.069778 6.89E-06 7.95E-06 0.031272 0.011671 1.32E-05 0.025024 0.256498 0.032453 0.071722 0.078218 0.016612 7.42E-06 8.48E-06 7.42E-06 0.008401 0.010385 0.009651 0.12671 1.22E-05 4.1E-05 1.17E-05 0.188629 8.48E-06 0.082495 0.144691 0.078214 0.070032 0.0000000 0.1362414 0.0918526 0.1242041 0.1464752 0.0531877 0.1267116 0.0804289 0.0926997 0.1023138 0.0911615 0.0770576 0.1358383 0.0000000 0.4253369 0.0486578 0.1237612 0.0843879 0.1190456 0.0990310 0.1369873 0.0756507 0.0847409 0.0979619 0.0598082 0.0768907 0.0732809 0.0622410 0.0553758 0 22 1 2 16 2 4 3 16 6 10 5 24 0 8 1 6 5 5 15 4 4 2 9 1 40 132 18 32 134 Table 3.10.B. (cont’d) 0.060924 Yanacancha 2 Direct standardization methods adjusting for sex and age categories (i.e., 18-44, 45-64, >=65 years). 3 Replacing null values and using the lower plausible value obtained in other districts. 4 Empirical best linear unbiased prediction (EBLUP) with Fay-Herriot model, using auxiliary data: population density, poverty rates, and human development index at district level. 0.044289 0.010068 0.0622228 4394 601 32 135 Figure 3.7. The Output of the final Fay-Herriot model1 for obtaining EBLUP2 estimates. 1: Model using Fayherriot Stata© command, including human development index, poverty proportions, and population density for each district as auxiliary data. 2: Empirical best linear unbiased prediction (EBLUP). 136 . fayherriot y idh_2019 pct_pobreza_total pob_densidad_2020 , variance(var) eblup(EBLUPFinal) mse(MSEFinal) sigmamethod(reml) arcsin reps(999) Iteration 0: f(p) = -79563.708 Iteration 1: f(p) = 28.517064 Iteration 2: f(p) = 28.693705 Iteration 3: f(p) = 28.695044 Iteration 4: f(p) = 28.695044 Bootstrap ......50.....100.....150.....200.....250.....300.....350.....400.....450.....500.....550.....600.....650.....700.....750.....800.....850.....900.....950.... Sigma2_u estimation method: reml N in sample = 223 Transformation of depvar: arcsine N out of sample = 0 EBLUP and MSE bias correction: none Sigma2_u = 0.0169 Adj R-squared = 0.1397 FH R-squared = 0.8277 ------------------------------------------------------------------------------ y | Coefficient Std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- HDI_2019 | -.2748443 .1563296 -1.76 0.079 -.5812447 .0315561 pct_poverty | -.0018669 .0011864 -1.57 0.116 -.0041922 .0004584 pop_density | -3.80e-06 2.13e-06 -1.78 0.075 -7.99e-06 3.79e-07 _cons | .4709515 .0984947 4.78 0.000 .2779054 .6639975 ------------------------------------------------------------------------------ Shapiro-Wilk test for normality: Residuals e (standardized) V = 9.738 p-value = 0.000 Random effects u V = 41.711 p-value = 0.000 ------------------------------------------------------------------------------ Table 3.11. Simple linear regressions between altitude and three depressed mood (DM) estimates, using the “arcsine of the estimate square root” transformation1 for all three estimates. Data from the “Noguchi Surveys,”2 Peru, 2003-2013. Model A: with Weighted proportion estimates (direct estimates) Variable Altitude (m) Intercept St.Err. 5.710e-06 0.01351199 19.21 [95% Conf Interval] 0.00003801 0.28615298 β Coef. 0.00002676 0.25952414 p-value <0.00001 <0.00001 0.00001551 0.23289531 t-value 4.69 0.12224400 Cameron & Trivedi’s decomposition of IM-Test Source chi2 df Model metrics Mean dependent var 0.31118119 R-squared 0.09049599 SD dependent var Number of obs 223 Heteroskedasticity 5.34 F-test 21.98958416 Prob > F 0.00000480 Skewness Akaike crit. (AIC) - 322.6828677 Bayesian crit. (BIC) -315.8685242 Kurtosis Total 5.93 11.88 23.15 p 0.0692 0.0149 0.0006 0.0001 2 2 1 4 Model B: with Standardized prevalence proportions (direct standardization by sex and age) Variable Altitude (m) St.Err. 5.860e-06 β Coef. 0.00002 p-value <0.001 t-value 3.71 0.0000102 [95% Conf Interval] 0.00003327 Intercept 0.25526 0.01387362 18.40 <0.001 0.2279189 0.28260189 Model metrics Mean dependent var R-squared 0.29719 0.058556 F-test Akaike crit. (AIC) 13.7456794 -310.902991 Bayesian crit. SD dependent var Number of obs Prob > F (BIC) 0.12336821 Source chi2 df p Cameron & Trivedi’s decomposition of IM-Test 223 Heteroskedasticity 1.54 0.0002647 -04.0886472 Kurtosis Skewness Total 2.14 10.60 14.28 2 1 1 4 0.04631 0.01433 0.0011 0.0064 137 Table 3.11. (cont’d) Model C: with EBLUP estimates Variable Altitude (m) β Coef. 0.000026 St.Err. 4.480e-06 t-value 5.76 Intercept 0.222405 0.01061463 20.95 Model metrics Mean dependent var 0.27226918 SD dependent var 0.09822218 p-value <0.00001 <0.00001 0.000017 [95% Conf Interval] <0.000035 0.2014856 0.24332 Cameron & Trivedi’s decomposition of IM-Test Source chi2 df R-squared 0.13061499 Number of 223 Heteroskedasticity 5.99 obs F-test 33.20267926 Prob > F 0.00000003 Skewness Akaike crit. (AIC) -430.32216 Bayesian crit. (BIC) -423.507815 Kurtosis Total 9.03 6.73 14.28 2 1 1 4 1 asin(x)= the radian value of the arcsine of x. 2 “Noguchi Surveys”: epidemiological surveys from the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health. 138 p 0.500 0.0027 0.0095 0.0002 Figure 3.8. Output results of Fay-Herriot model, analysis excluding forty districts with missing values and infinitesimal values of the standard errors in the weighted proportion estimation of depressed mood (i.e., ‘direct estimator’). 139 [95% Conf Interval] [0.00001449, 0.00002372] [0.2661643, 0.28733823] Table 3.12. Simple Linear regression between altitude and Depressed Mood with EBLUP estimates omitting infinitesimal and missing values of standard errors, using the “arcsine of the estimate square root” transformation1 of the EBLUP. Data from the “Noguchi Surveys,”2 Peru, 2003-2013. Variable Altitude (m) St.Err. 0.00001911 2.340e-06 p-value <0.01 t-value 8.17 β Coef. Intercept 0.27675127 .0053655 51.58 <0.01 66.70103338 Prob > F -621.74862668 Bayesian crit. (BIC) 0.31159345 SD dependent var 0.26928040 Number of obs Model Metrics Mean dependent var R-squared F-test Akaike crit. (AIC) Cameron & Trivedi’s decomposition of the IM-test Chi2 Source 11.13 Heteroskedasticity 3.31 Skewness 3.37 Kurtosis Total 17.81 1: asin(x)= the radian value of the arcsine of x. 2: “Noguchi Surveys”: epidemiological surveys from the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health. 0.05135301 183 <0.00000001 -615.32965437 p-value 0.0038 0.0688 0.0665 0.0013 df 2 1 1 4 140 Figure 3.9. Output results of regressing altitude deciles as predictors of Depressed Mood EBLUP estimates. Same as Model 4 from Table 3.9 but omitting infinitesimal and missing values of standard errors. 141 _cons .0708802 .0048997 14.47 0.000 .061277 .0804834 _Ialt10_10 .0348755 .0075509 4.62 0.000 .020076 .049675 _Ialt10_9 .0403036 .007304 5.52 0.000 .025988 .0546191 _Ialt10_8 .0461259 .0074212 6.22 0.000 .0315806 .0606712 _Ialt10_7 .0363326 .0075509 4.81 0.000 .0215332 .0511321 _Ialt10_6 .0343465 .007304 4.70 0.000 .0200309 .048662 _Ialt10_5 .0297004 .0074212 4.00 0.000 .0151551 .0442457 _Ialt10_4 .0313505 .0070112 4.47 0.000 .0176088 .0450921 _Ialt10_3 .0104455 .007304 1.43 0.153 -.00387 .0247611 _Ialt10_2 .0004275 .0071003 0.06 0.952 -.0134888 .0143438 EBLUPWoMiss Coefficient std. err. z P>|z| [95% conf. interval] OIM Log likelihood = 435.9474305 BIC = -901.1497 AIC = -4.655163Link function : g(u) = u [Identity]Variance function: V(u) = 1 [Gaussian]Pearson = .0913694861 (1/df) Pearson = .0005281Deviance = .0913694861 (1/df) Deviance = .0005281 Scale parameter = .0005281Optimization : ML Residual df = 173Generalized linear models Number of obs = 183Iteration 0: log likelihood = 435.94743 i.alt10 _Ialt10_1-10 (naturally coded; _Ialt10_1 omitted). xi: glm EBLUPWoMiss i.alt10, link(identity) Figure 3.10. Altitude of districts in nine surveys across the years. Data from the “Noguchi Surveys,”1 Peru, 2003-2013. 1: “Noguchi Surveys”: epidemiological surveys from the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health. 142 APPENDIX C: CHAPTER 4 In this Appendix, I will show two Stata© outputs with results mentioned in Chapter 4. First, the output results from the "Nestreg" command fitting nested models by sequentially adding blocks of variables and then reports comparison tests between the nested models included in Table 4.3. Then, in Figure 4.3, I will show the output from the covariate-adjusted logistic regression analysis of model 6 in Table 4.3. Finally, in Table 4.5, several results of survey logistic regressions between different altitude quantiles or categories and depressed mood will be presented. Figure 4.2. The output of Nested model statistics using the "Nestreg" command for model 6 from Table 4.3. 143 9 6.40 1 2855 0.0115 8 4.95 1 2855 0.0262 7 12.67 1 2855 0.0004 6 0.23 1 2855 0.6330 5 10.80 1 2855 0.0010 4 154.69 1 2855 0.0000 3 42.09 1 2855 0.0000 2 48.18 1 2855 0.0000 1 8.42 1 2855 0.0037 Block F df df Pr > F Block Design Figure 4.3. Estimated associations from covariate-adjusted analysis with depressed mood (DM) expressed as a function of each covariate included in model 6. Data from the "Noguchi Surveys,"1 Peru, 2003-2013. 1 "Noguchi Surveys": epidemiological surveys from the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health. 144 _cons -3.116399 .2675886 -11.65 0.000 -3.641085 -2.591713_Ilong_resi_1 -.6217411 .2457789 -2.53 0.011 -1.103663 -.139819_Icivil_sta_3 -.2840387 .1256191 -2.26 0.024 -.530352 -.0377253_Icivil_sta_2 .3160335 .0997787 3.17 0.002 .1203879 .511679_Imater_ton_3 -.3936253 .7313046 -0.54 0.590 -1.827564 1.040313_Imater_ton_2 .3619393 .1179571 3.07 0.002 .1306496 .5932289 _Ifemale_1 1.146587 .1007094 11.39 0.000 .9491167 1.344058 age .0093971 .0028097 3.34 0.001 .0038878 .0149064_Ialtitude__3 .4736971 .083853 5.65 0.000 .3092786 .6381156_Ialtitude__2 .3095751 .093563 3.31 0.001 .1261173 .493033 depress_mood Coefficient std. err. t P>|t| [95% conf. interval] Linearized Prob > F = 0.0000 F(9, 2847) = 29.53 Design df = 2,855Number of PSUs = 2,877 Population size = 9,816,114Number of strata = 22 Number of obs = 33,013Survey: Logistic regression(running logistic on estimation sample)i.long_resid _Ilong_resi_0-1 (naturally coded; _Ilong_resi_0 omitted)i.civil_stat3c _Icivil_sta_1-3 (naturally coded; _Icivil_sta_1 omitted)i.mater_tong3 _Imater_ton_1-3 (naturally coded; _Imater_ton_1 omitted)i.female _Ifemale_0-1 (naturally coded; _Ifemale_0 omitted)i.altitude_ter _Ialtitude__1-3 (naturally coded; _Ialtitude__1 omitted). xi: svy: logistic depress_mood i.altitude_ter age i.female i.mater_tong3 i.civil_stat3c i.long_resid, coef Table 4.5. Survey Logistic Regressions for associations between different altitude categories and depressed mood. Data from the "Noguchi Surveys,"1 Peru, 2003-2013. Survey Logistic regression: altitude continuous variable t-value St.Err. 6.03 0.00003 0.05201 -48.81 0.11273 SD dependent var p-value p<0.001 33080 F-test [95% Conf. Interval] 0.00013 -2.64075 0.00025 -2.43678 0.31626 36.32610 Coef. 0.0002 -2.53876 Coef. St.Err. 0.294 0.509 -2.55323 Altitude Constant Mean dependent var Number of obs Survey Logistic regression: altitude tertiles 3 quantiles of altitude 2nd 3rd Constant Mean dependent var Number of obs Survey Logistic regression: altitude quintiles 5 quantiles of altitude 2nd 3rd 4th 5th Constant Mean dependent var Number of obs 0.018 0.299 0.445 0.476 -2.52199 St.Err. Coef. t-value p-value [95% Conf. Interval] 3.27 0.08979 6.94 0.07337 -43.70 0.05843 0.11273 SD dependent var 0.001 p<0.001 33080 F-test 0.118 0.365 -2.6678 0.469 0.653 -2.43867 0.31626 24.16555 t-value p-value [95% Conf. Interval] 0.17 0.10752 2.75 0.1089 4.28 0.10401 4.78 0.09964 -35.83 0.0704 0.11273 SD dependent var 0.86 p<0.01 p<0.001 p<0.001 33080 F-test -0.192 0.086 0.241 0.281 -2.660 0.229 0.513 0.649 0.671 -2.383 0.31626 9.47256 p-value [95% Conf. Interval] Coef. St. Err. t-value Survey Logistic regression: altitude deciles 10 quantiles of altitude 2nd 3rd 4th 5th 6th 7th 8th 9th 10th Constant Mean dependent var Number of obs -0.08312 -0.18617 0.10567 -0.05839 0.51303 0.27388 0.63507 0.70497 -0.2247 -2.47724 -0.60 0.13893 -1.24 0.14986 0.80 0.13246 -0.45 0.13072 3.62 0.14178 1.94 0.14089 5.95 0.10671 6.06 0.11625 -1.87 0.12029 0.08093 -30.61 0.11273 SD dependent var .54971 .21422 .42509 .65512 p<0.001 0.052 p<0.001 p<0.001 0.06186 33080 F-test -.35553 -.48001 -.15406 -.31471 .23502 -.00237 .42584 .47703 -.46056 -2.63592 .1893 .10767 .3654 .19792 .79104 .55013 .84431 .93291 .01116 -2.31856 0.31626 14.95134 145 Table 4.5. (cont'd) Survey Logistic regression: altitude in four categories from the International Society of Mountain Medicine Coef. St.Err. t-value p-value [95% Conf. Interval] Altitude in 4 categories (m) 1500-2499 2500-3499 ≥3,500 Constant Mean dependent var Number of obs 0.51219 0.4737 0.21141 -2.51295 4.03 p<0.0001 0.12703 5.97 p<0.0001 0.07934 1.53 0.127 0.13832 0.04934 -50.93 0.11273 SD dependent var 33080 F-test 0.26312 0.31813 -0.0598 -2.6097 0.76127 0.62927 0.48262 -2.41619 0.31626 14.23816 Coef. St.Err. t-value Survey Logistic regression: High Altitude definition from International Society of Mountain Medicine Altitude in HA categories (m) 0.53114 ≥2,500 -2.36902 Constant 0.31626 Mean dependent var Number of obs 28.48359 33080 F-test 1: "Noguchi Surveys": epidemiological surveys from the "Honorio Delgado - Hideyo Noguchi" Peruvian National Institute of Mental Health. 0.07278 0.04557 0.11273 SD dependent var 0.38843 -2.45837 0.24572 -2.54773 [95% Conf. Interval] 5.34 p<0.0001 p-value -53.95 146 APPENDIX D: CHAPTER 5 Figure 5.1. Stata© output of the model with coca leaf use included. Altitude is in three categories. Beta Coefficients shown. 147 _cons -3.138 .2703271 -11.61 0.000 -3.668056 -2.607944 Yes .8001697 .1650433 4.85 0.000 .4765535 1.123786 cocause_lastm Yes -.5986412 .2493677 -2.40 0.016 -1.0876 -.1096822 long_resid single -.2788178 .1257789 -2.22 0.027 -.5254445 -.0321911 divorc/widow .319238 .1002576 3.18 0.001 .1226535 .5158226 civil_stat3c other -.3933291 .7346649 -0.54 0.592 -1.833857 1.047198Quechua/Aymara .3105942 .1231856 2.52 0.012 .0690524 .552136 mater_tong3 1.female 1.158246 .1014215 11.42 0.000 .9593789 1.357112 age .0092437 .0028219 3.28 0.001 .0037106 .0147769 3 .3632523 .0879249 4.13 0.000 .1908495 .5356552 2 .2703481 .0938718 2.88 0.004 .0862847 .4544115 altitude_ter depress_mood Coefficient std. err. t P>|t| [95% conf. interval] Linearized Design df = 2,855Number of PSUs = 2,877 Population size = 9,787,461Number of strata = 22 Number of obs = 32,962Survey: Generalized linear models(running glm on estimation sample)> i.cocause_lastm, family(binomial) link(logit). svy: glm depress_mood i.altitude_ter c.age i.female i.mater_tong3 i.civil_stat3c i.long_resid Figure 5.2. Stata © output of the model with coca-leaf use and product terms included. Altitude is in three categories. 148 _cons -3.146464 .2715267 -11.59 0.000 -3.678872 -2.614056 3#Yes -.2769698 .5688402 -0.49 0.626 -1.392349 .8384093 2#Yes .1944652 .5968856 0.33 0.745 -.9759053 1.364836altitude_ter#cocause_lastm Yes .7435659 .5571587 1.33 0.182 -.3489081 1.83604 cocause_lastm Yes -.5895157 .2505725 -2.35 0.019 -1.080837 -.0981943 long_resid single -.2782146 .1258165 -2.21 0.027 -.5249151 -.0315142 divorc/widow .3201676 .1002891 3.19 0.001 .1235212 .5168141 civil_stat3c other -.3905772 .7352566 -0.53 0.595 -1.832265 1.05111 Quechua/Aymara .3145321 .1230337 2.56 0.011 .0732882 .5557759 mater_tong3 1.female 1.156223 .1017639 11.36 0.000 .956685 1.355761 age .0092608 .0028237 3.28 0.001 .0037241 .0147974 3 .4208586 .0846753 4.97 0.000 .2548276 .5868895 2 .2596872 .0952188 2.73 0.006 .0729826 .4463918 altitude_ter depress_mood Coefficient std. err. t P>|t| [95% conf. interval] Linearized Design df = 2,855Number of PSUs = 2,877 Population size = 9,787,461Number of strata = 22 Number of obs = 32,962Survey: Generalized linear models(running glm on estimation sample)> i.cocause_lastm i.altitude_ter#i.cocause_lastm , family(binomial) link(logit). svy: glm depress_mood i.altitude_ter c.age i.female i.mater_tong3 i.civil_stat3c i.long_resid Figure 5.3. Stata © output of the model with coca-leaf use and product terms included. Altitude is in two categories. Coefficients shown. 149 _cons -3.066912 .2650815 -11.57 0.000 -3.586683 -2.547142 2#Yes .2942971 .3873822 0.76 0.447 -.4652801 1.053874altitude_2grp#cocause_lastm Yes .4510093 .3467164 1.30 0.193 -.2288307 1.130849 cocause_lastm Yes -.6149051 .2473904 -2.49 0.013 -1.099987 -.1298232 long_resid single -.2959775 .1260187 -2.35 0.019 -.5430743 -.0488807 divorc/widow .3190803 .1008701 3.16 0.002 .1212947 .5168658 civil_stat3c other -.4013506 .7520063 -0.53 0.594 -1.875881 1.07318 Quechua/Aymara .322034 .122373 2.63 0.009 .0820856 .5619823 mater_tong3 1.female 1.153983 .1016222 11.36 0.000 .9547231 1.353244 age .0090015 .002805 3.21 0.001 .0035015 .0145015 2.altitude_2grp .4288299 .0933748 4.59 0.000 .2457411 .6119188 depress_mood Coefficient std. err. t P>|t| [95% conf. interval] Linearized Design df = 2,855Number of PSUs = 2,877 Population size = 9,787,461Number of strata = 22 Number of obs = 32,962Survey: Generalized linear models(running glm on estimation sample)> i.cocause_lastm i.altitude_2grp#i.cocause_lastm , family(binomial) link(logit). svy: glm depress_mood i.altitude_2grp c.age i.female i.mater_tong3 i.civil_stat3c i.long_resid