HERD VARIABLES ASSOCIATED WITH MILKING EFFICIENCY OF DAIRY CATTLE   

By 

Rhyannon Moore-Foster 

 

 

 

 

 

 

 

 

 

A DISSERTATION 

Submitted to  

Michigan State University  

in partial fulfillment of the requirements  

for the degree of  

Comparative Medicine and Integrative Biology – Doctor of Philosophy 

 

2018 

 

 

 

 

 

 

 

 

HERD VARIABLES ASSOCIATED WITH MILKING EFFICIENCY OF DAIRY CATTLE 

ABSTRACT 

By 

Rhyannon Moore-Foster 

Over the past two decades, there has been a marked shift in herd size distribution among dairy 

farms in the U.S.; farms with fewer than 100 cows accounted for 49% of the country’s milk cows 

in 1992, but just 17% of milk cows in 2012. In contrast, farms with at least 1,000 cows 

accounted for 49% of all milk cows in 2012, an increase from 10% in 1992.  As overall herd size 

increases, dairy farms are also becoming more reliant on the hiring of non-family labor. 

However, greater reliance on employees has raised new challenges for training and compliance 

of critical farm protocols, including milking cows. Improper milking can impact the health of 

dairy cattle, particularly with respect to mastitis. Bovine mastitis is one of the most important 

diseases of dairy cattle in the United States and continues to cause major economic losses to the 

dairy industry, due to decreased farm productivity and quality of dairy foods. Crucially, the 

capacity of farms to train and educate personnel may play a key role in mastitis control and 

milking efficiency. However, many farms have not adapted management strategies to address 

changes in the employee training and education landscape. This ‘cultural lag’ in the dairy 

industry has warranted further study of the labor culture on dairy farms, and to find the means to 

assess employee performance and success of employee training, as measured by the milking 

efficiency. 

Therefore, the hypothesis of Chapter 2 was that employee and management factors affect 

milking efficiency of cows as measured by the proportion of cows within a herd with delayed 

milk ejection. Results of this study showed that herd size and mean stimulation time during pre-

 

milking preparation was a significant factor in the proportion of cows with delayed milk ejection. 

While past research indicates that stimulation time is important for proper milk ejection, it is a 

novel concept that stimulation time has the possibility of overriding many other factors that may 

play a role. However, this did not fully explain the employee factors that could potentially affect 

milking efficiency. To further investigate this relationship, Chapter 3 hypothesized that 

employee and management factors had an effect on the mean stimulation time during the pre-

milking protocol. The results of that study showed that increasing lag time during the pre-

milking routine was positively associated with increasing mean stimulation time. However, herd 

size and the number of passes during the pre-milking routine was negatively associated with 

mean stimulation time. These results suggest that employees who were expected to complete 

more tasks in a shorter amount of time were less likely to properly stimulate teats before milking, 

which as determined in Chapter 2, then leads to greater frequency of delayed milk ejection. 

In addition to delayed milk ejection, overmilking of cows (the time period when milking 

units are attached to teats with no milk flowing) is a critical measure of milking efficiency. 

Overmilking not only increases the amount of time that cows are unable to rest and eat, but also 

exposes teats to high vacuum levels and subsequently, increases the risk of mastitis. Chapter 4 

hypothesized that there were employee and management factors that are associated with median 

overmilking time in a herd. The results of this study found that herds with a shorter duration of 

daily milking periods had a greater proportion of overmilked cows. These results suggested that 

farm owners or operators who milk cows on farms that do not attain the full daily capacity of the 

milking equipment, may be making subjective decisions to milk cows inefficiently from 

overmilking.  

 

 

Copyright by 
RHYANNON MOORE-FOSTER 
2018 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

This dissertation is dedicated to my family, especially my parents, Chuck and Terri Moore. Most 

importantly also to my husband, Erik Foster. Who have all supported me unconditionally 

throughout this journey.  

Thank you for everything, I could not have done it without you.  

 

 

 

 

 

 

 

 

 

v 

 

 

 

 

 

 

A special thanks to:  

ACKNOWLEDGEMENTS 

 
 
 

(cid:1)  My mentor and major advisor, Dr. Ronald J. Erskine, for all of the guidance, mentorship 

and life advice. I appreciate your encouragement and support through this entire process.  

(cid:1)  Dr. Paul Bartlett, for the mentorship in statistics. 

(cid:1)  Dr. Bo Norby, for the mentorship in SAS and statistics.  

(cid:1)  Dr. Rebecca Schewe, for the mentorship in rural sociology and survey design.  

(cid:1)  Dr. Roger Thomson, for the mentorship in milking machine systems and parlor 

evaluations. 

(cid:1)  Trevor Lloyd-James, Caitlin McNichols, Leah Girard and Ellen Launstein for their 

assistance in data collection and entry.  

(cid:1)  The United States Department of Agriculture National Institute of Food and Agriculture 

(Agriculture and Food Research Initiative Competitive Grant no. 2013-68003-20339) for 

funding this study and my research.  

 

 

 

 

 

 

 

vi 

 

TABLE OF CONTENTS 

 

LIST OF TABLES  ...................................................................................................................................... ix 

LIST OF FIGURES  ..................................................................................................................................... x 

CHAPTER 1 MILK QUALITY AND MASTITIS, THE ROLE OF DAIRY EMPLOYEES AND THE 
CULTURAL LAG IN TRAINING – A LITERATURE REVIEW .............................................................. 1 
The Impact of Mastitis .................................................................................................................................. 2 
Mastitis Control Practices ............................................................................................................................. 7 
Milk Quality and Changes in the Dairy Demographic Landscape................................................................ 9 
Employee Behaviors and Milk Quality ....................................................................................................... 15 
Conclusion .................................................................................................................................................. 19 
REFERENCES ........................................................................................................................................... 21 

CHAPTER 2 DELAYED MILK EJECTION IN MICHIGAN DAIRY HERDS ...................................... 29 
Abstract ....................................................................................................................................................... 30 
Introduction ................................................................................................................................................. 31 
Materials and Methods ................................................................................................................................ 32 
Dairy Farm Selection ..................................................................................................................... 32 
 
Herd Profile and Management Culture .......................................................................................... 33 
 
 
Milking Dynamics and Parlor Behaviors ....................................................................................... 33 
 
Parlor Ergonomics ......................................................................................................................... 35 
Statistical Analysis ...................................................................................................................................... 37 
Dependent Variable ......................................................................................................................... 37 
 
 
Independent Variables ..................................................................................................................... 37 
Scale Factor Analysis ...................................................................................................................... 38 
 
Bivariate Analysis ........................................................................................................................... 38 
 
 
Multiple Linear Regression ............................................................................................................. 38 
Results ......................................................................................................................................................... 39 
Discussion and Conclusions ....................................................................................................................... 41 
Summary ..................................................................................................................................................... 44 
Acknowledgements  .................................................................................................................................... 46 
APPENDICES ............................................................................................................................................ 47 
Appendix A. Tables  ...................................................................................................................... 48 
 
 
Appendix B. Figures  ..................................................................................................................... 54 
REFERENCES ........................................................................................................................................... 56 

CHAPTER 3 STIMULATION TIME IN MICHIGAN DAIRY HERDS .................................................. 64 
Abstract ....................................................................................................................................................... 65 
Introduction ................................................................................................................................................. 66 
Materials and Methods ................................................................................................................................ 67 
Dairy Farm Selection ..................................................................................................................... 67 
 
Herd Profile and Management Culture .......................................................................................... 68 
 
Milking Dynamics and Parlor Behaviors ....................................................................................... 68 
 
 
Parlor Ergonomics ......................................................................................................................... 69 
Statistical Analysis ...................................................................................................................................... 70 
 
Dependent Variable ......................................................................................................................... 70 

 

vii 

 

Independent Variables ..................................................................................................................... 70 
 
Scale Factor Analysis ...................................................................................................................... 70 
 
Bivariate Analysis ........................................................................................................................... 71 
 
 
Multiple Linear Regression ............................................................................................................. 71 
Results ......................................................................................................................................................... 72 
Discussion and Conclusions ....................................................................................................................... 73 
Summary ..................................................................................................................................................... 76 
Acknowledgements ..................................................................................................................................... 77 
APPENDICES ............................................................................................................................................ 78 
Appendix A. Tables  ...................................................................................................................... 79 
 
 
Appendix B. Figures  ..................................................................................................................... 82 
REFERENCES ........................................................................................................................................... 83 

CHAPTER 4 OVERMILKING IN MICHIGAN DAIRY HERDS ............................................................ 85 
Abstract ....................................................................................................................................................... 86 
Introduction ................................................................................................................................................. 87 
Materials and Methods ................................................................................................................................ 88 
Dairy Farm Selection ..................................................................................................................... 88 
 
Herd Profile and Management Culture .......................................................................................... 88 
 
 
Milking Vacuum Dynamics and Parlor Behaviors ........................................................................ 89 
 
Parlor Ergonomics ......................................................................................................................... 90 
Statistical Analysis ...................................................................................................................................... 91 
Dependent Variable ......................................................................................................................... 91 
 
Independent Variables ..................................................................................................................... 91 
 
 
Scale Factor Analysis ...................................................................................................................... 92 
Bivariate Analysis ........................................................................................................................... 92 
 
 
Multiple Linear Regression ............................................................................................................. 92 
Results ......................................................................................................................................................... 93 
Discussion and Conclusions ....................................................................................................................... 94 
Summary ..................................................................................................................................................... 97 
Acknowledgements ..................................................................................................................................... 98 
APPENDICES ............................................................................................................................................ 99 
 
Appendix A. Tables  .................................................................................................................... 100 
 
Appendix B. Figures  ................................................................................................................... 102 
REFERENCES  ........................................................................................................................................ 105 

CHAPTER 5 CONCLUSION  .................................................................................................................. 114 

 

 

 

 

 

 

 

 

viii 

 

LIST OF TABLES 

 

 

Table 1. Descriptive data of herd profile and management culture variables from 64 Michigan 
dairy herds ......................................................................................................................................47 

Table 2. Bivariate analysis results for explanatory variables of herd-level proportion of cows 
with delayed milk ejection in 64 Michigan dairy herds .................................................................51 

Table 3. Final linear model for associations between delayed milk ejection and herd-level 
variables in 64 Michigan dairies ....................................................................................................52 

Table 4. Comparison of means for eligible independent variables for multivariable model 
between small (<300 cows) and large herds (≥300 cows) in 64 Michigan dairy herds.................53 

Table 5. Bivariate analysis results for explanatory variables of mean total stimulation time in the 
premilking routine in 64 Michigan dairy herds .............................................................................79  

Table 6. Final linear model for associations between mean total stimulation time during the 
premilking routine and herd-level variables in 64 Michigan dairies (herd size included) ............80 

Table 7. Comparison of means for eligible independent variables for multivariable model 
between small (<300 cows) and large herds (≥ 300 cows) in 64 Michigan dairy herds................81 

Table 8. Bivariate analysis results for explanatory variables of herd-level median duration of 
overmilking in 64 Michigan dairy herds ......................................................................................100 

Table 9. Final linear model for associations between median overmilking time and herd-level 
variables in 64 Michigan dairies ..................................................................................................101 

 

 

 

 

 

 

 

 

ix 

 

LIST OF FIGURES 

 

 

Figure 1. The upper plot is an example of VaDia® digital recording showing uninterrupted milk 
ejection. Vacuum is recorded (kpa) on the vertical axis. Each hash mark indicates a 15 s time 
interval on the horizontal axis.  Channel 1 (red) is indicates rear mouthpiece chamber vacuum, 
channel 2 (blue) front mouthpiece chamber vacuum, and channel 3 (green) short milk tube 
vacuum.  Symbols mark the start of milking (↓), start of incline phase of milk flow (▲), and 
end of milking (■). The lower plot is an exxample of VaDia® digital recording showing bimodal 
milk ejection. Vacuum is recorded (kpa) on the vertical axis. Each hash mark indicates a 15 s 
time interval on the horizontal axis. Channel 1 (red) indicates the rear mouthpiece chamber 
vacuum, channel 2 (blue) indicates the front mouthpiece chamber vacuum, channel 3 (green) 
indicates the short milk tube vacuum as a proxy for cluster vacuum.  Symbols mark the start of 
milking (↓), start of bimodal ejection (●), start of incline phase of milking (▲), and end of 
milking (■) .....................................................................................................................................54 

Figure 2. Histogram of time period between cluster attachment and attaining the incline phase of 
milk ejection (Time to milk let down) for 3,824 cows as recorded by digital vacuum .................55 

Figure 3. Histogram of mean total stimulation time (s) for 64 Michigan dairy herds ...................82 

Figure 4. Example of digital vacuum recording demonstrating overmilking. The vertical axis 
indicates vacuum (kPa). . The horizontal axis indicates time after cluster attachment divided into 
15 s intervals. Channels 1 and 2 (white arrow) represent the rear and front mouthpiece chambers, 
and channel 3 (black arrow) the short milk tube as a proxy for cluster vacuum.  Symbols mark 
the start of milking (◊), start of overmilking (▲) and end of milking (∆) ..................................102 

Figure 5. Histogram of duration of overmilking (s) of individual animals in 64 Michigan dairy 
herds as recorded by digital vacuum (n=3,824 cows)..................................................................103 

Figure 6. Correlation of median overmilking time (s) on horizontal axis vs. median unit on time 
by herd on the vertical axis (n=64) ..............................................................................................104 

 

 

 

x 

CHAPTER 1 

MILK QUALITY AND MASTITIS, THE ROLE OF DAIRY EMPLOYEES AND THE 

CULTURAL LAG IN TRAINING – A LITERATURE REVIEW 

 

 

 

R. Moore-Foster*, and R. J. Erskine*1 
 
 
 

*Department of Large Animal Clinical Sciences  
Michigan State University 
East Lansing, 48824 
 
 
 
 
1Corresponding Author 
R. J. Erskine 
736 Wilson Rd 
East Lansing, 48824 
517-355-9593 
Fax: 517-432-1042 
erskine@msu.edu  

 

 

 

 

 

 

 

 

1 

 

The Impact of Mastitis 

Bovine mastitis is one of the most important diseases of dairy cattle in the United States 

(USDA 2007) and continues to cause major economic losses to the dairy industry due to 

decreased farm productivity and quality of dairy foods (Ma et al., 2000; Losinger, 2005; Cha et 

al., 2011; Hogeveen et al., 2011). The National Mastitis Council estimates that reduced milk 

production accounts for 70% of the total loss associated with mastitis (NMC, 2013).  While quite 

variable, losses from clinical mastitis range from $95-200/case (Cha et al., 2011) with lost milk 

production from clinical cases ranging from 0.37 kg/cow/day to 3.7 kg/cow/day (Bartlett et al., 

1990; Seegers et al., 2003; Hagnestam-Nielsen et al., 2009; Halasa et al., 2009). However, 

additional losses result from discarded milk, treatment costs, and replacement costs (Gruet, 2001; 

Erskine et al., 2003; Hand et al., 2012; Heikkilä et al., 2012). Clinical mastitis is commonly 

identified by farm personnel due to symptoms such as abnormal milk, swollen quarters or 

systemic illness. However, most intramammary infections (IMI)  are often subclinical, i.e., there 

are no visually detectable changes in milk or the udder (Royster and Wagner, 2015). Thus, 

subclinical IMIs are more difficult to detect and only 25-30% of chronic subclinical infections 

show clinical signs (Barlow et al., 2009).  

Somatic cell counts (SCC) have been used for decades as a general indicator of udder 

health (Dohoo and Meek, 1982) and are comprised of a mixture of white blood cells including 

neutrophils and macrophages, as well as epithelial cells (Fox, 2013). Although many factors can 

affect individual cow SCC over the course of her lactation including days in milk, stress levels, 

seasonality and natural day to day variation, the most important variable is the presence of IMI in 

the udder (Dohoo and Meek, 1982; Archer et al., 2013).  Likewise, herd level or bulk tank SCC 

(BTSCC) are highly correlated with the prevalence of IMI in a dairy herd (Erskine et al., 1987).   

 

2 

 

Phagocytes are the bovine mammary gland’s primary response to infection and 

inflammation (Kehrli and Shuster, 1993). A threshold of 200,000 cells/mL of milk has been used 

as a benchmark to distinguish between an uninfected and infected udder (Dohoo and Meek, 

1982; Schepers et al., 1997; Fox, 2013). Pioneering work by Raubertas and Shook (1982) found 

that there was a direct linear relationship between milk yield loss and the log 2 transformation of 

somatic cell count (SCC), with estimates of 0.7 kg of milk lost per day in older cows (0.35 kg in 

first lactation cows) for each doubling of SCC, or unit increase in linear score.  Thus, a cow with 

a SCC of 200,000 cells/mL (linear score of 4) would lose 0.7 kg of milk per day as compared to 

a cow with a SCC of 100,000 cells/mL (linear score of 3).  However, more recent studies have 

re-examined the relationship between SCC and milk production losses and may better reflect the 

current production, management, and genetics of dairy operations than 30 years ago.   

Dairy Herd Improvement (DHI) records from over 2,800 dairy herds in Ontario were 

reviewed to determine the relationship between SCC and milk production (Hand et al., 2012).  

Similar to the Raubertas and Shook report, the Canadian study also found a positive association 

between daily milk loss and SCC; but the milk loss associated with increased SCC may be 

greater than previously suggested, especially in first lactation animals. Lactation milk loss 

increased from 165 to 919 kg over a lactation as average SCC increased from 100,000 cells/mL 

to 1,500,000 cells/mL. Unlike the Raubertas and Shook model, the impact of subclinical mastitis 

on milk yield loss, both in terms of kg of milk per day and percent of daily milk yield, increased 

with parity and milk production. Thus, the higher the milk production of a cow, the greater the 

losses related to increased SCC (Hand et al., 2012).  

In 2015, analysis of 164,423 individual cow records in Washington, Oregon and Idaho 

showed that cows with a somatic cell count (SCC) > 200,000 cells/mL at the first DHI test date 

 

3 

 

produced about 700 kg less milk than their herd mates with a SCC < 200,000 cells/mL over the 

course of the ensuing lactation (Kirkpatrick, 2015). Additionally, cows with high SCC at the first 

test date are 2.5 times more likely to have a case of clinical mastitis, as well as three times 

greater risk of being culled within the first 60 days in milk.  

At a herd level, increased mastitis also results in lost income from decreased premiums or 

increased  penalties  that  are  applied  relative  to  bulk  tank  SCC  (BTSCC).  BTSCC  greater  than 

200,000 cells/mL are associated with decreased productivity per cow, milk shelf-life, and  yield 

of raw milk products such as cheese (Klei et al., 1998; Ma et al., 2000). On a national level, 

reduced  milk  production  associated  with  BTSCC  ≥  200,000  cells/mL  results  in  an  annual 

economic loss of $ 3.1 billion to consumers (Losinger, 2005).  As of March 31, 2012, the USDA 

required that dairy exports to the European Union (E.U.) originate from farms with BTSCC at or 

below 400,000 cells/mL. 

Recently, studies have also linked clinical mastitis with reduced reproductive fertility, 

reporting between 17 to 52 additional open (or non-pregnant) days in cows with a case of clinical 

mastitis (Ahmadzadeh et al., 2009; Kirkpatrick, 2015). Loeffler et. al (1999) reported a >50% 

reduction in pregnancy risk three weeks after artificial insemination if cows had a clinical 

mastitis cases within 3 weeks after breeding. Fuenzalida et. al (2015) also found a decrease in the 

odds of pregnancy for cows that experienced a case of clinical mastitis during the breeding risk 

period (BRP) that was defined as 3 days prior to artificial insemination to 32 days after artificial 

insemination. Pregnancy rates during the BRP were 0.56, 0.67 and 0.75 for cows that had a first 

episode of clinical mastitis, first episode of subclinical mastitis, or were uninfected, respectively.  

Mastitis is the most common reason for antimicrobial drug use on dairy farms for adult 

dairy cattle (Wagner and Erskine, 2009). In 2014, an estimated 21.7% of the approximately 9 

 

4 

 

million cows in the U.S. were treated with antimicrobials for this disease (USDA, 2014), 

equating to over 1.9 million mastitis cases treated annually. Estimates of the percent of farms 

that use intramammary (IM) infusions varies from 90 to 98% (Zwald et al., 2004; USDA 2009).  

The use of IM dry cow treatments is also routinely used on many dairy farms to help control 

mastitis infection rates over the dry period. In a recent survey of 628 dairy herds, 75 % of herds 

reported use of blanket dry cow therapy (Schewe et al., 2015).  In a Wisconsin study, about 80% 

of all antimicrobial drugs used were for treatment or prevention of mastitis, which included dry 

cow therapy (Pol and Ruegg, 2007). In a Canadian study, IM administration of antimicrobials 

was estimated to account for 35% of all antimicrobial use on dairy farms, which was lower than 

antimicrobials administered systemically (38%; Saini et al., 2012).  However, in this study, the 

accounting of antimicrobial drug use was not just limited to adult dairy cattle.  

Several studies have examined attitudes and motivations of producers when it comes to 

mastitis treatment. In a study of treatment of clinical mastitis in 51 large dairy herds in 

Wisconsin (Oliveira and Ruegg, 2014), 71% of mastitis cases were treated with IM 

antimicrobials while another 43% received both IM and systemic antimicrobials. In the United 

States, no antimicrobials are currently labeled for systemic treatment of mastitis, thus in the 

Wisconsin study, all systemic treatments were considered as “extra-label.” Other studies have 

also noted a high proportion of extra-label drug use for the treatment of mastitis. Sawant et al. 

(2005) surveyed 113 herds from 13 countries and found 18% of herds used ceftiofur in an extra-

label manner to treat mastitis. Swinkels et. al (2015) surveyed 38 dairy herds in Germany and the 

Netherlands and found that 37 of 38 surveyed herds reported to “occasionally” or “frequently 

extend mastitis treatment.” They hypothesized that giving the best possible treatment gave the 

producer a feeling of being a “good farmer.” However, none of the farmers expressed concern 

 

5 

 

about the increased treatment costs, wasted milk, prolonged use of antibiotics or the potential of 

selecting for antibiotic resistant pathogens. Every tanker of milk is tested for beta-lactam 

residues, which per the USDA, has resulted in a decline over the past decade in positive violating 

residues (0.012% of tankers positive in 2015). While more than 90% of operations administer 

drugs that require a milk or meat withdrawal time, only 70% of operations do on-farm testing of 

milk for residues (USDA, 2014). 

Most cases of clinical mastitis on a dairy farm are considered mild or moderate and 

therapeutic protocols often rely on administration of IM antimicrobial drugs. However, the use of 

case definitions and criteria to decide if a case should be treated are important parts of treatment 

protocols. Basic considerations for treatment of clinical mastitis include 1) the severity of the 

case,  2) if the case is new or recurring 2) the number of quarters affected 3) the existence of 

concurrent health problems 4) stage of lactation, and 5) causative pathogen (Wagner and 

Erskine, 2009). In a study by Vaarst et al. (2002), 16 Danish dairy farmers were surveyed on the  

factors they considered to  treat a cow with mastitis. These included 1) symptoms (the severity of 

the case), 2) cow level factors (e.g., to what extent has a cow fulfilled goals of the producer and 

herd) 3) herd level factors (the current situation of the herd, including availability of replacement 

heifers, amount of milk production, milk quota), and 4) opportunities for alternatives (whether 

the producer regarded other practices as serious alternatives to antibiotic treatment). The authors 

concluded that symptom level was the dominant deciding factor in severe cases of clinical 

mastitis although what producers interpreted to be “severe” varied from farm to farm (Vaarst et 

al., 2002).  Despite the importance that dairy producers place on the severity of case 

presentation, the most important management tool that should be employed in therapeutic 

decisions for mastitis is the use of records of treatments and outcomes. However, Sawant et al 

 

6 

 

(2005) found that only 50% of surveyed herds kept written or computerized treatment records for 

antibiotics. Furthermore only 33 producers (29%) maintained complete records.  In a more recent 

survey, Kayitsinga et al (2017) found only 56% of surveyed herds maintained records of all 

treated cows and only 49% reviewed records before administering mastitis treatments. Past 

surveys done by Kaneene and Ahl (1987) of dairy producers in Michigan indicated that 

insufficient record keeping and poor knowledge about drug withdrawal periods among producers 

were important factors leading to drug residues in milk.  

Thus, mastitis has a variety of both short and long term impacts on a cow’s productivity 

ranging from discarded milk, decreased milk production, and extensive antimicrobial drug use. 

Case severity and individualized treatment protocols of the farm often determine the course of 

treatment. However, case definitions are highly variable between operations which also reflects a 

producer’s motivations and behaviors associated with treating mastitis.  

 

Mastitis Control Practices  

Controlling mastitis requires a multifactorial approach that includes milking protocols, 

equipment, cow immune status, and cow environment, such as housing, bedding, and ventilation. 

Historically, subclinical infections from contagious pathogens such as Streptococcus agalactiae 

and Staphylococcus aureus were found to be controlled by, post-milking teat dipping (PMTD), 

and blanket dry cow therapy (BDCT) with concurrent reductions in BTSCC (Erskine et al., 1987; 

Goodger et al., 1993; Hogan and Smith, 2012; Schewe et al., 2015).  This approach to mastitis 

control was summarized by the National Mastitis Council as a five point program that includes 

1) treat and record clinical cases, 2) post milking teat disinfection, 3) dry cow therapy, 4) cull 

 

7 

 

chronic cases, and 5) milk machine maintenance (NMC, 2013).  As the prevalence of contagious 

mastitis declined, mastitis control evolved to include management practices such as bedding and 

housing that reduce exposure to environmental pathogens, e.g., Streptococcus uberis, Klebsiella 

sp., and Escherichia coli. Recent surveys demonstrated that dairy producers have widely adopted 

standard practices associated with mastitis control; post-milking teat dipping, use of dry cow 

therapy, not using water during udder preparation before milking, and use of free stalls with 

inorganic bedding are practiced in about 83%, 86%, 78% and 88% of herds, respectively 

(Jayarao et al., 2004; Dufour et al., 2011; Schewe et al., 2015).   

As accepted practices for milk quality have become more widely adopted, and as our 

understanding of the disease advance, more recent studies have attempted to further discriminate 

the factors that are the most critical for mastitis control. However, results are variable due in part 

to the differences in the adoption of control practices, herd demographics, and facilities. Herds 

that house dry cows in free stalls or loose housing had lower BTSCC than those farms that used 

tie stalls, stanchions or outside lots. Mattresses and sand bedding for lactating cows and heifers 

was also associated with lower BTSCC (Wenz et al., 2007; Dufour et al., 2011), as was  

maintaining dry cow cleanliness and teat disinfection (Chassagne et al., 2005).  Barnouin et al. 

(2004; 2005) did a prospective study of 534 French herds to determine what management 

practices influenced bulk tank somatic cell counts (BTSCC). The probability of a herd belonging 

to a low somatic cell score group (top 5th percentile) was associated with: (1) regular use of teat 

dipping after mammary infusion at dry off; (2) culling cows with recurring clinical mastitis; (3) 

having experienced people provide herd management practices and (4) feeding a balanced ration. 

Additionally, use of internal teat sealants have been shown to be beneficial in lowering mastitis. 

Godden et al. (2003) found that the use of internal teat sealants decreased the risk of developing 

 

8 

 

mastitis over the dry period by 30%. Schewe et. al (2015) also found similar results, with lower 

BTSCC associated with the use of internal teat sealants and BDCT. However, as in the French 

study that linked lower BTSCC to the experience of farm personnel, the Schewe et al study 

found lower BTSCC were associated with the producer’s perspective that employees regularly 

complied with milking protocols.  

 

Milk Quality and Changes in the Dairy Demographic Landscape 

Due to extensive research, outreach education and the willingness of dairy producers to 

adopt proven mastitis control practices, herd average SCC have been decreasing over the past 

quarter century. (USDA, 2013; Norman, 2015). However, the dairy producer’s willingness to 

consistently practice mastitis control protocols varies and may impact milk quality.  A Dutch 

study found that the management approach and attitudes among herds with low BTSCC (< 

200,000 cells/mL) was  described as “clean and accurate”, as compared to herds with BTSCC 

between 250,000 and 400,000 cells/mL whose approach was described as “quick and dirty” 

(Barkema et al., 1999).  The low BTSCC herd group were found to have better overall farm 

hygiene and record keeping than the high BTSCC herd group.  In a study done by Khaitsa 

(2000), producers who perceived that BTSCC had not been a problem, or only a slight problem, 

had significantly lower BTSCC than farms where BTSCC was perceived to be a moderate or 

major problem. Thus, the approach and attitudes of the producer, and potentially of all farm 

personnel, to comply with the farm protocols, contributes to the status of milk quality on a dairy 

farm.  

 

9 

 

However, despite the importance of a producer’s approach to mastitis control, two major 

barriers hinder progress for milk quality. The first is the need to further explain dairy producer 

attitudes and motivations towards mastitis control practices; particularly social variables, that 

include cultural, economic and environmental variables. The second, and perhaps more 

important barrier, is the increasing reliance on non-family labor for most dairy herds, resulting 

from a dramatic increase in herd size and the proportion of milk produced by larger herds.  

 Over the past two decades, there has been a marked shift in herd size among dairy farms 

in the U.S. Farms with fewer than 100 cows accounted for 49% of the country’s milk cows in 

1992, but just 17% of milk cows in 2012. In contrast, farms with at least 1,000 cows accounted for 

49%  of  all  milk  cows  in  2012,  up  from  10%  in  1992  (MacDonald  and  Newton,  2014).  As 

variability in herd size increases, dairy farms are also becoming increasingly varied in terms of 

employment  practices  and  organization  (Jackson-Smith,  2000).  Additionally,  farms  with  more 

than 500 milking cows now account for 63% of the milk supply in the U. S.  (Bewley et al., 2001; 

Cross, 2006; von Keyserlingk et al., 2013). Thus, an increase in herd size has created a greater 

need for non-family labor (Bewley et al., 2001), with estimates of one employee for every 80 to 

100 cows (Douphrate et al., 2013). 

According to census 2000 data, the number of Mexican immigrants in the U.S. labor force 

nearly doubled from 1990-2000 with estimates ranging from anywhere from 2.6 million to 4.9 

million (Jenkins et al., 2009). The actual number of immigrant workers is hard to document for 

many reasons including: 1)  many temporary workers may not have been working at the time of 

the  survey  2)  some  employers  may  not  have  reported  “under  the  table”  employees,  and  3) 

employers don’t include part time workers (Harrison, 2009b). Regarding the dairy industry, 41 to 

50% of dairy farms in the U.S. rely on foreign labor (Baker and Chappelle, 2012; von Keyserlingk 

 

10 

 

et al., 2013) and 88.5% of surveyed immigrants were from Mexico, the remainder were comprised 

from Central and South America (Harrison et al., 2009a).  

There are many obstacles that immigrant employees must overcome. One of the biggest 

challenges includes language barriers. Dustmann et al. (2003) found that as English fluency 

increases, employment probabilities increase by about 22%.  However, there is a lack of 

published information regarding the importance of this problem with respect to dairy farms. In 

studies of farm employees done by Baker and Chappelle and the U.S. Department of Labor, only 

26% of surveyed employees were able to speak “some English”, 4% said they spoke English 

well, and 64% said they had little to no English language ability. Conversely, 68% stated that 

nobody in a management position spoke Spanish. Illiteracy may further complicate training of 

farm employees, especially with Spanish speaking employees. There is a need for programs 

pertaining to language training, cultural understanding, farm management skills and small 

business development assistance for minority farmers and farm workers (Harrison, 2009b; Baker 

and Chappelle, 2012).   

Even when language is not a barrier, past research has found that miscommunication still 

occurs between employees and management (Hadley et al., 2002; Román-Muñiz, 2007; Baker 

and Chappelle, 2012). Román-Muñiz et. al (2007) commented that a communication gap is likely 

to arise and the problem is further exacerbated where owners do not directly supervise daily 

tasks. Erskine et al. (2015) documented a pilot study where dairy herd managers and employees 

were interviewed with a similar set of human resource questions. Managers filled out a paper 

survey, employees were surveyed with the use of a hand-held clickers and computer acquisition 

of responses for anonymity.  Surveys were also translated into English or Spanish depending on 

the needs of the employees. While 11 out of 12 management teams believed that they provided 

 

11 

 

primary training for milking, 42% of English speaking and only 14% of Latino workers said that 

they were trained primarily by managers. From the employee’s perspective, most training was 

provided by other employees or was self-taught. 

Immigrants also experience difficulties with community, not only with language but also 

discrimination and cultural differences (Harrison, 2009a; Hagevoort et al., 2013). In a survey of 

267 workers in Wisconsin, 15% said they “didn’t go out”, or “did nothing” when asked about 

outside activities (Harrison, 2009a). Safety is also an issue with increased risk of injury, illness, 

or death occurring among workers with lower education levels, illiteracy and limited proficiency 

in English (Loh and Richardson, 2004; Smith et al., 2006; DaVila et al., 2011).  

When first hired, many employees on dairy farms are inexperienced with dairy work, 

including tasks such as milking and handling cows. This inexperience, coupled with many dairy 

producers feeling unqualified to manage and train employees, has created training and 

knowledge gaps for employees on many dairy farms. Many dairy producers realize that human 

resource management is a weakness in their operation (Stup et al., 2006) and dairy producers 

who expand their operations experience more difficulty and less satisfaction with human 

resources than any other facet of management (Bewley et al., 2001). Bewley et al. also noted that 

producers who built new facilities spent less time on farm work and more time on managing 

employees. However, producers who expanded into new facilities reported less difficulty in 

finding and retaining good employees compared to those producers who only modified their 

facilities. As the dairy industry is evolving, producers with larger herds spend more time hiring, 

training and managing employees and rely more on non-family labor. While satisfaction with 

employee morale and attitude, labor efficiency and ability to get farm work done also seems to 

increase as herd size increases, non-family labor is generally more expensive, less experienced 

 

12 

 

and less flexible than family labor (Bewley et al., 2001). Producers also recognized that there 

was a need to change how information was given to employees (Jansen and Lam, 2012).  

There has been little published regarding the motivations and attitudes of dairy farm 

workers, although some studies investigated these social variables in other agricultural 

enterprises. In 2011, Holmes participated in an observational study on a berry farm in 

Washington and documented the hierarchy among employees. He found that there is an inverse 

relationship between job responsibilities and stressors. As control over one’s job decreases, 

anxiety accumulates within positions of lower supervision and responsibility.  Kolstrup (2012) 

surveyed a combined 550 agricultural students, dairy farm workers and employers. Employees 

ranked the highest satisfaction scores to factors that included: 1) independence at work; 2) 

provision of free working clothes and 3) working with animals. Priorities for employees that 

chose to work and remain in agricultural field were 1) fun at work; 2) good leadership; 3) 

working with animals; 4) pride in the profession, and 5) job security.  

There are both external and internal influences on employee job satisfaction in the dairy 

industry. Larger herds often offer higher cash wages and total compensation, health insurance 

and housing compared to smaller herds. Qualitative observations show employees also enjoy 

non-cash benefits but underestimate their value (Fogleman, 1999). Valeeva et al. (2007) found 

that internal farm performance factors provide more motivation than external factors (including 

farm penalties or incentives) for employees. Internal non-monetary factors that related to internal 

esteem and taking pleasure in healthy animals was as equally motivating as monetary factors.  

 

 

13 

 

Shift fatigue has been documented in industries including the automotive industry and police 

work. When comparing an eight vs. twelve hour shift in the automotive industry, it was found 

that even though reaction time was faster for employees at the end of a twelve hour shift,  more 

errors were documented when compared to the end of the eight hour shift (Mitchell, 2000). Vila 

et al. (2002) documented how police officers often work twelve  hour shifts, noting there are 

both long term and short term consequences of fatigue. Short term consequences include 

sleepiness, diminished appetite, dulled recollection and slower reaction times. Long term 

consequences include increased stress levels and decreased abilities to cope with complex and 

threatening situations, interference of familial and social relationships that are connected to the 

community, as well as the perspective needed for sound decision making. However, little is 

known of the impact that shift fatigue and ergonomics have on the performance of employees on 

dairy farms.  

The demographics of dairy farming in the United States are changing and thus, the needs of 

the industry are also evolving. Many of the employees that are working on dairy farms now are 

from other countries and when working in the United States face illiteracy and language barriers. 

These are barriers that can be overcome with proper training and educational opportunities to 

encourage employee engagement. However, there is an overall lack of capacity within the dairy 

industry to provide adequate employee training and education. Understanding employee 

motivations and their needs will help contribute to employee retention, improve farm 

productivity, and reduce the cost of employee turnover.  

 

 

 

14 

 

Employee Behaviors and Milk Quality   

The application of proven milking protocols is a crucial part of producing high quality milk 

on a dairy farm by 1) maintaining sanitation, 2) reducing mastitis, 3) harvesting milk effectively 

and 4) ensuring that abnormal milk, including that from cows treated with antibiotics, does not 

end up in the food supply. Effective milking is based upon the triad of interactions between the 

milking system, the cow, and personnel who milk the cows. Employees can be part of a positive 

milking experience for the cows by practicing proper milking procedures, and moving cows to 

and from the parlor in a non-stressful manner. Current recommendations for milking procedures 

from the National Mastitis Council include the following; that a low stress environment needs to 

be provided for cows, the foremilk and udder should be checked for mastitis, teats should be 

disinfected  with an udder wash solution or pre-dip and then be dried completely with an 

individual towel, milking units should be attached within 120 seconds after initiation of 

stimulation and adjusted for proper alignment, vacuum should be shut off before removing units, 

and teats should be dipped immediately after unit removal with an effective disinfectant (NMC, 

2013). Perhaps the three most critical milking behaviors for effective milking are 1) ensuring 

proper pre-stimulation of teats and time period between stimulation and unit attachment to the 

udder (lag time), 2) ensuring that units are aligned on the udder properly during milking and that 

air vents that allow proper milk flow away from the cluster are patent, and 3) not allowing the 

cows to be overmilked.    

The anatomy of the bovine udder divides milk storage; 20% of the milk is held in the milk 

cistern at the bottom of the udder, known as “cisternal milk”, and 80% is in the mammary gland, 

known as “glandular milk”(Bruckmaier, 2005). While cisternal milk can be readily released, 

glandular milk which is held in the mammary gland alveoli requires the contraction of 

 

15 

 

myoepithelial cells to move milk through the ducts and into the cistern. Pre-milking stimulation 

of the teats is necessary to induce a neural reflex arc that causes oxytocin release from the 

pituitary gland into the circulation, which in turn activates the myoepithelial cells to force 

ejection of milk from the alveolar space (Bruckmaier, 2005). Weiss and Bruckmaier et al. (2005, 

2013) recommend that the duration of pre-stimulation before milking should be for 30-60 

seconds, however with low udder filling, longer pre-stimulation of 90 seconds can be helpful, 

which may not always practical in the parlor environment. They also found that the intensity of 

stimulation is less important than the duration. However, stimulation does not need to occur all at 

once and can be split into shorter tactile sessions, minimum of 15 seconds, followed by a latency 

of 45 seconds. Because of the short plasma half-life of oxytocin, no more than 2 minutes should 

pass between tactile stimulation and milking unit attachment (Bruckmaier et al., 2013). After 

initial stimulation, continuously elevated levels of oxytocin are needed through the entire 

milking process which is typically provided by liner (inflation) stimulation during pulsation. 

Even greater myoepithelial contraction is needed to eject milk out of an incompletely filled, as 

compared to, filled alveolus (Bruckmaier, 2007). Therefore, good udder stimulation is even more 

important for a late lactation cow compared to a cow that has recently calved. Inhibition of milk 

ejection due to a lack of oxytocin may also be caused by a variety of factors, including 

unfamiliar surroundings,  cortisol related to rough cow handling, or use of exogenous oxytocin 

injections (Bruckmaier, 2013), .  

Unfortunately, as labor costs increase and farm size increases, less time is spent on proper 

teat stimulation, therefore reducing time for udder preparation (Sandrucci et al., 2007). Proper 

udder stimulation and lag times relative to milk ejection can be measured by several indicators, 

however measures of peak milk flow are most commonly used. Bimodal milking, also known as 

 

16 

 

“biphasic milking,” is caused by the disturbance of oxytocin availability to myoepithelial cells 

which causes a time period between release of cisternal and glandular milk and thus a period 

when no milk is flowing. This delayed milk ejection has many negative health effects to the cow 

including penetration of the milking vacuum into the cistern which subsequently collapses the 

teat cavities, interrupts blood flow, allows the milking clusters to “climb” up the teats, causes air 

admission and increases bacterial exposure to the teat ends. This disrupts effective milk ejection 

during the remainder of milking even after delayed milk “let down” has been resolved 

(Bruckmaier, 2007).   

Several factors can cause delayed milk ejection. The most common is inadequate stimulation 

causing a lack of oxytocin to the myoepithelial cells to induce good milk let down. However, 

certain cow factors may also contribute to this problem, including longer days in milk (DIM). 

According to Sandrucci et. al (2007) the percentage of bimodal curves increased throughout 

lactation; 27% of cows < 150 days in milk (DIM) had bimodal curves and 40% of cows > 150 

DIM had bimodal curves. Tancin et. al (2007) also found an association between bimodal milk 

let down and longer overmilking phases and lower milk yield. This may have long term effects 

as investigators also found that a longer duration of the decline phase of the milk flow curve 

seems to have positive correlation with SCC, either for the single quarter or for the whole udder.  

The second important practice to ensure effective milking is to alleviate and reduce 

overmilking.  Overmilking is defined as when the milk flow to the teat cistern is less than the 

flow out of the teat canal (Rasmussen, 2004). Overmilking can be caused by improper automated 

take off settings on milking units (milking the cows dry) or the use of manual milking (when 

units are set to continuously milk until they are taken off by milking personnel). Overmilking 

results in longer unit on times (when units are attached to the udder), and thus longer exposure of 

 

17 

 

teat ends to vacuum.  In the short term, this can result in color changes to the teat wall which 

indicates vascular damage and swelling of the teat canal (Mein, 2001). Vascular changes and 

swelling causes the flow of milk to be impeded from being expelled efficiently. Over the long 

term, teat ends become roughened which can lead to increasing rates of clinical mastitis (NMC, 

1999; Neijenhuis et al., 2000; Paduch, 2012). Teat end changes can occur with chronic exposure 

(2 to 8 weeks) from overmilking that can result in teat end roughness and callosity 

(hyperkeratosis). Other cow variables that can affect teat end hyperkeratosis include teat end 

shape, cow production level and stage of lactation, as well as machine variables such as slow 

milking and over milking, take off settings for automatic cluster take offs and pre-milking udder 

preparation (Mein, 2001). Moderate and severely hyperkeratotic teat ends have a higher risk 

(ranging from 0.3 to 2.2-fold) of a clinical mastitis event during that lactation than normal teats 

(Neijenhuis et al., 2001; Breen et al., 2009; Paduch et al., 2012).  

Longer unit on times also decrease parlor efficiency by decreasing parlor turnover rate and 

increasing the time that cows are away from the barn. Increased time that cows wait in holding 

pens increases the time that they remain standing away from feed and stall rest. Watters et al. 

(2014) found that providing feed to cows 90 minutes post-milking decreases the risk of 

intramammary infection. DeVries et. al (2012) also found longer periods of cows standing in the 

holding pen were associated with poor udder hygiene.  

Milking personnel also maintain patent air vents and proper alignment of the milking units. 

Proper alignment of milking units is crucial for efficient and gentle milking. It also reduces the 

risk of liner slips which allow atmospheric air to enter the milking unit. This increases the risk of 

milk splashing upwards towards the open teat orifice and can introduce bacteria into the canal. 

Air vents help to maintain proper milking vacuum and move milk away from the cow along a 

 

18 

 

vacuum gradient towards the cluster. Vents also help prevent back flushing of milk towards the 

teat end in case of a slip (LeGassick, 2013).  

Along with the cow and milking machine, employees are the third piece of the triad for 

effective milking. They can be directly responsible for creating an environment for a good 

milking experience that includes all the points listed above; adequate teat stimulation that 

induces oxytocin-related milk let down, patent air vents and proper unit alignment, and 

prevention of overmilking.  

Thus, a complete evaluation of milk quality on a dairy farm should not only include the 

traditional parts of an evaluation such as the milking equipment and the cows, but should also 

provide insight into the employee’s work culture, training and education. As employees often 

handle and provide care to the animals on the farm daily, their knowledge and behaviors are 

critical for the success of the milking operation.  

 

Conclusion  

Mastitis continues to be a significant problem for the dairy industry. Proven control 

measures are available and have shown to be successful, however greater reliance on employees 

has raised new challenges for mastitis control because of a lack of capacity to train and educate 

personnel to follow accepted mastitis control protocols. This ‘cultural lag’ in the dairy industry 

has created a need to further study the labor culture on dairy farms, and find the means to assess 

employee performance and thus success of employee training. As shown with the Barkema et. al 

(1999) study, a dairy producers approach to farm management impacts milk quality. If a 

producer’s management style is “quick and dirty” their farms tend to be associated with lower 

 

19 

 

milk quality vs. those producers whose farm management style was “careful and precise”.  This 

dichotomy of producer approaches to mastitis control practices may also exist in a producer’s 

approach to employee training and management as well, which could potentially impact milk 

quality and food security. To address this crucial problem, the overall objectives of this thesis are 

to determine the impact of the labor culture, employee training, and employee behaviors on two 

critical outcomes of efficient milking, milk let down and prevention of overmilking.  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

20 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

REFERENCES 

21 

 

REFERENCES 

 

 

Ahmadzadeh, A., F. Frago, B. Shafii, J. C. Dalton, W. J. Price, and M. A. McGuire. 2009. Effect 
of clinical mastitis and other diseases on reproductive performance of Holstein cows. Anim. 
Reprod. Sci. 112:273-282. 
 
Archer, S. C., F. Mc Coy, W. Wapenaar, and M. J. Green. 2013. Association of season and herd 
size with somatic cell count for cows in Irish, English, and Welsh dairy herds. The Veterinary 
Journal 196:515-521. 
 
Baker, D. and D. Chappelle. 2012. Health Status and Needs of Latino Dairy Farmworkers in  
Vermont. Journal of Agromedicine 17:277-287. 
 
Barkema, H. W., J. D. Van der Ploeg, Y. H. Schukken, T. J. G. M. Lam, G. Benedictus, and A. 
Brand. 1999. Management Style and Its Association with Bulk Milk Somatic Cell Count and 
Incidence Rate of Clinical Mastitis. J. Dairy Sci. 82:1655-1663. 
 
Barlow, J. W., L. J. White, R. N. Zadoks, and Y. H. Schukken. 2009. A mathematical model 
demonstrating indirect and overall effects of lactation therapy targeting subclinical mastitis in 
dairy herds. Prev. Vet. Med. 90:31-42. 
 
Barnouin, J., S. Bord, S. Bazin, and M. Chassagne. 2005. Dairy Management Practices 
Associated with Incidence Rate of Clinical Mastitis in Low Somatic Cell Score Herds in France. 
J. Dairy Sci. 88:3700-3709. 
 
Barnouin, J., M. Chassagne, S. Bazin, and D. Boichard. 2004. Management Practices from 
Questionnaire Surveys in Herds with Very Low Somatic Cell Score Through a National Mastitis 
Program in France. J. Dairy Sci. 87:3989-3999. 
 
Bartlett, P. C., G. Y. Miller, C. R. Anderson, and J. H. Kirk. 1990. Milk Production and Somatic 
Cell Count in Michigan Dairy Herds. J. Dairy Sci. 73:2794-2800. 
 
Bewley, J., R. W. Palmer, and D. B. Jackson-Smith. 2001. An Overview of Experiences of 
Wisconsin Dairy Farmers who Modernized Their Operations. J. Dairy Sci. 84:717-729. 
 
Breen, J. E., M. J. Green, and A. J. Bradley. 2009. Quarter and cow risk factors associated with 
the occurrence of clinical mastitis in dairy cows in the United Kingdom. J. Dairy Sci. 92:2551-
2561. 
 
Bruckmaier, R. M. 2005. Normal and disturbed milk ejection in dairy cows. Domest. Anim. 
Endocrinol. 29:268-273. 
 

 

22 

 

Bruckmaier, R. M. 2013. Oxytocin from the pituitary or from the syringe: Importance and 
Consequences for Milking Machine in Dairy Cows. Pages 4-11 in Proc. NMC Annual Meeting 
2013. 
Bruckmaier, R. M., Wellnitz, O. 2007. Induction of milk ejection and milk removal in different 
production systems. J. Anim. Sci. 86:15-20. 
 
Cha, E., D. Bar, J. A. Hertl, L. W. Tauer, G. Bennett, R. N. González, Y. H. Schukken, F. L. 
Welcome, and Y. T. Gröhn. 2011. The cost and management of different types of clinical 
mastitis in dairy cows estimated by dynamic programming. J. Dairy Sci. 94:4476-4487. 
 
Chassagne, M., J. Barnouin, and M. Le Guenic. 2005. Expert Assessment Study of Milking and 
Hygiene Practices Characterizing Very Low Somatic Cell Score Herds in France. J. Dairy Sci. 
88:1909-1916. 
 
Cross, J. A. 2006. RESTRUCTURING AMERICA'S DAIRY FARMS*. Geographical Review 
96:1-23. 
 
DaVila, A., M. T. Mora, and R. GonzÁLez. 2011. English-Language Proficiency and 
Occupational Risk Among Hispanic Immigrant Men in the United States. Industrial Relations: A 
Journal of Economy and Society 50:263-296. 
 
Dohoo, I. R. and A. H. Meek. 1982. Somatic Cell Counts in Bovine Milk. The Canadian 
Veterinary Journal 23:119-125. 
 
Douphrate, D. I., G. R. Hagevoort, M. W. Nonnenmann, C. L. Kolstrup, S. J. Reynolds, M. 
Jakob, and M. Kinsel. 2013. The Dairy Industry: A Brief Description of Production Practices, 
Trends, and Farm Characteristics Around the World. Journal of Agromedicine 18:187-197. 
 
Dufour, S., A. Frechette, H. W. Barkema, A. Mussell, and D. T. Scholl. 2011. Effect of udder 
health management practices on herd somatic cell count. J. Dairy Sci. 94:563-579. 
 
Dustmann, C. and F. Fabbri. 2003. Language proficiency and labour market performance of 
immigrants in the UK*. The Economic Journal 113:695-717. 
 
Erskine, R. J., R. J. Eberhart, L. J. Hutchinson, and S. B. Spencer. 1987. Herd management and 
prevalence of mastitis in dairy herds with high and low somatic cell counts. J. Am. Vet. Med. 
Assoc. 190:1411-1416. 
 
Erskine, R. J., R. O. Martinez, and G. A. Contreras. 2015. Cultural lag: A new challenge for 
mastitis control on dairy farms in the United States. J. Dairy Sci 98:8240-8244. 
 
Erskine, R. J., S. Wagner, and F. J. DeGraves. 2003. Mastitis therapy and pharmacology. Vet. 
Clin. North Am. Food Anim. Pract. 19:109-138. 
 

 

23 

 

Fogleman, S., Milligan, R., Maloney, T., Knoblauch, W. 1999. Employee Compensation and Job 
Satisfaction on Dairy Farms in the Northeast. 1999 Annual Meeting, Aug 8-11, Nashville, TN 
from American Agricultural Economics Association. 
 
Fox, L. 2013. Can Milk Somatic Cells Get too Low? A Question to be Revisited. Pages 56-63 in 
Proc. Annual National Mastitis Conference. National Mastitis Council. 
 
Fuenzalida, M. J., P. M. Fricke, and P. L. Ruegg. 2015. The association between occurrence and 
severity of subclinical and clinical mastitis on pregnancies per artificial insemination at first 
service of Holstein cows. J. Dairy Sci 98:3791-3805. 
 
Goodger, W. J., T. Farver, J. Pelletier, P. Johnson, G. DeSnayer, and J. Galland. 1993. The 
association of milking management practices with bulk tank somatic cell counts. Prev. Vet. Med. 
15:235-251. 
 
Gruet, P., Maincent, P. , Berthelot, X. , Kaltsatos, V. 2001. Bovine Mastitis and IM drug delivery 
review and perspectives. Advanced Drug Delivery Reviews:245-259. 
 
Hadley, G. L., S. B. Harsh, and C. A. Wolf. 2002. Managerial and Financial Implications of 
Major Dairy Farm Expansions in Michigan and Wisconsin. J. Dairy Sci. 85:2053-2064. 
Hagevoort, G. R., D. I. Douphrate, and S. J. Reynolds. 2013. A review of health and safety 
leadership and managerial practices on modern dairy farms. Journal of agromedicine 18:265-
273. 
 
Hagnestam-Nielsen, C., U. Emanuelson, B. Berglund, and E. Strandberg. 2009. Relationship 
between somatic cell count and milk yield in different stages of lactation. J. Dairy Sci. 92:3124-
3133. 
 
Halasa, T., M. Nielen, A. P. W. De Roos, R. Van Hoorne, G. de Jong, T. J. G. M. Lam, T. van 
Werven, and H. Hogeveen. 2009. Production loss due to new subclinical mastitis in Dutch dairy 
cows estimated with a test-day model. J. Dairy Sci. 92:599-606. 
 
Hand, K. J., A. Godkin, and D. F. Kelton. 2012. Milk production and somatic cell counts: A 
cow-level analysis. J. Dairy Sci. 95:1358-1362. 
 
Harrison, J., Lloyd, Sarah and O'Kane, Trish. 2009a. Immigrant Dairy Workers in Rural 
Wisconsin - Briefing 4. in Program on Agricultural Technology Studies. M. Univeristy of 
Wisconsin, ed, University of Wisconsin, Madison. 
 
Harrison, J., Lloyd, Sarah and O'Kane, Trish. 2009b. Overview of Immigrant Workers on 
Wisconsin Dairy Farms - Briefing 1. in Program on Agricultural Technology Studies. 
 
Heikkilä, A. M., J. I. Nousiainen, and S. Pyörälä. 2012. Costs of clinical mastitis with special 
reference to premature culling. J. Dairy Sci. 95:139-150. 
 

 

24 

 

Hogan, J. and K. L. Smith. 2012. Managing Environmental Mastitis. Vet. Clin. North Am. Food 
Anim. Pract. 28:217-224. 
 
Hogeveen, H., K. Huijps, and T. Lam. 2011. Economic aspects of mastitis: New developments. 
N. Z. Vet. J. 59:16-23. 
 
Holmes, S. M. 2011. Structural vulnerability and hierarchies of ethnicity and citizenship on the 
farm. Med. Anthropol. 30:425-449. 
 
Jackson-Smith, D., Barham, B. . 2000. Dynamics of Dairy Industry Restructuring in Wisconsin. 
Research in Rural Sociology and Development 8:115-139. 
 
Jansen, J. and T. J. Lam. 2012. The role of communication in improving udder health. Vet. Clin. 
North Am. Food Anim. Pract. 28:363-379. 
 
Jayarao, B. M., S. R. Pillai, A. A. Sawant, D. R. Wolfgang, and N. V. Hegde. 2004. Guidelines 
for monitoring bulk tank milk somatic cell and bacterial counts. J. Dairy Sci. 87:3561-3573. 
 
Jenkins, P. L., S. G. Stack, J. J. May, and G. Earle-Richardson. 2009. Growth of the Spanish-
Speaking Workforce in the Northeast Dairy Industry. Journal of Agromedicine 14:58-65. 
 
Kaneene, J. B. and A. S. Ahl. 1987. Drug Residues in Dairy Cattle Industry: Epidemiological 
Evaluation of Factors Influencing Their Occurrence. J. Dairy Sci. 70:2176-2180. 
 
Kayitsinga, J., R. L. Schewe, G. A. Contreras, and R. J. Erskine. 2017. Antimicrobial treatment 
of clinical mastitis in the eastern United States: The influence of dairy farmers' mastitis 
management and treatment behavior and attitudes. J. Dairy Sci. 
 
Kehrli, M. E., Jr. and D. E. Shuster. 1993. Factors Affecting Milk Somatic Cells and Their Role 
in Health of the Bovine Mammary Gland. J. Dairy Sci. 77:619-627. 
 
Khaitsa, M. L., T. E. Wittum, K. L. Smith, J. L. Henderson, and K. H. Hoblet. 2000. Herd 
characteristics and management practices associated with bulk-tank somatic cell counts in herds 
in official Dairy Herd Improvement Association programs in Ohio. Am. J. Vet. Res. 61:1092-
1098. 
 
Kirkpatrick, M. A., Olson, Jerry D. 2015. Somatic Cell Counts at First Test: More Than a 
Number. Pages 53-56 in National Mastitis Council Annual Meeting 2015. NMC, Memphis, TN. 
 
Klei, L., J. Yun, A. Sapru, J. Lynch, D. Barbano, P. Sears, and D. Galton. 1998. Effects of Milk 
Somatic Cell Count on Cottage Cheese Yield and Quality. J. Dairy Sci. 81:1205-1213. 
 
Kolstrup, C. L. 2012. What factors attract and motivate dairy farm employees in their daily 
work? Work (Reading, Mass.) 41:5311-5316. 
 

 

25 

 

Labor, U. S. D. o. 2005. Findings from the National Agricultural Workers Survey (NAWS) 
2001-2002. Research Report No. 9. 
 
LeGassick, J. 2013. Liners should not be the first to blame. Pages 72-73 in Progressive 
Dairyman. Progressive Dairyman. 
 
Loeffler, S. H., M. J. de Vries, and Y. H. Schukken. 1999. The Effects of Time of Disease 
Occurrence, Milk Yield, and Body Condition on Fertility of Dairy Cows. J. Dairy Sci. 82:2589-
2604. 
 
Loh, K. and S. Richardson. 2004. Foreign-born workers: trends in fatal occupational injuries, 
1996-2001. Mon. Labor Rev. 127:42-53. 
 
Losinger, W. C. 2005. Economic impacts of reduced milk production associated with an increase 
in bulk-tank somatic cell count on US dairies. J. Am. Vet. Med. Assoc. 226:1652-1658. 
 
Ma, Y., C. Ryan, D. M. Barbano, D. M. Galton, M. A. Rudan, and K. J. Boor. 2000. Effects of 
Somatic Cell Count on Quality and Shelf-Life of Pasteurized Fluid Milk1. J. Dairy Sci. 83:264-
274. 
 
MacDonald, J. and D. Newton. 2014. Milk Production Continues Shifting to Large-Scale Farms. 
Amber Waves:7-1E,2E,3E,4E,5E,6E,7E. 
 
Mein, G. A., Neijenhuis, F., Morgan, W.F., Reinemann, D.J., Hillerton, J.E, Baines, J.R., 
Ohnstad, I., Timms, L., Britt, J.S., Farnsworth, R., Cook, N., Hemling, T. 2001. Evaluation of 
Bovine Teat Condition in Commercial Dairy Herds: 1. Non-Infectious Factors. in Proc. 2nd 
International Symposium on Mastitis and Milk Quality. 
 
Mitchell, R. J., Williamson, A.M. 2000. Evaluation of an 8  hour versus a 12 hour shift roster on 
employees at a power station. Appl. Ergon. 31:83-93. 
 
Neijenhuis, F., H. W. Barkema, H. Hogeveen, and J. P. T. M. Noordhuizen. 2000. Classification 
and longitudinal examination of callused teat ends in dairy cows. J. Dairy Sci. 83:2795-2804. 
 
Neijenhuis, F., H. W. Barkema, H. Hogeveen, and J. P. T. M. Noordhuizen. 2001. Relationship 
Between Teat-End Callosity and Occurrence of Clinical Mastitis. J. Dairy Sci. 84:2664-2672. 
 
NMC. 1999. Teat Lesions Can Lead to Milking Problems, Mastitis. NMC. 
 
NMC. 2013. Reccomended MIlking Protocols. in www.nmconline.org. N. M. Council, ed. 
National Mastitis Council, Verona, WI. 
 
Norman, H. D., Walton, L.M., Durr, J. 2015. Somatic cell counts of milk from Dairy Herd 
Improvement herds during 2015. Council on Dairy Cattle Breeding. 
 

 

26 

 

Paduch, J.-H., E. Mohr, and V. Krömker. 2012. The association between teat end hyperkeratosis 
and teat canal microbial load in lactating dairy cattle. Vet. Microbiol. 158:353-359. 
 
Paduch, J. H., Mohr, E., Kromker, V. 2012. The association between teat end hyperkeratosis and 
teat canal microbial load in lactating dairy cattle. Vet. Microbiol. 158:353-359. 
 
Rasmussen, M. D. 2004. Overmilking and Teat Condition. Pages 169-175 in National Mastitis 
Council Annual Meeting. NMC. 
 
Raubertas, R. F. and G. E. Shook. 1980. Relationship Between Lactation Measures of Somatic 
Cell Concentration and Milk Yield<sup>1</sup>. J. Dairy Sci. 65:419-425. 
 
Reccomended Milking Procedures. in NMC Fact Sheet. N. M. Council, ed. National Mastitis 
Council, Verona, WI. 
 
Román-Muñiz, I. N., Van Metre, D. C., Garry, F. B., & Smith, R. A. 2007. Dairy Worker 
Training Experiences. Pages 20-22 in Proc. Fortieth Annual Conference, American Association 
of Bovine Practitioners, Vancouver, British Columbia, Canada. 
 
Royster, E. and S. Wagner. 2015. Treatment of Mastitis in Cattle. Vet. Clin. North Am. Food 
Anim. Pract. 31:17-46. 
 
Sandrucci, A., A. Tamburini, L. Bava, and M. Zucali. 2007. Factors Affecting Milk Flow Traits 
in Dairy Cows: Results of a Field Study. J. Dairy Sci. 90:1159-1167. 
 
Schepers, A. J., T. J. G. M. Lam, Y. H. Schukken, J. B. M. Wilmink, and W. J. A. Hanekamp. 
1997. Estimation of Variance Components for Somatic Cell Counts to Determine Thresholds for 
Uninfected Quarters. J. Dairy Sci. 80:1833-1840. 
 
Schewe, R. L., J. Kayitsinga, G. A. Contreras, C. Odom, W. A. Coats, P. Durst, E. P. Hovingh, 
R. O. Martinez, R. Mobley, S. Moore, and R. J. Erskine. 2015. Herd management and social 
variables associated with bulk tank somatic cell count in dairy herds in the eastern United States.  
J. Dairy Sci 98:7650-7665. 
 
Seegers, H., C. Fourichon, and F. Beaudeau. 2003. Production effects related to mastitis and 
mastitis economics in dairy cattle herds. Vet. Res. 34:475-491. 
 
Smith, S. M., T. Perry, and D. Moyer. 2006. Creating a Safer Workforce. Prof. Saf. 51:20-25. 
Stup, R. E., J. Hyde, and L. A. Holden. 2006. Relationships between selected human resource 
management practices and dairy farm performance. J. Dairy Sci. 89:1116-1120. 
 
Swinkels, J. M., A. Hilkens, V. Zoche-Golob, V. Krömker, M. Buddiger, J. Jansen, and T. J. G. 
M. Lam. 2015. Social influences on the duration of antibiotic treatment of clinical mastitis in 
dairy cows. J. Dairy Sci 98:2369-2380. 
 

 

27 

 

Tančin, V., A. H. Ipema, and P. Hogewerf. 2007. Interaction of Somatic Cell Count and Quarter 
Milk Flow Patterns. J. Dairy Sci 90:2223-2228. 
 
USDA, A., VS. 2013. Determining US MIlk Quality Using Bulk-tank Somatic Cell Counts. A. 
United States Department of Agriculture, VS, , ed. 
 
USDA, A., VS, CEAH. 2008. Antibiotic Use on U.S. Dairy Operations, 2002 and 2007. Page 5. 
A. USDA, VS, CEAH, ed. USDA, Fort Collins, CO. 
 
USDA, A., VS, NAHMS. 2016. Dairy 2014 Milk Quality, MIlking Procedures and Mastitis on 
U.S. Dairies, 2014. Fort Collins, CO. 
 
Vaarst, M., B. Paarup-Laursen, H. Houe, C. Fossing, H.J. Andersen. 2002. Farmers' Choice of 
Medical Treatment of Mastitis in Danish Dairy Herds Based on Qualitative Research Interviews.  
J. Dairy Sci. 85:992-1001. 
 
Vila, B., G. B. Morrison, and D. J. Kenney. 2002. Improving Shift Schedule and Work-Hour 
Policies and Practices to Increase Police Officer Performance, Health, and Safety. Police 
Quarterly 5:4-24. 
 
von Keyserlingk, M. A., N. P. Martin, E. Kebreab, K. F. Knowlton, R. J. Grant, M. Stephenson, 
C. J. Sniffen, J. P. Harner, 3rd, A. D. Wright, and S. I. Smith. 2013. Invited review: 
Sustainability of the US dairy industry. J. Dairy Sci. 96:5405-5425. 
 
Wagner, S. A. and R. J. Erskine. 2009. CHAPTER 101 - Decision Making in Mastitis Therapy. 
Pages 502-509 in Food Animal Practice (Fifth Edition). D. E. A. M. Rings, ed. W.B. Saunders, 
Saint Louis. 
 
Weiss, D. and R. M. Bruckmaier. 2005. Optimization of Individual Prestimulation in Dairy 
Cows. J. Dairy Sci. 88:137-147. 
 

Wenz, J. R., S. M. Jensen, J. E. Lombard, B. A. Wagner, and R. P. Dinsmore. 2007. Herd 
Management Practices and Their Association with Bulk Tank Somatic Cell Count on United 
States Dairy Operations. J. Dairy Sci. 90:3652-3659. 

 

 

 

 

 

 

 

 

28 

 

 

 

CHAPTER 2 

DELAYED MILK EJECTION IN MICHIGAN DAIRY HERDS 

Herd-level variables associated with delayed milk ejection in Michigan dairy herds  

 

R.  Moore-Foster*,  B.  Norby*,  R.  L.  Schewe†,  R.  Thomson‡,  P.  C.  Bartlett*,  and  R.  J. 
Erskine*1 
  
 
 

*Department of Large Animal Clinical Sciences  
and 
‡Department of Animal Science 
Michigan State University 
East Lansing, 48824 
 
† Department of Sociology 
Syracuse University, Syracuse, NY 13244  
 
 
 
1Corresponding Author 
R. J. Erskine 
736 Wilson Rd 
East Lansing, 48824 
517-355-9593 
Fax: 517-432-1042 
erskine@msu.edu  

 

 

 

 

 

 

29 

 

Abstract 

The  objective  of  this  study  was  to  determine  which  herd-level  variables  were  associated  with 

delayed milk ejection (bimodal milk let down) in 64 Michigan dairy herds. Median herd size was 

294 cows (range 59 to 2,771 cows). For each herd, milking protocols were observed and milk flow 

dynamics were estimated by use of digital vacuum recorders. Surveys were also administered to 

the producers to measure mastitis management practices and attitudes. Milk flow dynamics were 

recorded for a total of 3,824 cow milkings, with a mean of 60 milkings per herd (range of 11 to 

154). Backwards multivariable analysis was used to determine which of the 47 herd-level milking 

and management variables were associated with delayed milk ejection (cows with milk let-down 

periods between milking cluster attachment and the incline phase of milk flow of  > 30 s). Delayed 

milk  ejection  occurred  in  an  average  of  25%  of  the  cows  in  each  herd  (range  0  to  75%).  A 

multivariable model found that the proportion of cows in a herd with delayed milk ejection was 

negatively associated with mean total time of tactile stimulation during pre-milking routines and 

positively associated with herd size. In summary, delayed milk ejection is more likely in herds 

with milking practices that may emphasize parlor efficiency over milking efficiency.  

Key words: delayed milk ejection, bimodal milking, milking behaviors 

 

 

 

 

 

30 

 

 

Introduction 

Milk  ejection  is  the  active  transport  of  alveolar  milk  through  the  milk  ducts  into  the  cisternal 

compartment  following  contraction  of  myoepithelial  cells  that  surround  the  mammary  alveoli 

(Bruckmaier  and  Wellnitz,  2007).  Delayed  milk  ejection  (DME),  most  often  exhibiting 

bimodality,  is  caused  by  the  disturbance  of  oxytocin  availability  to  myoepithelial  cells,  which 

interrupts milk flow between release of cisternal and alveolar milk (Bruckmaier and Blum, 1996; 

Bruckmaier  and  Blum,  1998).  Delayed  milk  ejection  has  negative  health  effects  for  the  udder, 

including penetration of the milking vacuum into the cistern which subsequently can collapse the 

teat cavity and interrupt blood flow (Bruckmaier and Wellnitz, 2007). This may in turn, allow the 

milking clusters to “climb” up the teats, cause air admission and increase bacterial exposure to the 

teat ends. This then disrupts effective milk ejection during the remainder of milking, even after 

DME has been resolved and is associated with decreased milk yield. (Bruckmaier, 2007; Samoré 

et al., 2011).   

 Several factors have been associated with increased risk of DME. The most common human 

behavioral  factors  are  inadequate  pre-milking  stimulation  and  improper  lag  time  (Bruckmaier, 

2007; Sandrucci et al., 2007; Watters et al., 2011). Weiss and Bruckmaier (2005) recommended 

that the duration of pre-stimulation before milking should be 30 to 60 s, however with low udder 

filling, longer pre-stimulation periods of 90 s can be helpful. This amount of stimulation may not 

be  practical  in  many  parlor  environments.  Other  researchers  have  found  that  stimulation  by 

stripping before milking did not affect milk flow or yield (Wagner and Ruegg, 2002). Because of 

the  short  plasma  half-life  of  oxytocin,  no  more  than  2  minutes  should  pass  between  tactile 

stimulation and milking unit attachment (Bruckmaier et al., 2013).  

 

31 

 

Additionally, cow factors such as increasing days in milk (DIM) may also contribute to DME. 

According to Sandrucci et al. (2007), the percentage of milk flow curves with DME increases in 

later lactation; 27% of  cows < 150 DIM and 41% of cows  > 150 DIM  had delayed milk flow 

curves. This agrees with Kaskous and Bruckmaier (2011) that found a higher proportion of DME 

in cows with less alveolar filling of milk before milking compared to cows with greater levels of 

alveolar filling; a factor typical of cows milked more frequently per day or in late lactation.  

As herd size increases in the U.S. dairy industry, farms are increasingly reliant on hired labor 

versus traditional owner-family labor (Baker and Chappelle, 2012; von Keyserlingk et al., 2013). 

This dynamic may impact milking efficiency; for example, Sandrucci et al. (2007) found that as 

farm size increases, less time is spent on proper teat stimulation, therefore reducing time for udder 

preparation.  However,  other  labor  variables  that  may  affect  milk  ejection,  or  their  relative 

importance to previously identified factors, have not been fully investigated. The purpose of this 

study was to further explore which herd-level variables are associated with DME.  

 

Materials and Methods 

Dairy Farm Selection 

This  study  was  part  of  a  larger  project  in  which  124  dairy  herds  in  Florida,  Michigan,  and 

Pennsylvania  participated  in  a  15-month  trial  to  develop  an  evaluation  to  assess  mastitis  and 

antimicrobial drug use. Sixty-four Michigan dairy herds were visited by the investigators twice 

between January, 2016 and May, 2017. Enrolled herds participated in the Dairy Herd Improvement 

(DHI) individual cow somatic cell count (SCC) option and had a herd size ≥ 70 cows. Because 

the overall objectives of the umbrella project included employee-related factors on milk quality 

 

32 

 

and antimicrobial use on dairy farms, organic dairies and herds that milked with automated milking 

systems were excluded.  All survey information was collected following approval and performed 

within the guidelines set by the Institutional Review Board of Michigan State University.  

 

Herd Profile and Management Culture 

During the initial herd visit, project investigators explained the study design and conducted a herd 

profile to record milking times and groups, type of milking facility, housing, employee structure, 

and  other  general  information.  Within  30  to  60  d,  the  investigators  returned  to  conduct  a  milk 

quality evaluation that included 1) milking behaviors and proficiency, 2) milking systems, 3) cow 

environment, 4) monitoring and therapy of infected cows, and 5) farm management culture. To 

capture information relative to the management culture, we interviewed dairy producers and/or 

managers  with  an  84  question  survey  relative  to  their  mastitis  control  practices,  attitudes  and 

behaviors (Schewe et al., 2015). Additionally, a separate 16 question human resources survey was 

administered 

to  describe  producer/manager  beliefs  and  practices  regarding  employee 

communication,  training  and  education.  Approximately  90  minutes  was  needed  to  conduct  the 

surveys and review the project with each producer.     

 

Milking Dynamics and Parlor Behaviors 

We evaluated milking vacuum with VaDia® digital recorders (Biocontrol, Rakkestad, Norway). 

Four vacuum channels were employed for each individual cow evaluation by attaching 2.4 mm 

(internal diameter) silicon tubing to the following positions on the milking cluster: 1) rear quarter 

liner  mouthpiece  to  record  vacuum  in  the  mouthpiece  chamber  (MPC),  2)  front  quarter  liner 

 

33 

 

mouthpiece, 3) short milk tube (SMT) as a proxy for cluster vacuum, and 4) short pulsation tube 

to record pulsation. Previous recordings by the investigators from 40 cows had demonstrated the 

consistency  between  simultaneous  vacuum  recordings in  the  short  milk  tube  and  actual  cluster 

vacuum by insertion of a needle into the bowl of the cluster (not reported).  

Each cow-milking event was continuously recorded from the time that milking units were attached 

until units were removed, either automatically by milk flow sensors or manually by employees. 

We identified four phases of flow intensity: incline, plateau, decline, and overmilking as described 

by Tančin et al. (2007) for our intrepretive guidelines and determined the following phases for our 

study, start of milking, start of the incline phase of milk flow, start of overmilking (end of the 

decline phase and near static levels in the SMT and MPC vacuum), and end of milking. All vacuum 

recordings  were  downloaded  and  then  reviewed  using  the  VaDia  Suite  software  (Biocontrol, 

Rakkestad,  Norway)  by  two  investigators  (Moore-Foster  and  Erskine)  and  inter-operator 

agreement of interpretation was tested in six herds before the beginning of the trial. We interpreted 

our vacuum data using the principle that milk vacuum in the cluster is the inverse of milk flow, as 

suggested by Schukken et al. (2005). More recently, Penry et al. (2018) determined that increasing 

MPC vacuum is also associated with lower milk flow as a consequence of teat end congestion. 

For each cow, the start of the incline phase was marked when the vacuum level in at least one of 

the two MPC channels decreased to < 13.5 kpa (4 inHg) and the SMT vacuum decreased from 

maximum and fluctuated by ≥ 3.4 kpa (1 inHg). The time interval between the start of milking 

(cluster attachment) and the start of the incline phase was then calculated and termed by the 

VaDia software as “Let Down Time” (LDT). The end of milking was determined to occur when 

vacuum returned to 0 kpa. (Figure 1). We defined cows with a LDT of > 30 s as having DME 

and those with a LDT of ≤ 30 s as having normal milk ejection. This cut off for DME was based 

 

34 

 

on the distribution of LDT for all cows in the study (Figure 2).  Thus, DME was analyzed as a 

dichotomous variable. Bimodal milk ejection (example in lower plot of Figure 1) was not 

observed for any cows that had a LDT of ≤ 30 s. Conversely, 96% of cows with a LDT > 30 s 

demonstrated bimodal milk ejection, and the remaining 4% of the cows in this category 

demonstrated high MPC vacuum and low SMT vacuum fluctuations (indicative of low milk 

flow) for a period of time despite a lack of an initial drop in vacuum after cluster attachment 

(indicative of bimodality).  

During the vacuum analysis for each herd, the quality of air hoses, liner alignment, and air vent 

patency were also recorded. System vacuum capacity was also tested with a “unit fall off test”; 

and results determined by National Mastitis Council standards (National Mastitis Council, 2012).  

In addition to the VaDia and cluster evaluation, we observed milking procedures and protocols to 

determine the milking routine, the lag time and time of tactile stimulation of teats. A minimum of 

four milking strings (defined as one side of a parlor) were recorded during milking preparation in 

parlors. In platform or tie stall milking operations, stimulation and lag times were recorded for at 

least four cows per unit. When there was more than one person milking, milking behaviors were 

observed for a minimum of two milking strings for each person.  

 

Parlor Ergonomics 

Measures of parlor ergonomics were estimated by determining the width of the parlor workspace, 

the height from the floor to the platform, light availability, duration of a milking shift, and the 

availability of breaks during the shift.  Lighting in the parlor was measured with a light meter, both 

in the middle of the parlor floor (mean of three separate locations) and at the level of the teat, under 

 

35 

 

the cow (mean of six to ten locations). Additionally, for parallel, herringbone and side in/side out 

parlors, we measured the length of the parlor floor to derive an estimate of the minimum linear 

distance  that  each  person  milking  cows  might  travel  during  a  milking  shift,  tie  stall  milking 

systems were excluded. This was calculated by the following formula:  

Total Distance per Shift = (D*PTR)*T 

D = Distance each employee traveled per parlor load (both sides of the parlor) of cows (m/load) 

PTR = Parlor turnover rate, or number of milking loads in the parlor per hour (loads/h) 

T = shift length (h) 

The distance each employee traveled for each parlor load  (D) was estimated by measuring the 

length of the parlor floor multiplied by proportion of stalls that each employee visited during each 

pass  of  the  milking  routine.  This  distance  was  then  multiplied  by  the  number  of  passes  in  the 

routine for both sides of the parlor and then doubled (to account for forward and reverse direction 

for each pass).   

The formula for total distance (m) each employee traveled per load of cows was: 

D = (L*P*N)*2a*2b 

L = length of parlor (m) 

P= proportion of stalls prepped by employee 

N = number of premilking passes per cow 

2a = factor to account for forward and reverse direction between passes 

2b = factor to account for duplication of distance for each side of parlor per load of cows 

 

36 

 

For example, if one employee milked in a parlor with 20 stalls (10 cows per side) with a 10 m 

length of the parlor, and three separate passes were made for each cow during preparation or 

cluster attachment:  

D = 10 m (L)*1.0 (P) *3 (N) *2*2= 120 m  

 

Statistical Analysis 

Dependent Variable  

The dependent variable for this study was the proportion of VaDia evaluated cows with DME in 

each herd. The proportion of cows with DME was entered into Microsoft Excel (Microsoft Corp., 

Redmond, WA, USA) for data management and SAS (ver. 9.4; SAS Institute, 2012) was used for 

descriptive and analytical analyses.  

Independent Variables  

The independent variables were divided into three categories 1) management culture and human 

resources, 2) employee behaviors and 3) parlor factors. Management culture and human resource 

variables included milking technician turnover rates (defined as number of milking technicians 

hired per year divided by number of positions available), as well as survey questions that included 

frequency  of  employee  training,  manager  attitudes  about  parlor  turnover  rates  and  how  often 

managers  communicated  with  employees  on  personal  matters.  The  average  pay  rate  and  other 

benefits were also included, if applicable. Employee behaviors included mean total stimulation 

time, lag time, distance traveled per shift, number of stalls per pass and parlor flow rates (described 

as cows milked per hour, cows milked per employee per hour, and milking strings per hour). Parlor 

 

37 

 

factors included rail height, light intensity, frequency of milking per day, if 95% of liners were 

properly aligned within the cluster (yes = 1, no = 0) and if 95% of air vents on the cluster were 

open (yes = 1, no = 0).  

Scale Factor Analysis 

Initially, a frequency distribution was graphed for each independent variable in our model and 

normality was checked by visually inspecting the results of residual analysis and ANOVA. 

Scales were constructed with three to five variables that researchers deemed most relevant. Three 

scales were tested that accounted for pre-milking procedures, management attitudes and parlor 

flow. However, only the pre-milking procedures scale was significant (Eigenvalue > 1 and 

Chronbach’s α > 0.7), which included three variables: mean time of stimulation at first touch, 

mean total time of tactile stimulation (natural-log transformation), and average lag time 

(Eigenvalue = 2.04 and Chronbach’s α = 0.75).  

Bivariate Analysis 

To decide which independent variables were eligible for the multivariable analysis, associations 

between  the  proportion  of  DME  within  a  herd  and  explanatory  continuous  variables  were 

investigated  using  Pearson’s  product  moment  correlation  coefficient.  For  binary  (nominal) 

variables, the dependent variable was compared to the independent variables using a Chi-square 

goodness-of-fit  test.  Any  variables  with  an  initial  cutoff  of  P  <  0.20  (2-tails)  were  considered 

eligible for inclusion in the multivariable model.  

Multiple Linear Regression 

 Using multiple regression and the type-III F-test, an automated backwards-stepwise elimination 

procedure was used to build the final multivariable model until only significant covariates (P < 

 

38 

 

0.05) were retained. Biologically relevant two-way (or first-order) interactions were also evaluated 

in the model, however none were significant. Herd size was also analyzed as a confounder, the 

change  to  the  coefficients  and  R2  values  were  greater  than  10%  for  mean  total  time  of  tactile 

stimulation. Thus, herd size was included as a confounder in the model. The residual distribution 

was assessed visually for normality and homoscedasticity. 

 

Results 

Median herd size was 294 milking cows and ranged from 59 to 2,771 cows (Table 1). Mean 

daily milk per milking cow was 36.8 kg, the three-month DHI geometric mean SCC was 136,800 

cells/mL and 62/64 herds (97%) employed non-family labor. This was representative of DHI 

data for herds in Michigan during 2016; mean herd size was reported to be 257 cows, mean daily 

milk 37.8 kg and mean SCC 157,000 cells/mL (Norman et al., 2017). A total of 3,824 cows 

(mean of 60 ± 29 recordings per herd) were evaluated by VaDia analysis, which averaged 23.5% 

of the milking herd. The mean percentage of cows with DME was 25.0% (Table 1; 95% CI 20.1 

to 30.0%). Mean time of stimulation during the first pre-milking pass was 8.0 s (95% CI 6.6 to 

9.3 s) and mean total stimulation time was 14.2 s (95% CI 11.9 to 16.5 s). Mean lag time, from 

first stimulation to cluster attachment, was 103 s (95% CI 96 to 111 s). Recording flaws (e.g., 

tubes that were disconnected during milking, milking clusters that fell off the udder during 

milking and were re-attached, battery failure, etc.) resulted in less than 5% of the VaDia 

recordings; cows with these events could not be accurately assessed for milk flow dynamics 

relative to DME and were excluded. 

 

39 

 

Based on the bivariable analysis, 28 variables were significant at P < 0.20, however nine 

variables with the lowest P-values and that were most biologically significant were chosen for 

multivariate analysis based on the following categories: 1) herd characteristics 2) employees 3) 

parlor and 4) pre-milking procedures (Table 2). Herd size (natural log), frequency of milkings 

per day, number of milking stalls, milking shift length, cows milked per full time employee-hour, 

number of cows milked per hour (natural log), mean lag time, mean stimulation time (natural log 

transformation), and mean light illumination in the parlor (candela) were all included in the 

multivariate analysis. 

 Using backward stepwise regression, the final multivariable model found that total mean 

stimulation time (natural log) to be negatively and herd size (natural log) to be positively 

associated with the proportion of cows with DME in a herd (Table 3).  

Because of herd size being retained in the final model, we performed a two sample t-test to 

compare means for each of the eligible variables, stratified by herd size, to better understand if 

exclusion of some of the eligible variables was potentially explained by association with herd 

size (Table 4). The cut-off point selected for large (≥ 300 cows) and small (< 300 cows) dairies 

was based on the criteria that we used for the umbrella project of the present study (Schewe et 

al., 2015). Other than milking frequency, mean lag time, and illumination, means for all other 

variables differed (P < 0.05) between small and large herds. Total stimulation time in smaller 

herds (19.2 s) was more than twice that of larger herds (8.8 s). The effect of stimulation time on 

DME was tested for an interaction with herd size, but was not significant (P = 0.67). 

 

 

 

40 

 

Discussion and Conclusions 

 

Our intent was to determine which herd variables, and their estimated relative importance, may 

influence the extent of DME in dairy herds. The 64 herds in this study had a mean geometric 

SCC below the U.S. and Michigan DHI average. Even among this population of herds, we found 

that on average, 25% of cows had delayed milk flow after cluster attachment, which was similar 

to past research that reported proportion of cows with DME between 22% to 35% (Sandrucci et 

al., 2007; Samoré et al., 2011; Watters et al., 2011).  

In our study, herds that stimulated teats for a longer duration of time had a lower proportion of 

cows with DME. This is similar with previous literature that also found pre-milking stimulation 

is an important precursor for oxytocin release to induce milk ejection (Weiss and Bruckmaier, 

2005; Bruckmaier, 2013). Additionally, Watters et al. (2015) reported that the frequencies of 

bimodal milking among cows that had clusters attached immediately, attached after dipping and 

forestripping followed by a 30s lag time, or attached after dipping, forestripping followed by a 

90s lag time, were 21%, 14%, and 7%, respectively.  In a study that was similar in size to ours 

(nearly 2,500 cows), Sandrucci et al. (2007) reported that bimodal milk ejection decreased from 

47% for cows with no udder preparation to 30% for cows that were cleaned and forestripped. As 

much as 60 s of stimulation has been suggested for efficient milk ejection, but stimulation of at 

least 15 s can induce milk ejection, if a lag time of at least 45 s is included (Wagner and Ruegg, 

2002; Weiss and Bruckmaier, 2005; Bruckmaier, 2013). These goals were similar to the mean 

total stimulation time (14.2 s) and lag time (103 s) observed in our study. However, the duration 

of both stimulation and lag time may not be as critical for efficient milk ejection in cows that 

have greater udder filling at milking, such as cows in early lactation, as compared to cows in 

later lactation (Kaskous and Bruckmaier, 2007).  

41 

 

 

Our study disagrees with Wagner and Ruegg (2002) who found that stripping teats before 

milking was not associated with milk flow rates, milk yield or unit on times. However, in that 

study, only 24 multiparous cows from one farm were included and were milked twice per day. 

Furthermore, both groups of cows (pre-stripped or not pre-stripped) in that study were vigorously 

dried with towels for 10 to 15 s before cluster attachment, although this procedure was done just 

before cluster attachment. In our study, we had considerable diversity in milking systems and 

protocols, and 41% of herds (26 of 64) were milked three times per day. Thus, our project may 

have offered a broader scope of variables that need to be considered in the relationship between 

stimulation and DME.    

Sandrucci et al. (2007) found that as farm size increased, less time was spent on teat stimulation, 

therefore reducing time for udder preparation. Our study agreed with these results, in finding that 

total stimulation time in smaller herds was more than twice that of larger herds. Conversely, the 

length of a work shift, the number of cows that were milked per employee-hour and the number 

of milking units each employee operated during the milking routine were greater in larger herds. 

The increased workload of employees in larger herds compared to smaller herds suggests there 

may be more emphasis on cow throughput, or parlor efficiency, in larger dairies. Thus, taken 

together with the increased risk of DME, larger herds may be sacrificing milking efficiency for 

parlor efficiency. Additionally, larger dairies have more complex employee structures and many 

dairies report difficulty in training and maintaining protocols, which is associated with higher 

bulk tank SCC (Erskine et al., 2015; Schewe et al., 2015). This problem could be potentially 

exacerbated in herds that have a protocol culture that leads to a higher proportion of cows with 

DME.  

 

42 

 

Surprisingly, we did not find an association between DME and lag time between preparation and 

cluster attachment. Previous reports found that lag times are a key factor in reducing bimodal 

milk flow and improving overall milking time efficiency (Weiss and Bruckmaier, 2005; Watters 

et al, 2012).  Additionally, the frequency of bimodal milk ejection among cows without a lag 

interval was 54%, but decreased to 36%  and 24% for cows with a lag time of < 60 s and ≥ 60 s, 

respectively (Sandrucci et al., 2007). However, in our study herds the mean lag time across all 

herds was within the suggested goals for efficient milk ejection, and the minimum was 33 s. 

Thus, the overwhelming proportion of herds had a lag time that was considered to be sufficient 

to augment oxytocin-induced milk ejection, which could have masked the importance of this 

variable in our model. In contrast, the range of variation in teat stimulation was nearly 20 fold 

across herds which allowed for more discrimination of comparisons among our study herds. Past 

research has also suggested an overriding effect of teat stimulation relative to other factors 

(Weiss and Bruckmaier, 2005; Ambord and Bruckmaier, 2009). Thus, the more critical question 

raised from this study may regard why some herds attain more stimulation than others.  

There were several limitations in this study, e.g., our inability to reliably evaluate vacuum 

dynamics in cows that had a milking cluster fall off during milking. It is possible that some of 

these interruptions in milking may be related to DME. However, the proportion of these events 

was low relative (< 5%) to the total number of cows evaluated and in most instances, 

determining the time points of the milking vacuum curve that included a detached milking unit 

was deemed to be highly subjective.  

To our knowledge, this is the largest study that has used vacuum recordings, rather than an 

electronic milk flow meter, as an indicator for milk flow dynamics. Despite this extrapolation, 

the distinct partitioning of cows that began the start of the incline phase of milking either ≤ 30 s 

 

43 

 

or those that were distributed > 30 s was remarkable and readily apparent when reviewing a 

scatterplot (not shown) of the histogram in Figure 2. Our dichotomous criteria for DME attained 

very similar results in terms of the proportion of milkings that resulted in delayed LDT and 

bimodal ejection in previous studies (Sandrucci et al., 2007; Samoré et al., 2011) and the 

association between stimulation and DME.   

The sample size of cows within each herd depended on herd size, milking shift length, and cow 

throughput while milking. In smaller herds, (milking shifts less than four hours) we were present 

throughout the entire milking. However, especially in larger herds, it was difficult to evaluate all 

milking groups of cows across all employee shifts. Thus, in herds with short intervals between 

milking shifts, we recorded milking events in portions of two consecutive shifts to gain a wider 

perspective of milking behaviors. Also, we attached VaDia recorders on several milking clusters 

in each herd to capture any variation resulting from milking position among cows. Finally, 

numerous individual cow factors such as stage of lactation, genetics, teat anatomy and parity 

have been reported as influencing milk ejection, which were not accounted for in our herd-level 

analyses.  

 

Summary 

Decreasing the amount of stimulation during the pre-milking routine was the most critical factor 

that increased the likelihood of DME in dairy herds. Increased frequency of DME was more 

likely as herd size increased. Although other variables were found to be associated with DME 

from bivariable analysis, the low level of variation among herds for some variables such as lag 

time, or the effect that herd size may have on other variables, possibly masked the importance of 

 

44 

 

these variables in our multivariable model. Given the strength of the relationship between teat 

stimulation and DME in our model, further research should investigate the management 

procedures that are directly associated with teat stimulation. 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

45 

 

Acknowledgements 

This project was supported by Agriculture and Food Research Initiative Competitive Grant no. 

2013-68004-20439 from the USDA National Institute of Food and Agriculture. The authors 

would like to acknowledge participating dairy producers and their employees for their 

willingness to aid in data collection. Also, the authors thank Leah Girard, Ellen Launstein, 

Caitlin McNichols and Trevor Walling for their technical assistance.  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

46 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

APPENDICES 

47 

 

Table 1. Descriptive data of herd profile and management culture variables from 64 Michigan dairy herds.    

Appendix A. Tables 

 

Variable 

Percentage of cows with DME1  

Mean total stimulation time (s) 

Median overmilking (s) 

Herd size (Milking cows) 

Milking frequency per day 

Milking units in parlor or barn 

Time parlor in operation per day (h) 

Employee shift length (h) 

Employee break length (min) 

Number of milking personnel per shift 

Mean 
(LCL, UCL)2 
25.0  
(20.0, 30.0) 
14.2  
(11.9, 16.5) 
47.2            
(38.6, 55.9) 
451  
(324, 578) 
2.41 
(2.28, 2.52) 
20.5 
(17.3, 23.7) 
12.4  
(10.9, 14.1) 
5.2  
(4.5, 5.9) 
10.2 
(4.7, 15.8) 
2.0 
(1.8, 2.2) 

Minimum  25th 

Median  75th 

Maximum 

percentile 

percentile 

0 

2.4 

9.0 

59 

6.8 

7.8 

21.0 

136 

23.3 

32.8 

11.2 

17.1 

40.0 

61.5 

75.0 

40.8 

201 

294 

484 

2,771 

2.00 

2.00 

2.00 

3.00 

3.00 

5.00 

12.00 

16.00 

24.00 

72.00 

3.00 

7.00 

12.50 

17.00 

24.00 

1.3 

0.0 

1.0 

3.0 

0.0 

2.0 

4.4 

0.0 

2.0 

6.9 

0.0 

2.0 

12.0 

90.0 

5.0 

9.00 

Milking personnel did other chores during milking shift3  0.34 m           

0.00 

0.00 

0.00 

0.00 

(0.05, 0.64) 

48 
 

 

(Table 1 cont’d) 
Number of milking units operated per milking personnel  10.6 

 

Parlor turnover rate (number of times the parlor emptied 
and filled in an hour) 

Number of cows milked per h 

Number of cows milked per h per individual milker 

Proportion of milking employee turnover per yr  

Proportion of total farm employee turnover per yr 

Hourly pay6 ($) 

Other benefits received3 

Mean lag time (s)4  

Mean time of stimulation at first touch (s) 

Are cluster vent holes open?3 

Are liners properly installed?3 

Mean illumination milking area (candela) 

 
 

(9.4, 11.8) 
4.4  
(4.1, 4.6) 
85  
(71, 100) 
45 
(40, 49) 
0.47 
(0.30, 0.63) 
0.40  
(0.24, 0.56) 
9.79  
(8.88, 10.70) 
0.66  
(0.53, 0.77) 
103 
(96, 110) 
8.0  
(6.6, 9.3) 
0.80  
(0.70, 0.90) 
0.94 
(0.88, 1.00) 
488 
(387, 589) 
 

49 
 

 
2.5 

2.8 

32 

19 

 
8.0 

3.7 

49 

32 

 
10.0 

 
12.0 

4.3 

4.6 

67 

39 

103 

56 

0.00 

0.00 

0.33 

0.50 

0.00 

0.00 

0.25 

0.50 

 
24.0 

8.9 

360 

120 

3.00 

3.00 

0.00 

10.00 

11.00 

11.50 

15.00 

0.00 

0.00 

1.00 

1.00 

33 

2.0 

84 

3.6 

102 

123 

6.3 

11.2 

0.00 

1.00 

1.00 

1.00 

0.00 

1.00 

1.00 

1.00 

1.00 

165 

21.1 

1.00 

1.00 

11 

 

199 

421 

650 

1,845 

 

 

 

 

 

(Table 1 cont’d) 
Mean illumination at the level of the teat (candela) 

Length of the parlor (m) 

Number of passes per cow during milking prep  

Number of cows prepped per pass  

0.20 

Proportion of stalls prepped per milking personnel  
 
Distance traveled per milking personnel per milking 
shift (m) 
 
1 DME=Delayed Milk Ejection; time between milking cluster and milk let down > 30s 
2 LCL, UCL=Lower and Upper Confidence Limits 
30 = no and 1 = yes 
4Time from first touch to unit attachment, seconds 
5Pay of $0.00 was reported for some family members 
 
 
 

108 

 

 
6 

 
17 

 
33 

 
66 

 
251 

0.00 

6.79 

8.40 

12.05 

27.45 

1.0 

0.0 

2.0 

3.0 

0.50 

200 

 

3.0 

4.5 

3.0 

6.0 

0.50 

0.50 

4.0 

10.0 

1.00 

383 

614 

1417 

 

 

 

 
48 
(37, 60) 
  
9.68 
(8.40, 10.96) 
2.7  
(2.5, 2.9) 
4.9 
(4.45, 5.5) 
0.59 
(0.53, 0.65) 
483  
(400, 566) 
 

50 
 

 

Table 2. Bivariate analysis results for explanatory variables of herd-level proportion of cows 
with delayed milk ejection in 64 Michigan dairy herds  

 

 

Variable 

 

Herd size (natural log) 

Milking frequency per day 

Number of milking units 

Employee shift length (h) 

Cows milked per fulltime 
employee-hour (natural log) 
 
Milking units operated per 
employee 
Mean lag time (s) 

Mean total stimulation time 
(s; natural log) 
Mean illumination in the 
parlor (candela) 

P-value 

<.0001 
 
 0.06 
 
0.0004 
 
0.0048 
 
0.0054 
 

0.0078 
 
0.0080 
 
<.0001 
 
0.0256 

51 
 

 

Table 3. Final linear model for associations between delayed milk ejection and herd-level 
variables in 64 Michigan dairies 

Parameter 
 
Intercept 

 
Mean total 
stimulation time 
(natural log) 

Estimate 
 
4.32 

 
-1.30 

Herd size (natural 
log) 

0.599 

SE 
 
2.38 

 
0.43 

0.28 

P-value 
 
0.074 
 
 
<.005 

0.034 

R2 
 
0.356 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

52 
 

 

Table 4. Comparison of means for eligible independent variables for multivariable model 
between small (<300 cows) and large herds (≥300 cows) in 64 Michigan dairy herds 

 

Variable 

Milking frequency 
per day 
 
Number of 
milking units  
 
Employee shift 
length (h) 
 
Cows milked per 
full-time employee 
hr (natural log) 
 
Units operated per 
employee 
 
Mean lag time (s) 

 
Mean total 
stimulation time 
(s, natural log) 
 
Mean illumination 
in parlor (candela) 

Small herds mean 
(SEM) 
2.47 (0.09) 

Large herds mean 
(SEM) 
2.33 (0.09) 

 
13.47 (0.96) 

 
3.52 (0.27) 

 
26.70 (2.50) 

 
6.82 (0.46) 

P-value 

0.27 

 
<.05 

 
<.05 

3.58 (0.06) 

3.85 (0.07) 

<.05 

8.67 (0.50) 

12.35 (0.94) 

<.05 

110.0 (5.70) 

96.90 (4.70) 

 
2.82 (0.10) 

 
2.16 (0.09) 

 
396 (73) 

 
564 (68) 

0.08 

 
<.05 

 
0.10 

 

 

 

 

 

 

 

 

 

 

53 
 

 

Appendix B. Figures 

 

Figure 1. The upper plot is an example of VaDia® digital recording showing uninterrupted milk 
ejection. Vacuum is recorded (kpa) on the vertical axis. Each hash mark indicates a 15 s time 
interval on the horizontal axis.  Channel 1 (red) is indicates rear mouthpiece chamber vacuum, 
channel 2 (blue) front mouthpiece chamber vacuum, and channel 3 (green) short milk tube 
vacuum.  Symbols mark the start of milking (↓), start of incline phase of milk flow (▲), and 
end of milking (■). The lower plot is an exxample of VaDia® digital recording showing bimodal 
milk ejection. Vacuum is recorded (kpa) on the vertical axis. Each hash mark indicates a 15 s 
time interval on the horizontal axis. Channel 1 (red) indicates the rear mouthpiece chamber 
vacuum, channel 2 (blue) indicates the front mouthpiece chamber vacuum, channel 3 (green) 
indicates the short milk tube vacuum as a proxy for cluster vacuum.  Symbols mark the start of 
milking (↓), start of bimodal ejection (●), start of incline phase of milking (▲), and end of 
milking (■). 

 

 ↓▲                                                                                  
■

 ↓ ●           ▲                                                                  
■

 

54 
 

 

Figure 2. Histogram of time period between cluster attachment and attaining the incline phase of 

milk ejection (Time to milk let down) for 3,824 cows as recorded by digital vacuum. 

2500

2000

1500

1000

500

0

s
w
o
C

 
f
o

 
r
e
b
m
u
N

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

≤ 10  11 to 20 21 to 30 31 to 40 41 to 50 51 to 60 61 to 70 71 to 80 81 to 90 91 to
100

>101

Time of Milk Let Down (s)

 

 

55 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

REFERENCES 

56 
 

 

 

REFERENCES 

Ahmadzadeh, A., F. Frago, B. Shafii, J. C. Dalton, W. J. Price, and M. A. McGuire. 2009. Effect 
of clinical mastitis and other diseases on reproductive performance of Holstein cows. Anim. 
Reprod. Sci. 112:273-282. 
 
Archer, S. C., F. Mc Coy, W. Wapenaar, and M. J. Green. 2013. Association of season and herd 
size with somatic cell count for cows in Irish, English, and Welsh dairy herds. The Veterinary 
Journal 196:515-521. 
 
Baker, D. and D. Chappelle. 2012. Health status and needs of Latino dairy farmworkers in 
Vermont. Journal of Agromedicine 17:277-287. 
 
Barkema, H. W., J. D. Van der Ploeg, Y. H. Schukken, T. J. G. M. Lam, G. Benedictus, and A. 
Brand. 1999. Management Style and Its Association with Bulk Milk Somatic Cell Count and 
Incidence Rate of Clinical Mastitis. J. Dairy Sci. 82:1655-1663. 
 
Barlow, J. W., L. J. White, R. N. Zadoks, and Y. H. Schukken. 2009. A mathematical model 
demonstrating indirect and overall effects of lactation therapy targeting subclinical mastitis in 
dairy herds. Prev. Vet. Med. 90:31-42. 
 
Barnouin, J., S. Bord, S. Bazin, and M. Chassagne. 2005. Dairy Management Practices 
Associated with Incidence Rate of Clinical Mastitis in Low Somatic Cell Score Herds in France. 
J. Dairy Sci. 88:3700-3709. 
 
Barnouin, J., M. Chassagne, S. Bazin, and D. Boichard. 2004. Management Practices from 
Questionnaire Surveys in Herds with Very Low Somatic Cell Score Through a National Mastitis 
Program in France. J. Dairy Sci. 87:3989-3999. 
 
Bartlett, P. C., G. Y. Miller, C. R. Anderson, and J. H. Kirk. 1990. Milk Production and Somatic 
Cell Count in Michigan Dairy Herds. J. Dairy Sci. 73:2794-2800. 
 
Bewley, J., R. W. Palmer, and D. B. Jackson-Smith. 2001. An overview of experiences of 
Wisconsin dairy farmers who modernized their operations. J. Dairy Sci. 84:717-729. 
 
Breen, J. E., M. J. Green, and A. J. Bradley. 2009. Quarter and cow risk factors associated with 
the occurrence of clinical mastitis in dairy cows in the United Kingdom. J. Dairy Sci. 92:2551-
2561. 
 
Bruckmaier, R. M. 2005. Normal and disturbed milk ejection in dairy cows. Domest. Anim. 
Endocrinol. 29:268-273. 
 

57 
 

 

Bruckmaier, R. M. 2013. Oxytocin from the pituitary or from the syringe: Importance and 
Consequences for milking machine in dairy cows. Pages 4-11 in Proc. NMC Annual Meeting 
2013. 
 
Bruckmaier, R. M., Wellnitz, O. 2007. Induction of milk ejection and milk removal in different 
production systems. J. Anim. Sci. 86:15-20. 
 
Cha, E., D. Bar, J. A. Hertl, L. W. Tauer, G. Bennett, R. N. González, Y. H. Schukken, F. L. 
Welcome, and Y. T. Gröhn. 2011. The cost and management of different types of clinical 
mastitis in dairy cows estimated by dynamic programming. J. Dairy Sci. 94:4476-4487. 
 
Chassagne, M., J. Barnouin, and M. Le Guenic. 2005. Expert Assessment Study of Milking and 
Hygiene Practices Characterizing Very Low Somatic Cell Score Herds in France. J. Dairy Sci. 
88:1909-1916. 
 
 
Cross, J. A. 2006. RESTRUCTURING AMERICA'S DAIRY FARMS*. Geographical Review  
96:1-23. 
 
DÁVila, A., M. T. Mora, and R. GonzÁLez. 2011. English-Language Proficiency and 
Occupational Risk Among Hispanic Immigrant Men in the United States. Industrial Relations: A 
Journal of Economy and Society 50:263-296. 
 
Dohoo, I. R. and A. H. Meek. 1982. Somatic Cell Counts in Bovine Milk. The Canadian  
Veterinary Journal 23:119-125. 
 
Douphrate, D. I., G. R. Hagevoort, M. W. Nonnenmann, C. L. Kolstrup, S. J. Reynolds, M. 
Jakob, and M. Kinsel. 2013. The Dairy Industry: A Brief Description of Production Practices, 
Trends, and Farm Characteristics Around the World. Journal of Agromedicine 18:187-197. 
 
Dufour, S., A. Frechette, H. W. Barkema, A. Mussell, and D. T. Scholl. 2011. Effect of udder 
health management practices on herd somatic cell count. J. Dairy Sci. 94:563-579. 
 
Dustmann, C. and F. Fabbri. 2003. Language proficiency and labour market performance of 
immigrants in the UK*. The Economic Journal 113:695-717. 
 
Erskine, R. J., R. J. Eberhart, L. J. Hutchinson, and S. B. Spencer. 1987. Herd management and 
prevalence of mastitis in dairy herds with high and low somatic cell counts. J. Am. Vet. Med. 
Assoc. 190:1411-1416. 
 
Erskine, R. J., R. O. Martinez, and G. A. Contreras. 2015. Cultural lag: A new challenge for 
mastitis control on dairy farms in the United States. J. Dairy Sci 98:8240-8244. 
 
Erskine, R. J., S. Wagner, and F. J. DeGraves. 2003. Mastitis therapy and pharmacology. Vet. 
Clin. North Am. Food Anim. Pract. 19:109-138. 
 

58 
 

 

Fogleman, S., Milligan, R., Maloney, T., Knoblauch, W. 1999. Employee Compensation and Job 
Satisfaction on Dairy Farms in the Northeast. 1999 Annual Meeting, Aug 8-11, Nashville, TN 
from American Agricultural Economics Association. 
 
Fox, L. 2013. Can Milk Somatic Cells Get too Low? A Question to be Revisited. Pages 56-63 in 
Proc. Annual National Mastitis Conference. National Mastitis Council. 
 
Fuenzalida, M. J., P. M. Fricke, and P. L. Ruegg. 2015. The association between occurrence and 
severity of subclinical and clinical mastitis on pregnancies per artificial insemination at first 
service of Holstein cows. J. Dairy Sci 98:3791-3805. 
 
Goodger, W. J., T. Farver, J. Pelletier, P. Johnson, G. DeSnayer, and J. Galland. 1993. The 
association of milking management practices with bulk tank somatic cell counts. Prev. Vet. Med. 
15:235-251. 
 
Gruet, P., Maincent, P. , Berthelot, X. , Kaltsatos, V. 2001. Bovine Mastitis and IM drug delivery 
review and perspectives. Advanced Drug Delivery Reviews:245-259. 
 
Hadley, G. L., S. B. Harsh, and C. A. Wolf. 2002. Managerial and Financial Implications of 
Major Dairy Farm Expansions in Michigan and Wisconsin. J. Dairy Sci. 85:2053-2064. 
 
Hagevoort, G. R., D. I. Douphrate, and S. J. Reynolds. 2013. A review of health and safety 
leadership and managerial practices on modern dairy farms. Journal of agromedicine 18:265-
273. 
 
Hagnestam-Nielsen, C., U. Emanuelson, B. Berglund, and E. Strandberg. 2009. Relationship 
between somatic cell count and milk yield in different stages of lactation. J. Dairy Sci. 92:3124-
3133. 
 
Halasa, T., M. Nielen, A. P. W. De Roos, R. Van Hoorne, G. de Jong, T. J. G. M. Lam, T. van  
Werven, and H. Hogeveen. 2009. Production loss due to new subclinical mastitis in Dutch dairy 
cows estimated with a test-day model. J. Dairy Sci. 92:599-606. 
 
Hand, K. J., A. Godkin, and D. F. Kelton. 2012. Milk production and somatic cell counts: A 
cow-level analysis. J. Dairy Sci. 95:1358-1362. 
 
Harrison, J., Lloyd, Sarah and O'Kane, Trish. 2009a. Immigrant Dairy Workers in Rural 
Wisconsin - Briefing 4. in Program on Agricultural Technology Studies. M. Univeristy of 
Wisconsin, ed, University of Wisconsin, Madison. 
 
Harrison, J., Lloyd, Sarah and O'Kane, Trish. 2009b. Overview of Immigrant Workers on 
Wisconsin Dairy Farms - Briefing 1. in Program on Agricultural Technology Studies. 
Heikkilä, A. M., J. I. Nousiainen, and S. Pyörälä. 2012. Costs of clinical mastitis with special 
reference to premature culling. J. Dairy Sci. 95:139-150. 
 

59 
 

 

Hogan, J. and K. L. Smith. 2012. Managing Environmental Mastitis. Vet. Clin. North Am. Food 
Anim. Pract. 28:217-224. 
 
Hogeveen, H., K. Huijps, and T. Lam. 2011. Economic aspects of mastitis: New developments. 
N. Z. Vet. J. 59:16-23. 
 
Holmes, S. M. 2011. Structural vulnerability and hierarchies of ethnicity and citizenship on the 
farm. Med. Anthropol. 30:425-449. 
 
Jackson-Smith, D., Barham, B. . 2000. Dynamics of Dairy Industry Restructuring in Wisconsin. 
Research in Rural Sociology and Development 8:115-139. 
 
Jansen, J. and T. J. Lam. 2012. The role of communication in improving udder health. Vet. Clin. 
North Am. Food Anim. Pract. 28:363-379. 
 
Jayarao, B. M., S. R. Pillai, A. A. Sawant, D. R. Wolfgang, and N. V. Hegde. 2004. Guidelines 
for monitoring bulk tank milk somatic cell and bacterial counts. J. Dairy Sci. 87:3561-3573. 
 
Jenkins, P. L., S. G. Stack, J. J. May, and G. Earle-Richardson. 2009. Growth of the Spanish-
Speaking Workforce in the Northeast Dairy Industry. Journal of Agromedicine 14:58-65. 
 
Kaneene, J. B. and A. S. Ahl. 1987. Drug Residues in Dairy Cattle Industry: Epidemiological 
Evaluation of Factors Influencing Their Occurrence. J. Dairy Sci. 70:2176-2180. 
 
Kayitsinga, J., R. L. Schewe, G. A. Contreras, and R. J. Erskine. 2017. Antimicrobial treatment 
of clinical mastitis in the eastern United States: The influence of dairy farmers' mastitis 
management and treatment behavior and attitudes. J. Dairy Sci. 
 
Kehrli, M. E., Jr. and D. E. Shuster. 1993. Factors Affecting Milk Somatic Cells and Their Role 
in Health of the Bovine Mammary Gland. J. Dairy Sci. 77:619-627. 
 
Khaitsa, M. L., T. E. Wittum, K. L. Smith, J. L. Henderson, and K. H. Hoblet. 2000. Herd 
characteristics and management practices associated with bulk-tank somatic cell counts in herds 
in official Dairy Herd Improvement Association programs in Ohio. Am. J. Vet. Res. 61:1092-
1098. 
 
Kirkpatrick, M. A., Olson, Jerry D. 2015. Somatic Cell Counts at First Test: More Than a 
Number. Pages 53-56 in National Mastitis Council Annual Meeting 2015. NMC, Memphis, TN. 
 
Klei, L., J. Yun, A. Sapru, J. Lynch, D. Barbano, P. Sears, and D. Galton. 1998. Effects of Milk 
Somatic Cell Count on Cottage Cheese Yield and Quality. J. Dairy Sci. 81:1205-1213. 
 
Kolstrup, C. L. 2012. What factors attract and motivate dairy farm employees in their daily 
work? Work (Reading, Mass.) 41:5311-5316. 
Labor, U. S. D. o. 2005. Findings from the National Agricultural Workers Survey (NAWS) 
2001-2002. Research Report No. 9. 

60 
 

 

 
LeGassick, J. 2013. Liners should not be the first to blame. Pages 72-73 in Progressive 
Dairyman. Progressive Dairyman. 
 
Loeffler, S. H., M. J. de Vries, and Y. H. Schukken. 1999. The Effects of Time of Disease 
Occurrence, Milk Yield, and Body Condition on Fertility of Dairy Cows. J. Dairy Sci. 82:2589-
2604. 
 
Loh, K. and S. Richardson. 2004. Foreign-born workers: trends in fatal occupational injuries, 
1996-2001. Mon. Labor Rev. 127:42-53. 
 
Losinger, W. C. 2005. Economic impacts of reduced milk production associated with an increase  
in bulk-tank somatic cell count on US dairies. J. Am. Vet. Med. Assoc. 226:1652-1658. 
 
Ma, Y., C. Ryan, D. M. Barbano, D. M. Galton, M. A. Rudan, and K. J. Boor. 2000. Effects of 
Somatic Cell Count on Quality and Shelf-Life of Pasteurized Fluid Milk1. J. Dairy Sci. 83:264-
274. 
 
MacDonald, J. and D. Newton. 2014. Milk Production Continues Shifting to Large-Scale Farms. 
Amber Waves:7-1E,2E,3E,4E,5E,6E,7E. 
 
Mein, G. A., Neijenhuis, F., Morgan, W.F., Reinemann, D.J., Hillerton, J.E, Baines, J.R.,  
Ohnstad, I., Timms, L., Britt, J.S., Farnsworth, R., Cook, N., Hemling, T. 2001. Evaluation of 
bovine teat condition in commercial dairy herds: 1. Non-infectious factors. in Proc. 2nd 
International Symposium on Mastitis and Milk Quality. 
 
Mitchell, R. J., Williamson, A.M. 2000. Evaluation of an 8  hour versus a 12 hour shift roster on 
employees at a power station. Appl. Ergon. 31:83-93. 
 
Neijenhuis, F., H. W. Barkema, H. Hogeveen, and J. P. T. M. Noordhuizen. 2000. Classification 
and longitudinal examination of callused teat ends in dairy cows. J. Dairy Sci. 83:2795-2804. 
 
Neijenhuis, F., H. W. Barkema, H. Hogeveen, and J. P. T. M. Noordhuizen. 2001. Relationship 
between teat-end callosity and occurrence of clinical mastitis. J. Dairy Sci. 84:2664-2672. 
 
NMC. 1999. Teat Lesions Can Lead to Milking Problems, Mastitis. NMC. 
 
NMC. 2013. Reccomended MIlking Protocols. in www.nmconline.org. N. M. Council, ed. 
National Mastitis Council, Verona, WI. 
 
Norman, H. D., Walton, L.M., Durr, J. 2015. Somatic cell counts of milk from Dairy Herd 
Improvement herds during 2015. Council on Dairy Cattle Breeding. 
Paduch, J.-H., E. Mohr, and V. Krömker. 2012. The association between teat end hyperkeratosis 
and teat canal microbial load in lactating dairy cattle. Vet. Microbiol. 158:353-359. 
 

61 
 

 

Rasmussen, M. D. 2004. Overmilking and teat condition. Pages 169-175 in National Mastitis 
Council Annual Meeting. NMC. 
 
Raubertas, R. F. and G. E. Shook. 1980. Relationship Between Lactation Measures of Somatic 
Cell Concentration and Milk Yield<sup>1</sup>. J. Dairy Sci. 65:419-425. 
 
Román-Muñiz, I. N., Van Metre, D. C., Garry, F. B., & Smith, R. A. 2007. Dairy Worker 
Training Experiences. Pages 20-22 in Proc. Fortieth Annual Conference, American Association 
of Bovine Practitioners, Vancouver, British Columbia, Canada. 
 
Royster, E. and S. Wagner. 2015. Treatment of Mastitis in Cattle. Vet. Clin. North Am. Food 
Anim. Pract. 31:17-46. 
 
Sandrucci, A., A. Tamburini, L. Bava, and M. Zucali. 2007. Factors Affecting Milk Flow Traits 
in Dairy Cows: Results of a Field Study. J. Dairy Sci. 90:1159-1167. 
 
Schepers, A. J., T. J. G. M. Lam, Y. H. Schukken, J. B. M. Wilmink, and W. J. A. Hanekamp. 
1997. Estimation of Variance Components for Somatic Cell Counts to Determine Thresholds for 
Uninfected Quarters. J. Dairy Sci. 80:1833-1840. 
 
Schewe, R. L., J. Kayitsinga, G. A. Contreras, C. Odom, W. A. Coats, P. Durst, E. P. Hovingh, 
R. O. Martinez, R. Mobley, S. Moore, and R. J. Erskine. 2015. Herd management and social 
variables associated with bulk tank somatic cell count in dairy herds in the eastern United States. 
J. Dairy Sci 98:7650-7665. 
 
Seegers, H., C. Fourichon, and F. Beaudeau. 2003. Production effects related to mastitis and 
mastitis economics in dairy cattle herds. Vet. Res. 34:475-491. 
 
Smith, S. M., T. Perry, and D. Moyer. 2006. Creating a Safer Workforce. Prof. Saf. 51:20-25. 
 
Stup, R. E., J. Hyde, and L. A. Holden. 2006. Relationships between selected human resource 
management practices and dairy farm performance. J. Dairy Sci. 89:1116-1120. 
 
Swinkels, J. M., A. Hilkens, V. Zoche-Golob, V. Krömker, M. Buddiger, J. Jansen, and T. J. G. 
M. Lam. 2015. Social influences on the duration of antibiotic treatment of clinical mastitis in 
dairy cows. J. Dairy Sci 98:2369-2380. 
 
Tančin, V., A. H. Ipema, and P. Hogewerf. 2007. Interaction of Somatic Cell Count and Quarter 
Milk Flow Patterns. J. Dairy Sci 90:2223-2228. 
 
USDA, A., VS. 2013. Determining US MIlk Quality Using Bulk-tank Somatic Cell Counts. A. 
United States Department of Agriculture, VS, , ed. 
 
USDA, A., VS, CEAH. 2008. Antibiotic Use on U.S. Dairy Operations, 2002 and 2007. Page 5. 
A. USDA, VS, CEAH, ed. USDA, Fort Collins, CO. 
 

62 
 

 

USDA, A., VS, NAHMS. 2014. Dairy 2014 milk quality, mIlking procedures and mastitis on 
U.S. dairies, 2014. Fort Collins, CO. 
 
Vaarst, M., B. Paarup-Laursen, H. Houe, C. Fossing, H.J. Andersen. 2002. Farmers' Choice of 
Medical Treatment of Mastitis in Danish Dairy Herds Based on Qualitative Research Interviews. 
J. Dairy Sci. 85:992-1001. 
 
Vila, B., G. B. Morrison, and D. J. Kenney. 2002. Improving Shift Schedule and Work-Hour 
Policies and Practices to Increase Police Officer Performance, Health, and Safety. Police 
Quarterly 5:4-24. 
 
von Keyserlingk, M. A., N. P. Martin, E. Kebreab, K. F. Knowlton, R. J. Grant, M. Stephenson, 
C. J. Sniffen, J. P. Harner, 3rd, A. D. Wright, and S. I. Smith. 2013. Invited review: 
Sustainability of the US dairy industry. J. Dairy Sci. 96:5405-5425. 
 
Wagner, S. A. and R. J. Erskine. 2009. CHAPTER 101 - Decision Making in Mastitis Therapy. 
Pages 502-509 in Food Animal Practice (Fifth Edition). D. E. A. M. Rings, ed. W.B. Saunders, 
Saint Louis. 
 
Weiss, D. and R. M. Bruckmaier. 2005. Optimization of individual prestimulation in dairy cows. 
J. Dairy Sci. 88:137-147. 
 

Wenz, J. R., S. M. Jensen, J. E. Lombard, B. A. Wagner, and R. P. Dinsmore. 2007. Herd 
Management Practices and Their Association with Bulk Tank Somatic Cell Count on United 
States Dairy Operations. J. Dairy Sci. 90:3652-3659. 

 

 

 

 

 

 

 

 

 

 

63 
 

 

 

 

CHAPTER 3 

STIMULATION TIME IN MICHIGAN DAIRY HERDS 

Herd level variables associated with premilking stimulation time in Michigan dairy herds 

R.  Moore-Foster*,  B.  Norby*,  R.  L.  Schewe†,  R.  Thomson‡,  P.  C.  Bartlett*,  and  R.  J. 
Erskine*1 
  
  
 
 

*Department of Large Animal Clinical Sciences  
and 
‡Department of Animal Science 
Michigan State University 
East Lansing, 48824 
 
† Department of Sociology 
Syracuse University, Syracuse, NY 13244  
 
 
 
1Corresponding Author 
R. J. Erskine 
736 Wilson Rd 
East Lansing, 48824 
517-355-9593 
Fax: 517-432-1042 
erskine@msu.edu  

 

 

 

 

 

64 
 

 

Abstract 

The objective of this study was to determine the herd level variables that were associated with total 

stimulation  time  during  the  premilking  routine  in  64  Michigan  dairy  herds  with  non-family 

employees. The mean herd size was 452 cows (range 59 to 2,771 cows) and the three month DHI 

geometric  mean  SCC  was  136,795  cells/mL.  For  each  herd,  surveys  were  administered  to  the 

producers to gather mastitis management practices and attitudes. Additionally, milking protocols 

were  observed  and  milk  flow  dynamics  were  measured  by  use  of  digital  vacuum  recorders. 

Backwards multivariate regression analysis was used to determine which of 47 herd-level milking 

and management variables were associated with mean duration of total stimulation time. Mean 

total  stimulation  time  was  14.2  s  (range  of  2.4  to  40.8  s)  and  was  positively  associated  with 

increasing  lag  time  (time  interval  between  first  stimulation  and  cluster  attachment).  Total 

stimulation  time  was  negatively  associated  with  greater  time  intervals  between  milkings  and 

increasing  number  of  passes  to  each  cow  in  the  premilking  routine.  In  summary,  increased 

stimulation time is more likely in herds that foster a lower sense of urgency of cow throughput 

during milking, as evidenced by an association with longer lag times despite fewer preparation 

passes per cow and longer intervals between milkings.  

Key words: stimulation time, milking protocols, employee management 

65 
 

 

Introduction 

The anatomy of the bovine udder divides milk storage; 20% of the milk is held in the milk 

cistern, known as “cisternal milk”, and 80% is in the alveoli and ductules, known as “alveolar or 

glandular milk” (Bruckmaier, 2005). While cisternal milk can be readily released, glandular milk 

requires the contraction of myoepithelial cells that surround the mammary alveoli to eject the 

milk through the ducts into the cisternal compartment (Bruckmaier and Wellnitz, 2007). 

Premilking stimulation of the teats is necessary to induce a neural reflex arc that causes oxytocin 

release from the pituitary gland into the circulation, which in turn activates the myoepithelial 

cells to force milk ejection (Bruckmaier, 2005). Weiss and Bruckmaier (2005) recommended that 

the duration of pre-stimulation before milking should be 30 to 60 s, however with low udder 

filling, longer pre-stimulation periods of 90 s can be helpful, which may not be practical in some 

parlor environments. Also, the intensity of stimulation is less important than the duration (Weiss 

and Bruckmaier, 2005). However, stimulation does not need to occur all at once and can be split 

into shorter tactile sessions, followed by a latency of 45 s (Bruckmaier, 2013). Because of the 

short plasma half-life of oxytocin, no more than 2 minutes should pass between tactile 

stimulation and milking unit attachment (Bruckmaier, 2013). As greater myoepithelial 

contraction is needed to eject milk out of an incompletely filled, as compared to, filled alveolus 

(Bruckmaier, 2007), udder stimulation is more important for late lactation cows as compared to 

recently calved cows (Sandrucci et al., 2007). Additionally, increased risk of bimodal milk 

ejection has been pre-associated with inadequate premilking stimulation (Sandrucci et al., 2007; 

Bruckmaier, 2013; Watters et al., 2015), although other researchers have suggested that 

stimulation by stripping before milking did not affect milk flow or yield (Wagner and Ruegg, 

2002).  

66 
 

 

As herd size increases in the U.S. dairy industry, farms are increasingly reliant on hired labor 

versus traditional owner-family labor (Baker and Chappelle, 2012; von Keyserlingk et al., 2013). 

This dynamic may impact milking efficiency; for example, Sandrucci et al. (2007) found that as 

farm size increases, less time is spent on proper teat stimulation, therefore reducing time for 

udder preparation. Recently, we completed a study that found that inadequate premilking 

stimulation and increasing herd size were the most important variables that are associated with 

increased bimodal milking (Moore-Foster et al., 2018). Thus, in order to further elucidate the 

factors related to milking efficiency, the objective of this study was to determine the herd level 

variables, including labor culture, that were associated with the duration of premilking 

stimulation in 64 Michigan dairy herds. 

 

Materials and Methods 

Dairy Farm Selection 

This  study  was  part  of  a  larger  project  in  which  124  dairy  herds  in  Florida,  Michigan,  and 

Pennsylvania  participated  in  a  15  month  trial  to  develop  an  evaluation  to  assess  mastitis  and 

antimicrobial  drug  use.  The  64  Michigan  dairy  herds  that  participated  in  the  larger  study  were 

visited  by  the  investigators  twice  between  January,  2016  and  May,  2017.  Enrolled  herds 

participated  in  the  Dairy  Herd  Improvement  (DHI)  individual  cow  somatic  cell  count  (SCC) 

option  and  had  a  herd  size  >  70  cows.    Because  the overall  objectives  of  the  umbrella  project 

included employee-related factors that might impact milk quality and antimicrobial use on dairy 

farms, organic dairies and herds that milked with automated milking systems were excluded.    

 

67 
 

 

Herd Profile and Management Culture 

During the initial herd visit, project investigators explained the study design and collected a herd 

profile to record milking times and groups, type of milking facility, housing, employee structure, 

and other general information from surveys and observation. Later, the investigators returned to 

conduct a milk quality evaluation that included 1) milking behaviors and proficiency, 2) milking 

systems,  3)  cow  environment,  4)  monitoring  and  therapy  of  infected  cows,  and  5)  farm 

management  culture.  To  capture  information  relative  to  the  5)  management  culture,  we 

interviewed  dairy  producers  and/or  managers  with  a  survey  relative  to  their  mastitis  control 

practices, attitudes and behaviors (IMPAB). Additionally, a separate human resources survey was 

administered to describe the farm management culture’s producer/manager beliefs and practices 

regarding employee communication, training and education. All survey information was collected 

following approval and within the guidelines of the Institutional Review Board of Michigan State 

University. 

Milking Dynamics and Parlor Behaviors  

While assessing milking behaviors, we evaluated milking vacuum from 3,862 cows (mean of 60 

± 29 recordings per herd) with VaDia® digital recorders (Biocontrol, Rakkestad, Norway). The 

methods and interpretation of this analysis were previously described and used to determine the 

proportion of cows within a herd with delayed (bimodal) milk ejection (Moore-Foster et al., 2018). 

Additionally, the quality of air hoses, liner alignment, and air vent patency were recorded.   

In addition to the VaDia and cluster evaluation, we observed milking protocols to determine the 

milking routine, the time interval between each preparation step and cluster attachment (lag time) 

and time spent stimulating the teats. We defined the total duration of teat stimulation (TS) as the 

68 
 

 

cumulative time spent applying tactile stimulation to the teats, exclusive of the time period between 

periods of tactile stimulation. For our analysis, we included wiping, drying, and stripping of the 

teats,  or  use  of  automated  teat  brush  equipment,  but  not  spraying  or  dipping  of  teats  with  a 

germicide  where  no  direct  tactile  contact  was  made.  A  minimum  of  four  milking  strings  were 

recorded  during  milking  preparation  in  parlors.  In  platform  or  tie  stall  milking  operations, 

stimulation and lag times were recorded for at least four cows per cluster. When there was more 

than one person milking, milking behaviors were observed for a minimum of two milking strings 

for each person.  

Parlor Ergonomics 

Measures of parlor ergonomics were estimated by determining the width of the parlor workspace, 

the height from the floor to the platform, light availability, duration of a milking shift, and the 

availability of breaks during the shift.  Lighting in the parlor was measured with a light meter, both 

in the middle of the parlor floor (mean of three separate locations) and at the level of the teat, under 

the cow (mean of six to ten locations). Additionally, for parallel, herringbone and side in/side out 

parlors, we measured the length of the parlor floor to derive an estimate of the minimum linear 

distance that each person milking cows might travel during a milking shift. For tie stall operations, 

we estimated the median linear distance traveled between cows during preparation and multiplied 

this distance by the number of cows that were milked by each employee. The formula to calculate 

this distance was previously described (Moore-Foster et al., 2018).  

 

 

 

69 
 

 

Statistical Analysis 

Dependent Variable  

The dependent variable for this study was mean TS during the premilking routine in each herd, as 

determined  by  observation  by  the  researchers.  This  data  was  entered  into  Microsoft  Excel 

(Microsoft Corp., Redmond, WA, USA) for data management and imported into SAS (ver. 9.4; 

SAS Institute, 2012) for descriptive and statistical analyses.  

Independent Variables  

The independent variables were divided into three categories 1) management culture and human 

resources, 2) employee behaviors and 3) parlor factors. Management culture and human resource 

variables  include  milking  technician  turnover  rates  (defined  as  number  of  milking  technicians 

hired  per  year  divided  by  number  of  positions  available,  reported  by  the  producer),  as  well  as 

survey  questions  that  include  frequency  of  employee  training,  manager  attitudes  about  parlor 

turnover rates and how often managers communicate with employees on personal matters. Also, 

herd compensation including pay rate, and if applicable, other benefits were recorded. Employee 

behaviors included lag time, estimated distance traveled in the parlor per shift, number of stalls 

per  milking  preparation  pass  and  parlor  flow  rates  (described  as  cows  milked  per  hour,  cows 

milked per employee per hour, and milking strings per hour). Parlor factors include rail height, 

light intensity, how many times cows were milked per day, and if 95% of liners were properly 

aligned (yes = 1, no = 0) led and if 95% of air vents on the cluster were patent (yes = 1, no = 0).  

Scale Factor Analysis 

 Initially  a  frequency  distribution  was  set  up  for  each  independent  variable  in  our  model  and 

normality was checked. We performed Principal Component Factor Analysis to determine if it was 

70 
 

 

possible  and  appropriate  to  combine  independent  variables  into  composite  scales  or  factors  to 

reflect for management attitudes and parlor flow variables, however alpha and Eigenvalues were 

not significant for any scales (Eigenvalue >1 and Chronbach’s α >0.7).   

Bivariate Analysis 

Right  skewed  continuous  level  variables  were  log-transformed  before  the  bivariate  analysis  so 

normality was attained. To decide which independent variables should be included in the model 

for  multivariate  analysis,  associations  between  TS  and  explanatory  continuous  variables  were 

investigated  using  Pearson’s  product  moment  correlation  coefficient.  For  binary  (nominal) 

variables, the dependent variables were compared to the independent variables using an Adjusted 

Wald test for the significance of the relationship. Any variables with an initial cutoff of P < 0.20 

(2-tails) were considered eligible for inclusion in the multivariable model.  

Multiple Linear Regression  

Using multiple regression and the type-III F-test, an automated backwards-stepwise elimination 

analysis was used to build the final multivariable model until only significant covariates (P < 0.05) 

were retained. To test for potential multicollinearity, interactions were also checked, however no 

significant relationships were found. Herd size was also analyzed as a confounder, the change to 

the coefficients and R2 values were minimal for the explanatory variables of  1) lag time and 2) 

number of passes during the prep procedure, However the R2 was greater than 10% for shift length. 

Thus,  herd  size  was  considered  as  a  confounder  in  the  model.  The  residual  distribution  was 

assessed visually for normality and homoscedasticity. 

 

 

71 
 

 

Results 

The mean herd size was 451 (median: 294) milking cows, ranging from 59 to 2,771 cows (Table 

1). 94% (61/64) of herds milked with automatic cluster detachers. Mean daily milk production for 

each herd was 17,848 kg [95% Confidence Interval (CI): 16,404 to 23,324 kg]. The three-month 

DHI  geometric  mean  SCC  was  136,795  cells/mL  and  62/64  herds  (97%)  reported  using  hired 

labor. These statistics are comparable to the average farm in the U.S. as reported by USDA With 

93% of large farms using ATD, the U.S. average bulk tank SCC at 194,000 cells/mL and 83% of 

surveyed herds reporting they hire non-family labor (USDA, 2014; Determining U.S. Milk Quality 

Using Bulk-Tank Somatic Cell Counts, 2015, 2016).  

Based  on  the  bivariate  analysis,  13  variables  were  significantly  related  to  TS  (Table  5)  in  the 

following categories: 1) management culture and human resources – herd size, hours of operation 

per day for the milking system, employee (or manager) shift length (includes time spent milking 

and additional chores, if shift length spanned multiple milking shifts, break length was deducted), 

break  length  for  employees  during  their  shift,  number  of  milking  units  operated  per  employee 

during  premilking  preparation,  number  of  passes  to  each  cow  in  the  preparation  routine,  2) 

employee behaviors - number of cows per employee (or manager) hour prepped in the parlor, mean 

lag time , and 3) parlor  factors - number of operating milking units, 24 h milk production, the 

number of cows milked per hour, rail height, and illumination in the milking workspace. Using 

backward stepwise regression, our preliminary multivariate regression found number of passes in 

the premilking routine and herd size (natural log) to be negatively associated with mean TS (Table 

6). Mean TS was positively associated with mean lag time of the premilking routine.  

Because of herd size being retained in the final model, we performed a two sample t-test to 

compare means for each of the eligible variables, stratified by herd size, to better understand if 

72 
 

 

exclusion of some of the eligible variables was potentially explained by association with herd 

size (Table 7). Means between large and small herds for cows milked per hour and shift length 

differed (P < 0.05). The cut-off point selected for large (≥ 300 cows) and small (< 300 cows) 

dairies was based on the criteria used for the umbrella project of the present study (Schewe et al., 

2015).  

Discussion and Conclusions 

The impetus to investigate TS as the outcome variable in this study was a previous analysis of 

from the same 64 dairy herds that determined that herd size and TS are the two key factors that are 

associated with delayed milk ejection, or bimodal milk let down (Moore-Foster et al., 2018).  This 

agreed  with  other  studies  that  described  the  importance  of  udder  preparation,  including 

forestripping and adequate lag times, in improving milking efficiency and decreasing bimodality 

(Sandrucci, et al., 2007; Bruckmaier, 2013; Moore-Foster et al., 2018). For example, in a study 

that was similar in size to ours (nearly 2,500 cows), Sandrucci et al. (2007) reported that bimodal 

milk ejection decreased from 47% for cows with no udder preparation to 30% for cows that were 

cleaned and forestripped.  

As much as 60 s of stimulation has been suggested for efficient milk ejection, but stimulation of 

at least 15 s before milking is ideal for optimal milk let down, if a lag time of at least 45 s is 

included  (Wagner  and  Ruegg,  2002;  Weiss  and  Bruckmaier,  2005;  Bruckmaier,  2013).  These 

guidelines  were  similar  (mean  TS;  14.2  s),  or  exceeded  by  (mean  lag  time;  103  s)  observed 

behaviors  in  our  study.  However,  the  duration  of  both  stimulation  and  lag  time  may  not  be  as 

critical to promote efficient milk ejection in cows that have greater udder filling at milking, such 

as cows in early lactation, as compared to cows to cows in later lactation (Kaskous and Bruckmaier, 

2007). In a Wisconsin study, stripping as part of premilking routine did not increase milk flow or 

73 
 

 

yield in cows compared to cows that were not stripped (Wagner and Ruegg, 2002). However, both 

stripped and non-stripped cows were vigorously dried for 10 to 15 s just before cluster attachment 

(Wagner and Ruegg, 2002). In our study, only 54% of herds stimulated teats > 10 s during the 

initial tactile pass (germicide application excluded). This indicates that many herds might have an 

opportunity to increase milking efficiency by increasing the proportion of stimulation earlier in the 

routine to optimize lag time intervals.  

Our results determined that as the number of passes in the premilking routine and herd size increase 

within herds, TS decreased. Conversely, as the lag time of the routine increased, TS increased. It 

seems paradoxical that fewer passes in a routine but a longer lag time are both associated with 

greater TS. However, this may suggest that herds with greater TS have a routine that results in 

more stripping and wiping for each cow and less travelling between cows. This is underscored by 

the trend (Table 7) that smaller herds averaged more TS and lag time than larger herds. Herd size 

was  determined  to  be  a  confounder  in  this  analysis.  Thus,  the  impact  on  TS  of  a  number  of 

explanatory variables that were found to be strongly correlated on bivariate anaylsis, such as shift 

length, the number of  cows milked per hour, and the the number of milking units, were likely 

masked due to herd size.   

Collectively, our analysis suggests that a milking routine culture that promotes milking efficiency 

as compared to parlor efficiency results in better TS of the teats and as found in a previous report 

(Moore-Foster et al., 2018) less bimodal milking. For example, personnel in smaller herds, who 

milk  cows  in  shorter  time  shifts,  that  are  under  less  pressure  to  finish  milking  before  the  next 

milking shift begins, or don’t need to make as many passes to each cow before cluster attachment, 

may be more willing to strip and dry teats more thoroughly. Sandrucci et al (2007) found that as 

farm size increases, less time is spent on proper teat stimulation, therefore reducing time for udder 

74 
 

 

preparation and increasing bimodality. Our study agreed with these results, mean TS in smaller 

herds (16.8 s) was nearly twice that of larger herds (8.7 s) Previously, we also found an association 

between greater herd size and a greater proportion of cows with delayed milk ejection (Moore-

Foster et al., 2018). In contrast, the USDA-NAHMS  Dairy 2014 survey cited that 86% of herds 

with ≥ 500 cows forestripped during their milking routine as opposed to 66% in herds with < 500 

cows. However, this qualitative information was self-reported by the dairy producers and TS for 

each herd of the entire milking routine was not measured. Likewise, in a survey of nearly 1,200 

herds across Canada (Belage et al., 2017), 82% and 84% of herds forestripped and dried udders 

before milking, respectively. There was no difference in these milking practices relative to herd 

size, however, the categories for herd size were < 37, 37 to 80, and > 80 cows, and the third quartile 

of herds had 80 cows. Thus, the use of hired labor was likely to be very different between the 

Canadian and our study. 

The sample size of cows within each herd depended on herd size, milking shift length, and cow 

throughput while milking. In smaller herds, (milking shifts less than four hours) we were present 

throughout the entire milking. However, especially in larger herds, it was difficult to evaluate all 

employees and all milking  groups of  cows across all shifts. Thus, in herds with short intervals 

between milking shifts, we recorded milking events in portions of two consecutive shifts to gain a 

wider  perspective  of  milking  behaviors.  Additionally,  we  decided  to  include  as  our  dependent 

variable only the actual time of tactile stimulation and not the total time of the premilking routine. 

Anecdotally, w Also, milking routines were often disrupted because of washing clusters or parlor 

platforms,  acquiring  clean  towels  from  laundry  bins,  gathering  cows  from  holding  pens, 

reattaching or adjusting units on cows throughout the parlor, and filling teat dip applicators. Thus, 

there is no assurance that our observed mean TS for each herd was consistently practiced for every 

75 
 

 

milking  string  of  cows.  Likewise,  as  previously  suggested  by  Wagner  and  Ruegg  (2002),  it  is 

difficult to know the contribution of premilking germicide application on milk ejection, as opposed 

to  true  tactile  stimulation.  However,  all  of  our  herds  applied  a  premilking  germicide  by  either 

spraying,  dipping,  or  foaming.  Thus,  due  to  the  universal  nature  of  this  procedure,  our  study 

focused on tactile stimulation. To this end, we found considerable variation (2s to 41s) between 

herds, as opposed to a higher degree of consistency with germicide application and lag times.    

 

Summary 

Smaller herd size, fewer passes to each cow in the premilking routine and greater lag times are 

associated with increasing TS in dairy herds. Variables such as greater shift length and the number 

of cows milked per hour may also impact TS but are confounded by herd size. Thus, a milking 

culture  that  emphasizes  milking  efficiency  rather  than  parlor  efficiency  may  lead  to  better 

premilking  stimulation.  With  increasing  consolidation  of  the  dairy  industry,  milking  routines 

should be evaluated to determine if adequate TS is practiced to help maintain teat and udder health.  

 

 

 

 

 

 

 

76 
 

 

Acknowledgements 

This project was supported by Agriculture and Food Research Initiative Competitive Grant no. 

2013-68004-20439 from the USDA National Institute of Food and Agriculture.  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

77 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

APPENDICES 

78 
 

 

Appendix A. Tables 

 

Table 5. Bivariate analysis results for explanatory variables of mean total stimulation time in the 
premilking routine in 64 Michigan dairy herds 

 

Variable 

Herd size (natural log 
transformation) 

P-value 

<.0001 

Number of milking units  

<.0001 

24 h milk production  (kgs) 

<.0001 

Time of milking system 
operation (natural log, h/day) 

0.0002 

Shift length (h) 

Break length for employees 
during shift (min) 
 
Number of milking units 

<.0001 

0.0142 

<.0001 

Cows milked per hour (natural 
log) 

<.0001 

Number of cows prepped/ h 
per operator 

0.0012 

Mean lag time (min) 

Mean illumination in milking 
workspace (candela) 

Rail height in parlor (natural 
log, cm) 

Number of passes in 
preparation routine 

0.0008 

0.0418 

0.1217 

0.0022 

79 
 

 

Table 6. Final linear model for associations between mean total stimulation time during the 
premilking routine and herd-level variables in 64 Michigan dairies (herd size included) 

 

Parameter 
 
Intercept 
 
Herd Size  
(natural log) 

Estimate 
 
4.43 

-0.31 

Mean lagtime (s) 

0.008 

Number of passes 

-0.30 

SE 
 
0.56 

0.11 

0.002 

0.08 

80 
 

R2 
 
0.58 
 
 

 

 

P-value 
 
<.0001 

0.0093 

0.0001 

0.0003 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

Table 7. Comparison of means for eligible independent variables for multivariable model 
between small (<300 cows) and large herds (≥ 300 cows) in 64 Michigan dairy herds 

Variable 

Mean total 
stimulation time 
(s, natural log) 

Small herds mean 
(SEM) 

Large herds mean 
(SEM) 

P-value 

2.82 (0.10) 

2.16 (0.09) 

<.05 

Mean lag time (s)  110.0 (5.7) 

96.9 (4.7) 

0.08 

Number of passes  2.70 (0.13) 

2.65 (0.13) 

0.78 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

81 
 

 

 

Appendix B. Figures 

Figure 3. Histogram of mean total stimulation time (s) for 64 Michigan dairy herds. 

y
c
n
e
u
q
e
r
F

30

25

20

15

10

5

0

 

 

 

 

 

 

 

 

 

 

 

 

 

 

≤ 5 6 to 10 11 to
15

16 to
20

21 to
25

26 to
30

31 to
35

36 to
40

41 to
45

Time (s)

 

82 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

REFERENCES 

83 
 

 

REFERENCES 

 
 
 

Baker, D. and D. Chappelle. 2012. Health status and needs of Latino dairy farmworkers in 
Vermont. Journal of Agromedicine 17:277-287. 
 
Belage, E., S. Dufour, C. Bauman, A. Jones-Bitton, and D. F. Kelton. 2017. The Canadian 
National Dairy Study 2015—Adoption of milking practices in Canadian dairy herds. J. Dairy 
Sci. 100:3839-3849. 
 
Bruckmaier, R. M. 2005. Normal and disturbed milk ejection in dairy cows. Domest. Anim. 
Endocrinol. 29:268-273. 
 
Bruckmaier, R. M. 2013. Oxytocin from the pituitary or from the syringe: Importance and 
Consequences for milking machine in dairy cows. Pages 4-11 in Proc. NMC Annual Meeting 
2013. 
 
Bruckmaier, R. M., Wellnitz, O. 2007. Induction of milk ejection and milk removal in different 
production systems. J. Anim. Sci. 86:15-20. 
Determining U.S. Milk Quality Using Bulk-Tank Somatic Cell Counts, 2015. 2016. Page 6. V. 
Services, ed. USDA-APHIS. 
 
Tančin, V., A. H. Ipema, and P. Hogewerf. 2007. Interaction of Somatic Cell Count and Quarter 
Milk Flow Patterns. J. Dairy Sci 90:2223-2228. 
 
USDA, A., VS, NAHMS. 2014. Dairy 2014 milk quality, mIlking procedures and mastitis on 
U.S. dairies, 2014. Fort Collins, CO. 
 
von Keyserlingk, M. A., N. P. Martin, E. Kebreab, K. F. Knowlton, R. J. Grant, M. Stephenson, 
C. J. Sniffen, J. P. Harner, 3rd, A. D. Wright, and S. I. Smith. 2013. Invited review: 
Sustainability of the US dairy industry. J. Dairy Sci. 96:5405-5425. 
 
Wagner, A. M. and P. L. Ruegg. 2002. The effect of manual forestripping on milking 
performance of Holstein dairy cows. J. Dairy Sci. 85:804-809. 
 

Weiss, D. and R. M. Bruckmaier. 2005. Optimization of individual prestimulation in dairy cows. 
J. Dairy Sci. 88:137-147. 

 
 

 

 

 

84 
 

 

 

 

CHAPTER 4 

OVERMILKING IN MICHIGAN DAIRY HERDS 

Herd level variables associated with overmilking in Michigan dairy herds  

 

R.  Moore-Foster*,  B.  Norby*,  R.  L.  Schewe†,  R.  Thomson‡,  P.  C.  Bartlett*,  and  R.  J. 
Erskine*1 
  
  
 
 

*Department of Large Animal Clinical Sciences  
and 
‡Department of Animal Science 
Michigan State University 
East Lansing, 48824 
 
† Department of Sociology 
Syracuse University, Syracuse, NY 13244  
 
 
 
1Corresponding Author 
R. J. Erskine 
736 Wilson Rd 
East Lansing, 48824 
517-355-9593 
Fax: 517-432-1042 
erskine@msu.edu  

 

 

 

 

85 
 

 

Abstract 

The objective of this study was to determine the herd level variables that were associated with 

overmilking in 64 Michigan dairy herds. Mean herd size was 451 cows (range 59 to 2,771 cows 

and three month DHI geometric mean SCC was 136,795 cells/mL. For each herd, surveys were 

administered to the producers to gather mastitis management practices and attitudes. Additionally, 

milking  protocols  were  observed  and  milk  flow  dynamics  were  determined  by  use  of  digital 

vacuum recorders. Milk flow dynamics were recorded for a total of 3,824 cows throughout the 

study, with a mean of 60 cows per herd (range of 11 to 154 cows per herd). Backwards multivariate 

analysis was used to determine which of 45 herd-level milking and management variables were 

associated with median duration of overmilking. Across all herds, median duration of overmilking 

was 56 s (range of 14 to 172 s) and was negatively associated with the duration of each milking 

shift (Radj

2 = 0.13). The percent of cows in each herd that were overmilked from reattachment of 

milking units after cessation of milk letdown was not correlated to the median time of overmilking. 

Given the low coefficient of determination, unaccounted variables, such as equipment function or 

manual detachment on the part of milking operators, may play an important role in overmilking in 

addition to longer time intervals between milking shifts.   

Key words: overmilking, milking protocols, employee management 

 

 

86 
 

 

Introduction 

Overmilking (OM) is defined as when milk flow to the teat cistern is less than the flow out of the 

teat  canal  (Rasmussen,  2004).  This  can  be  caused  by  improper  settings  for  automatic  cluster 

detachers (ACD) that remove clusters after a prolonged period of low milk flow, or by subjective 

manual  removal  of  clusters  by  milking  personnel.  Overmilking  results  in  longer  milking  times 

(duration of time that milking clusters are attached to the udder), and thus greater exposure of teat 

ends to high vacuum levels. In the short term, this can result in congestion within the teat wall, 

with subsequent damage and swelling of the teat canal (Mein, 2001). This also impedes efficient 

milk flow, and if the exposure to excessive vacuum from OM continues, teat-end condition can be 

compromised (Neijenhuis et al., 2000; Mein, 2001; Edwards et al., 2013). Moderate and severely 

hyperkeratotic teat ends have a higher risk of clinical mastitis than normal teats ends (Neijenhuis 

et al., 2001; Breen et al., 2009; Paduch et al., 2012). Tančin et al. (2007) reported that quarters 

with  high  SCC  (>  500  x  103  cells/mL)  had  longer  OM  phases  of  milk  flow  as  compared  with 

quarters with low SCC (< 200 x 103 cells/mL).  

Despite  knowledge  of  the  deleterious  effects  of  OM,  and  milking  machine  factors  that  may 

contribute to OM,  little is known of herd level milking behaviors and labor factors that may be 

associated  with  the  frequency  of  OM.  Additionally,  as  herd  size  increases  in  the  U.S.  dairy 

industry,  farms  are  increasingly  reliant  on  hired  labor  (Baker  and  Chappelle,  2012;  von 

Keyserlingk et al., 2013) and the impact that the labor culture may have on OM is largely unknown. 

The objective of this study was to determine the herd level variables, including the labor culture, 

that were associated with OM in 64 Michigan dairy herds. 

 

87 
 

 

Materials and Methods 

Dairy Farm Selection 

This  study  was  part  of  a  larger  project  in  which  124  dairy  herds  in  Florida,  Michigan,  and 

Pennsylvania  participated  in  an  18-month  trial  to  develop  an  evaluation  to  assess  mastitis  and 

antimicrobial drug use. Sixty-four Michigan dairy herds were visited by the investigators twice 

between January, 2016 and May, 2017. Enrolled herds participated in the Dairy Herd Improvement 

(DHI) individual cow somatic cell count (SCC) option and had a herd size ≥ 70 cows. Because 

the overall objectives of the umbrella project included employee-related factors on milk quality 

and antimicrobial use on dairy farms, organic dairies and herds that milked with automated milking 

systems were excluded.  All survey information was collected following approval and performed 

within the guidelines set by the Institutional Review Board of Michigan State University.  

Herd Profile and Management Culture 

During the initial herd visit, project investigators explained the study design and conducted a herd 

profile to record milking times and groups, type of milking facility, housing, employee structure, 

and  other  general  information.  Within  30  to  60  d,  the  investigators  returned  to  conduct  a  milk 

quality evaluation that included 1) milking behaviors and proficiency, 2) milking systems, 3) cow 

environment, 4) monitoring and therapy of infected cows, and 5) farm management culture. To 

capture information relative to the management culture, we interviewed dairy producers and/or 

managers  with  an  84  question  survey  relative  to  their  mastitis  control  practices,  attitudes  and 

behaviors (Schewe et al., 2015). Additionally, a separate 16 question human resources survey was 

administered 

to  describe  producer/manager  beliefs  and  practices  regarding  employee 

88 
 

 

communication,  training  and  education.  Approximately  90  minutes  was  needed  to  conduct  the 

surveys and review the project with each producer.  

Milk Vacuum Dynamics and Analysis 

While assessing milking behaviors, we evaluated milking vacuum from 3,824 cows (mean of 60 

± 29 recordings per herd) with VaDia® digital recorders (Biocontrol, Rakkestad, Norway). Four 

vacuum channels were employed for each individual cow evaluation by attaching 2.4 mm (internal 

diameter) silicon tubing to record vacuum on the following positions of the milking cluster: 1) rear 

quarter liner mouthpiece chamber (MPC), 2) front quarter MPC, 3) short milk tube (SMT) as an 

estimate for cluster vacuum, and 4) a short pulsation tube. Previous recordings from 40 cows had 

demonstrated the consistency between vacuum recordings in the short milk tube and actual cluster 

vacuum by insertion of a needle into the bowl of the cluster (Erskine, not reported).  

Each  cow-milking  event  was  continuously  recorded  from  the  time  that  milking  clusters  were 

attached until clusters were removed, either automatically by milk flow sensors or manually by 

employees. We employed the four phases of flow intensity: incline, plateau, decline, and OM as 

described  by  Tančin  et  al.  (2007)  for  our  intrepretive  guidelines  and  determined  the  following 

phases for our study, start of milking, start of the incline phase of milk flow, start of OM (end of 

the decline phase and near static levels in the SMT and MPC vacuum), and end of milking. All 

vacuum  recordings  were  downloaded  and  then  reviewed  using  the  VaDia  Suite  software 

(Biocontrol, Rakkestad, Norway). Previous research has suggested that milk flow is the inverse of 

vacuum levels in the cluster (Schukken, 2005). Additionally, Penry et al. (2018) determined that 

increasing MPC acts to increase teat-end congestion and reduce milk flow rates, i.e., higher MPC 

vacuum is associated with lower milk flow. 

89 
 

 

We employed the same interpretation and analysis of milk flow dynamics as previously described 

(Moore-Foster et al., 2018). For each cow, the start of the incline phase  was marked when the 

vacuum level in at least one of the two MPC channels decreased to < 13.5 kpa (4 inHg) and the 

SMT vacuum decreased from maximum and fluctuated from 3.4 to 10.2 kpa. Overmilking was 

determined to begin when MPC vacuum increased and reached a plateau above 13.5 kpa in both 

channels and the range of SMT vacuum fluctuations narrowed to < 3 kpa. The end of milking was 

determined to occur when vacuum returned to 0 kpa. (example of VaDia recording, Figure 1).  

During the vacuum analysis, the quality of air hoses, liner alignment, and air vent patency were 

also  recorded.  Additionally,  we  observed  milking  procedures  and  protocols  to  determine  the 

milking routine, the time interval between each preparation step and cluster attachment (lag time) 

and time spent stimulating the teats. A minimum of four milking strings were recorded during 

milking preparation in parlors. In platform or tie stall milking operations, stimulation and lag times 

were recorded for at least four cows per cluster. When there was more than one person milking, 

milking behaviors were observed for a minimum of two milking strings for each person.  

Parlor Ergonomics 

Measures of parlor ergonomics were estimated by determining the width of the parlor workspace, 

the height from the floor to the platform, light availability, duration of a milking shift, and the 

availability of breaks during the shift. Illumination of the milking workspace was measured with 

a light meter, both in the middle of the parlor floor (mean of three separate locations) and at the 

level of the teat, under the cow (mean of six to ten locations, depending on the number of milking 

units). Additionally, we measured the length of the milk space floor to derive an estimate of the 

minimum linear distance that each person milking cows might travel during a milking shift. The 

formula to calculate this distance was previously described (Moore-Foster et al., 2018).  

90 
 

 

Statistical Analysis 

Dependent Variable 

The dependent variable for this study was the median duration of OM for cows in each herd, as 

determined by VaDia analysis. Median OM was selected as our outcome variable rather than mean 

OM after reviewing the skewed distribution of the time period of OM for all the cows included in 

the study (Figure 2). All data was entered into Microsoft Excel (Microsoft Corp., Redmond, WA, 

USA) for data management and imported into SAS (ver. 9.4; SAS Institute, 2012) for descriptive 

and analytical statistical analyses.  

Independent Variables  

The independent variables were divided into three categories 1) management culture and human 

resources, 2) employee behaviors and 3) parlor factors. Management culture and human resource 

variables include milking operator turnover rates (defined as number of operators hired per year 

divided by number of positions available), as well as survey questions that include frequency of 

employee  training,  manager  attitudes  about  parlor  turnover  rates  and  how  often  managers 

communicate with employees on personal matters. The average pay rate, and if applicable, other 

benefits are also included. Operator behaviors included total stimulation time, lag time, distance 

traveled per shift, number of stalls per pass and the proportion of cows-milkings within each herd 

that  had  units  reattached  after  milking  was  finished  (as  determined  by  VaDia  analysis).  Parlor 

factors include rail height, illumination, how many times cows are milked per day, and if 95% of 

liners are properly aligned (yes = 1, no = 0) led and if 95% of air vents on the cluster are open (yes 

= 1, no = 0).  

 

91 
 

 

Scale Factor Analysis  

Initially  a  frequency  distribution  was  set  up  for  each  independent  variable  in  our  model  and 

normality  was  checked.  Scales  were  tested,  accounting  for  management  attitudes  and  beliefs 

regarding  labor.  However,  alpha  and  Eigenvalues  were  not  significant  (Eigenvalue  >1  and 

Chronbach’s α >0.7) for all of the scales.   

Bivariate Analysis 

Continuous independent variables that were right skewed were log-transformed before bivariate 

analysis  to  attain  normality.  To  decide  which  independent  variables  should  be  included  in  the 

model for multivariate analysis, associations between the median OM and explanatory continuous 

variables  were  investigated  using  Pearson’s  product  moment  correlation  coefficient  using  SAS 

(ver.  9.4;  SAS  Institute,  2012).  For  binary  (nominal)  variables,  the  dependent  variable  was 

compared  to  the  independent  variables  using  an  Adjusted  Wald  test  for  the  significance  of  the 

relationship. Any variables with an initial cutoff of P < 0.20 (2-tails) were considered eligible for 

inclusion in the multivariable model.  

Multiple Linear Regression  

Using multiple regression and the type-III F-test, an automated backwards-stepwise elimination 

procedure was used to build the final multivariable model until only significant covariates (P < 

0.05)  were  retained.  Interactions  were  also  analyzed,  however  no  significance  was  found.  The 

residual distribution was assessed visually for normality and homoscedasticity. 

 

 

92 
 

 

Results 

The mean herd size was 451 (median: 294) milking cows, ranging from 59 to 2,771 cows (Table 

1). 97% (62/64) of surveyed herds milked with ACD. The three-month DHI geometric mean SCC 

was 136,795  cells/mL and 62/64 herds (97%) reported using hired labor. The mean percent of 

milking  cows  digitally  evaluated  for  vacuum  dynamics  in  each  herd  was  23.5%.  The  median 

duration of OM was 47s (95% CI: 38.6s to 55.9s) with a mean of 55% (95% CI: 49.5% to 61.1%) 

of cows within each herd overmilked by at least 30 s. Median milking time for all herds was 324 

s (Table 1; 95% CI: 302, 346) and was found to be positively correlated with median OM (r = 

0.670; Figure 3).  

Based on bivariable analysis, 25 variables were significant at P < 0.20, however ten variables with 

the lowest P-values and deemed the most significant by investigators were chosen for explanatory 

variables  for  multivariable  analysis  based  on  the  following  categories  (Table  8):  1)  herd 

characteristics including number of times milking per day, kgs of milk produced daily and herd 

size 2) employee characteristics including employee turnover rates, shift length, break length, total 

distance each employee covers during a milking shift, 3) parlor characteristics including hours the 

parlor is in operation a day, if 95% of vents are clear on milking units, 4) and consistency of pre-

milking  behaviors  within  each  herd,  as  measured  by  the  standard  deviation  of  time  of  teat 

stimulation at first touch between cows.  

Using  backward  stepwise  regression,  the  final  multivariable  model  found  shift  length  to  be 

negatively associated with median OM (Table 9). Thus for every 0.04 hour (2.4 min) decrease in 

shift length decrease there is a one second increase in median OM time.   

 

93 
 

 

Discussion and Conclusions  

The impact of OM on teat end health and congestion was visually described nearly a quarter of a 

century ago (Shearn and Hillerton, 1996). In our study, median OM time across all herds was > 

45  s.  This  indicates  that  many  herds  have  an  opportunity  to  increase  milking  efficiency  and 

enhance cow health, not only by improving teat heath and reducing mastitis (Neijenhuis et al., 

2001; Breen et al., 2009; Paduch et al., 2012), but also by decreasing the duration of time for cows 

within the holding pen or milking area. In our study, MPC vacuum during the OM period was 

typically > 27 kPa, or more than twice that of what is considered desirable for teat health (Mein, 

2001). Penry et al. (2018) found that high MPC vacuum is highly correlated with lower teat canal 

cross  sectional  area,  which  in  turn  indicates  increased  teat  congestion.  Thus,  OM  can  lead  to 

congestion within the teat wall, with subsequent damage  and swelling of the teat canal  (Mein, 

2001).  During  the  massage  phase  of  the  pulsation  cycle,  the  compressive  load  of  the  liners 

facilitates venous flow and removal of interstitial fluid (Paulrud, 2005). However, d uring periods 

when the milk flow is low or absent, removal of blood and interstitial fluids may be insufficient so 

that  congestion  and  edema  occur  (Paulrud,  2005).  Thus,  our  study  suggests  that  a  high 

proportion of a dairy herds may be susceptible to deleterious effects of teat congestion as a result 

of OM. This was striking because the mean SCC in our study herds was < 150,000 cells/mL  

Our results determined that as the duration of milking shifts decrease within herds, the frequency 

of OM increases. This seems to be a paradoxical result, as observed from our data, herds with 

greater median OM periods were more likely to have longer milking times per cow. This could 

lead to slower cow flow through the parlor, longer waiting times in the holding pens, and thus 

longer shift lengths. However, the negative association between shift length and median OM time 

in our results could suggest the relationship between milking time, OM, and shift length maybe 

94 
 

 

more complex. OM may occur if automatic detachers are not properly maintained or detachment 

is  set  to  milk  cows  too dry.  78%  of  the  herds  in  this  study  stated  that  they  have  a  protocol  or 

schedule for milking equipment maintenance and 83% of the herds evaluate their entire milking 

system  at  least  once  a  year.  However,  many  routine  equipment  maintenance  programs  do  not 

include ACD service, nor did we assess detacher performance.   

From a behavioral perspective, the increased risk of OM could reflect an attitude among operators 

who work within the labor culture of shorter milking shifts; there may be less pressure to finish 

milking before the next milking shift begins.  Herds that do not emphasize parlor turnover as a 

crucial management goal may also inadvertently allow employees to have more time to observe 

cows  and  subjectively  assess  if  milk  is  still  present  in  the  udder,  and  consequently  make  the 

mistaken  decision  to  continue  milking.  Our  study  tried  included  a  survey  question  asking 

producers to score the importance of improving parlor turnover rates. However, this question did 

not reach the bivariate cut off to be included as an explanatory variable in the multivariate analysis. 

Our results were in agreement with an Irish study (O’Brien et al., 2012) that found as parlor size 

(number of units) increases, operator idle time decreases. Conversely, when ACD were not in use, 

OM increased when the number of units per operator increased, which suggests that OM can occur 

when operators are milking under pressure for greater cow flow (O’Brien et al., 2012).  However, 

except for two herds that milked in a tie-stall barn, all of our study herds had ACD. Overmilking 

can also occur from reattaching already removed clusters after cows are completely milked.  We 

found  no  association  between  the  percent  of  cows  that  were  reattached  and  median  OM  time. 

Anecdotally, some herd managers also suggested that employees, who are were all paid by the 

hour in this study, might be financially incentivized to milk cows a little longer, thus reducing the 

parlor flow and increasing the shift length and pay.   

95 
 

 

Although individual cow variables were not measured in this study, Tančin et al. (2007) reported 

that bimodal milk ejection and SCC were associated with increased duration of overmilking. We 

did not find this association between delayed milk ejection and OM in our study. This may be due 

to other cow factors that were not measured in our research. Indeed, Sandrucci et al (2007) found 

that as farm size increases, less time is spent on proper teat stimulation, therefore reducing time 

for udder preparation. In a previous study, we determined that less teat stimulation was associated 

with increased bimodal milking (Moore-Foster et al., 2018). Taken together, an intriguing dynamic 

that  suggests  as  herds  try  to  increase  parlor  throughput,  less  preparation  is  given  to  each  cow, 

resulting in more bimodal milking, but herds that “take their time” to effectively prepare udders 

before milking may be more susceptible to OM. Both of these potential problems with milking 

efficiency are often related to operator behaviors. Thus, with the increasing reliance on hired labor 

in the dairy industry and the lack of capacity on many farms to effectively train employees, more 

research and programs to train and educate employees on dairy farms may be beneficial (Bewley 

et al., 2001).  

There were several limitations in this study. We used vacuum as a qualitative, not quantitative 

measure  of  milk  flow.  Recording  flaws  (i.e.  disconnected  tubes,  failed  batteries  on  the  VaDia 

recorders) resulted in less than 5% of the recordings, but this data could not be assessed for the 

occurrence of OM and was thus excluded. The sample size of recorded cows within each herd 

varied depending on herd size, the duration of the milking shift, and the time period that we were 

present for the parlor evaluation. In smaller herds, (milking shifts less than four hours) we were 

present throughout the entire milking. However,  it was difficult to evaluate all employees, and 

milking groups for all shifts, which was especially true in larger herds. Thus, in these herds, we 

recorded milking events in portions of at least two shifts and several milking groups to serve as 

96 
 

 

indicators for the entire herd. Also, we placed multiple VaDia units at diverse milking positions in 

a  parlor  to  capture  as  much  variation  as  possible  relative  to  lag  time  and  individual 

employees/producers.  Additionally,  numerous  individual  cow  factors  such  stage  of  lactation, 

genetics, teat anatomy and parity could play a role in milk flow and the propensity to overmilk, 

which were not accounted for in our study.  

 

 

Summary 

Decreasing shift length is associated with increasing OM times in dairy herds. Thus, management 

factors  that  result  in  more  relaxed  parlor  flow  may  lead  to  greater  subjective  determination  of 

completion of milking in cows, which may decrease milking and parlor efficiency. In these herds, 

it is important to assure that detaching equipment is well maintained and milking operators are 

trained to either better determine when milking is complete or rely on ATD.   

  

 

 

 

 

 

 

97 
 

 

Acknowledgements 

This project was supported by Agriculture and Food Research Initiative Competitive Grant no. 

2013-68004-20439 from the USDA National Institute of Food and Agriculture.  

 

 

 

 

 

 

 

 

 

 

 

 

 

98 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

APPENDICES 

99 
 

 

Appendix A. Tables 

 

Table 8. Bivariate analysis results for explanatory variables of herd-level median duration of 
overmilking in 64 Michigan dairy herds  

 

 

Variable 

P-value 

Herd size (natural log 
transformation) 

0.12 

Shift length (hours)  

0.0035 

Employee break length 
(min) 

0.13 

Number of times milking 
per day 

0.03 

Milk produced in 24 hrs 
(kgs) 

Total distance employees 
traveled during a shift 
(natural log, meters) 
 
Hours milking per day 
(natural log) 

Employee turnover in 12 
months 

0.02 

0.03 

0.08 

0.19 

Are 95% of vents open on 
milking units? 

0.03 

Variability of stimulation 
at first touch  

0.16 

 
 
 
 
 

 

100 

 

 

Table 9. Final linear model for associations between median overmilking time and herd-level 
variables in 64 Michigan dairies  

Parameter 
Intercept 

Estimate 
1.78 

 
Shift length (h) 

 
-0.04 

SE 
0.08 

 
0.01 

P-value 
<.0001 

 
0.0035 

R2 
0.13 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

101 

 

 

Appendix B. Figures 

 

Figure 4. Example of digital vacuum recording demonstrating overmilking. The vertical axis 
indicates vacuum (kPa). . The horizontal axis indicates time after cluster attachment divided into 
15 s intervals. Channels 1 and 2 (white arrow) represent the rear and front mouthpiece chambers, 
and channel 3 (black arrow) the short milk tube as a proxy for cluster vacuum.  Symbols mark 
the start of milking (◊), start of overmilking (▲) and end of milking (∆). 

 

 

 

 

 

 

 

 

 

 

102 

 

 

Figure 5. Histogram of duration of overmilking (s) of individual animals in 64 Michigan dairy 
herds as recorded by digital vacuum (n=3,824 cows) 

900

800

700

600

500

400

300

200

100

0

s
w
o
C
 
f
o
 
r
e
b
m
u
N

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

≤ 10  11 to 2021 to 3031 to 4041 to 5051 to 6061 to 7071 to 8081 to 90 91 to
100

>101

Overmilking Time (sec)

 

103 

 

 

Figure 6. Correlation of median overmilking time (s) on horizontal axis vs. median unit on time 
by herd on the vertical axis (n=64) 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

)
c
e
s
(
 
e
m
T
k
l
i

i

 

 

M
n
a
d
e
M

i

500

450

400

350

300

250

200

0

50

100

150

200

Median Over Milking (s)

 

104 

 

 

 

 

 

 

 

 
 
 
 
 
 
 
 
 
 

 

 

 

 

 

 

 

 

REFERENCES 

105 

 

 

REFERENCES 

 
 
 
Ahmadzadeh, A., F. Frago, B. Shafii, J. C. Dalton, W. J. Price, and M. A. McGuire. 2009. Effect 
of clinical mastitis and other diseases on reproductive performance of Holstein cows. Anim. 
Reprod. Sci. 112:273-282. 
 
Ambord, S. and R. M. Bruckmaier. 2009. Milk flow-controlled changes of pulsation ratio and  
pulsation rate affect milking characteristics in dairy cows. J. Dairy Res. 76:272-277. 
 
Archer, S. C., F. Mc Coy, W. Wapenaar, and M. J. Green. 2013. Association of season and herd 
size with somatic cell count for cows in Irish, English, and Welsh dairy herds. The Veterinary 
Journal 196:515-521. 
 
Baker, D. and D. Chappelle. 2012. Health status and needs of Latino dairy farmworkers in 
Vermont. Journal of Agromedicine 17:277-287. 
 
Barkema, H. W., J. D. Van der Ploeg, Y. H. Schukken, T. J. G. M. Lam, G. Benedictus, and A. 
Brand. 1999. Management Style and Its Association with Bulk Milk Somatic Cell Count and 
Incidence Rate of Clinical Mastitis. J. Dairy Sci. 82:1655-1663. 
 
Barlow, J. W., L. J. White, R. N. Zadoks, and Y. H. Schukken. 2009. A mathematical model 
demonstrating indirect and overall effects of lactation therapy targeting subclinical mastitis in 
dairy herds. Prev. Vet. Med. 90:31-42. 
 
Barnouin, J., S. Bord, S. Bazin, and M. Chassagne. 2005. Dairy Management Practices 
Associated with Incidence Rate of Clinical Mastitis in Low Somatic Cell Score Herds in France. 
J. Dairy Sci. 88:3700-3709. 
 
Barnouin, J., M. Chassagne, S. Bazin, and D. Boichard. 2004. Management Practices from 
Questionnaire Surveys in Herds with Very Low Somatic Cell Score Through a National Mastitis 
Program in France. J. Dairy Sci. 87:3989-3999. 
 
Bartlett, P. C., G. Y. Miller, C. R. Anderson, and J. H. Kirk. 1990. Milk Production and Somatic 
Cell Count in Michigan Dairy Herds. J. Dairy Sci. 73:2794-2800. 
 
Belage, E., S. Dufour, C. Bauman, A. Jones-Bitton, and D. F. Kelton. 2017. The Canadian 
National Dairy Study 2015—Adoption of milking practices in Canadian dairy herds. J. Dairy 
Sci. 100:3839-3849. 
 
Bewley, J., R. W. Palmer, and D. B. Jackson-Smith. 2001. An overview of experiences of 
Wisconsin dairy farmers who modernized their operations. J. Dairy Sci. 84:717-729. 
 

106 

 

 

Breen, J. E., M. J. Green, and A. J. Bradley. 2009. Quarter and cow risk factors associated with 
the occurrence of clinical mastitis in dairy cows in the United Kingdom. J. Dairy Sci. 92:2551-
2561. 
Bruckmaier, R. M. 2005. Normal and disturbed milk ejection in dairy cows. Domest. Anim. 
Endocrinol. 29:268-273. 
 
Bruckmaier, R. M. 2013. Oxytocin from the pituitary or from the syringe: Importance and 
Consequences for milking machine in dairy cows. Pages 4-11 in Proc. NMC Annual Meeting 
2013. 
 
Bruckmaier, R. M. and J. W. Blum. 1996. Simultaneous recording of oxytocin release, milk 
ejection and milk flow during milking of dairy cows with and without prestimulation. J. Dairy 
Res. 63:201-208. 
 
Bruckmaier, R. M., Wellnitz, O. 2007. Induction of milk ejection and milk removal in different 
production systems. J. Anim. Sci. 86:15-20. 
 
Cha, E., D. Bar, J. A. Hertl, L. W. Tauer, G. Bennett, R. N. González, Y. H. Schukken, F. L. 
Welcome, and Y. T. Gröhn. 2011. The cost and management of different types of clinical 
mastitis in dairy cows estimated by dynamic programming. J. Dairy Sci. 94:4476-4487. 
 
Chassagne, M., J. Barnouin, and M. Le Guenic. 2005. Expert Assessment Study of Milking and 
Hygiene Practices Characterizing Very Low Somatic Cell Score Herds in France. J. Dairy Sci. 
88:1909-1916. 
 
Cross, J. A. 2006. Restructuring America's Dairy Farms. Geographical Review 96:1-23. 
 
DaVila, A., M. T. Mora, and R. GonzÁLez. 2011. English-Language Proficiency and 
Occupational Risk Among Hispanic Immigrant Men in the United States. Industrial Relations: A 
Journal of Economy and Society 50:263-296. 
Determining U.S. Milk Quality Using Bulk-Tank Somatic Cell Counts, 2015. 2016. Page 6. V. 
Services, ed. USDA-APHIS. 
 
Dohoo, I. R. and A. H. Meek. 1982. Somatic Cell Counts in Bovine Milk. The Canadian 
Veterinary Journal 23:119-125. 
 
Douphrate, D. I., G. R. Hagevoort, M. W. Nonnenmann, C. L. Kolstrup, S. J. Reynolds, M. 
Jakob, and M. Kinsel. 2013. The Dairy Industry: A Brief Description of Production Practices, 
Trends, and Farm Characteristics Around the World. Journal of Agromedicine 18:187-197. 
 
Dufour, S., A. Frechette, H. W. Barkema, A. Mussell, and D. T. Scholl. 2011. Effect of udder 
health management practices on herd somatic cell count. J. Dairy Sci. 94:563-579. 
 
Dustmann, C. and F. Fabbri. 2003. Language proficiency and labour market performance of 
immigrants in the UK*. The Economic Journal 113:695-717. 
 

107 

 

 

Edwards, J. P., B. O'Brien, N. Lopez-Villalobos, and J. G. Jago. 2013. Overmilking causes 
deterioration in teat-end condition of dairy cows in late lactation. J. Dairy Res. 80:344-348. 
Erskine, R. J., R. J. Eberhart, L. J. Hutchinson, and S. B. Spencer. 1987. Herd management and 
prevalence of mastitis in dairy herds with high and low somatic cell counts. J. Am. Vet. Med. 
Assoc. 190:1411-1416. 
 
Erskine, R. J., R. O. Martinez, and G. A. Contreras. 2015. Cultural lag: A new challenge for 
mastitis control on dairy farms in the United States. J. Dairy Sci 98:8240-8244. 
 
Erskine, R. J., S. Wagner, and F. J. DeGraves. 2003. Mastitis therapy and pharmacology. Vet. 
Clin. North Am. Food Anim. Pract. 19:109-138. 
 
Fogleman, S., Milligan, R., Maloney, T., Knoblauch, W. 1999. Employee Compensation and Job 
Satisfaction on Dairy Farms in the Northeast. 1999 Annual Meeting, Aug 8-11, Nashville, TN 
from American Agricultural Economics Association. 
 
Fox, L. 2013. Can Milk Somatic Cells Get too Low? A Question to be Revisited. Pages 56-63 in 
Proc. Annual National Mastitis Conference. National Mastitis Council. 
 
Fuenzalida, M. J., P. M. Fricke, and P. L. Ruegg. 2015. The association between occurrence and 
severity of subclinical and clinical mastitis on pregnancies per artificial insemination at first 
service of Holstein cows. J. Dairy Sci 98:3791-3805. 
 
Goodger, W. J., T. Farver, J. Pelletier, P. Johnson, G. DeSnayer, and J. Galland. 1993. The 
association of milking management practices with bulk tank somatic cell counts. Prev. Vet. Med. 
15:235-251. 
 
Gruet, P., Maincent, P. , Berthelot, X. , Kaltsatos, V. 2001. Bovine Mastitis and IM drug delivery 
review and perspectives. Advanced Drug Delivery Reviews:245-259. 
 
Hadley, G. L., S. B. Harsh, and C. A. Wolf. 2002. Managerial and Financial Implications of 
Major Dairy Farm Expansions in Michigan and Wisconsin. J. Dairy Sci. 85:2053-2064. 
 
Hagevoort, G. R., D. I. Douphrate, and S. J. Reynolds. 2013. A review of health and safety 
leadership and managerial practices on modern dairy farms. Journal of agromedicine 18:265-
273. 
 
Hagnestam-Nielsen, C., U. Emanuelson, B. Berglund, and E. Strandberg. 2009. Relationship 
between somatic cell count and milk yield in different stages of lactation. J. Dairy Sci. 92:3124-
3133. 
 
Halasa, T., M. Nielen, A. P. W. De Roos, R. Van Hoorne, G. de Jong, T. J. G. M. Lam, T. van 
Werven, and H. Hogeveen. 2009. Production loss due to new subclinical mastitis in Dutch dairy 
cows estimated with a test-day model. J. Dairy Sci. 92:599-606. 
 

108 

 

 

Hand, K. J., A. Godkin, and D. F. Kelton. 2012. Milk production and somatic cell counts: A 
cow-level analysis. J. Dairy Sci. 95:1358-1362. 
Harrison, J., Lloyd, Sarah and O'Kane, Trish. 2009a. Immigrant Dairy Workers in Rural 
Wisconsin - Briefing 4. in Program on Agricultural Technology Studies. M. Univeristy of 
Wisconsin, ed, University of Wisconsin, Madison. 
 
Harrison, J., Lloyd, Sarah and O'Kane, Trish. 2009b. Overview of Immigrant Workers on 
Wisconsin Dairy Farms - Briefing 1. in Program on Agricultural Technology Studies. 
 
Heikkilä, A. M., J. I. Nousiainen, and S. Pyörälä. 2012. Costs of clinical mastitis with special 
reference to premature culling. J. Dairy Sci. 95:139-150. 
 
Hogan, J. and K. L. Smith. 2012. Managing Environmental Mastitis. Vet. Clin. North Am. Food 
Anim. Pract. 28:217-224. 
 
Hogeveen, H., K. Huijps, and T. Lam. 2011. Economic aspects of mastitis: New developments. 
N. Z. Vet. J. 59:16-23. 
 
Holmes, S. M. 2011. Structural vulnerability and hierarchies of ethnicity and citizenship on the 
farm. Med. Anthropol. 30:425-449. 
 
Jackson-Smith, D., Barham, B. . 2000. Dynamics of Dairy Industry Restructuring in Wisconsin. 
Research in Rural Sociology and Development 8:115-139. 
 
Jansen, J. and T. J. Lam. 2012. The role of communication in improving udder health. Vet. Clin. 
North Am. Food Anim. Pract. 28:363-379. 
 
Jayarao, B. M., S. R. Pillai, A. A. Sawant, D. R. Wolfgang, and N. V. Hegde. 2004. Guidelines 
for monitoring bulk tank milk somatic cell and bacterial counts. J. Dairy Sci. 87:3561-3573. 
 
Jenkins, P. L., S. G. Stack, J. J. May, and G. Earle-Richardson. 2009. Growth of the Spanish-
Speaking Workforce in the Northeast Dairy Industry. Journal of Agromedicine 14:58-65. 
 
Kaneene, J. B. and A. S. Ahl. 1987. Drug Residues in Dairy Cattle Industry: Epidemiological  
Evaluation of Factors Influencing Their Occurrence. J. Dairy Sci. 70:2176-2180. 
 
Kayitsinga, J., R. L. Schewe, G. A. Contreras, and R. J. Erskine. 2017. Antimicrobial treatment  
of clinical mastitis in the eastern United States: The influence of dairy farmers' mastitis 
management and treatment behavior and attitudes. J. Dairy Sci. 
 
Kehrli, M. E., Jr. and D. E. Shuster. 1993. Factors Affecting Milk Somatic Cells and Their Role  
in Health of the Bovine Mammary Gland. J. Dairy Sci. 77:619-627. 
 
Khaitsa, M. L., T. E. Wittum, K. L. Smith, J. L. Henderson, and K. H. Hoblet. 2000. Herd 
characteristics and management practices associated with bulk-tank somatic cell counts in herds 

109 

 

 

in official Dairy Herd Improvement Association programs in Ohio. Am. J. Vet. Res. 61:1092-
1098. 
Kirkpatrick, M. A., Olson, Jerry D. 2015. Somatic Cell Counts at First Test: More Than a 
Number. Pages 53-56 in National Mastitis Council Annual Meeting 2015. NMC, Memphis, TN. 
 
Klei, L., J. Yun, A. Sapru, J. Lynch, D. Barbano, P. Sears, and D. Galton. 1998. Effects of Milk 
Somatic Cell Count on Cottage Cheese Yield and Quality. J. Dairy Sci. 81:1205-1213. 
 
Kolstrup, C. L. 2012. What factors attract and motivate dairy farm employees in their daily 
work? Work (Reading, Mass.) 41:5311-5316. 
Labor, U. S. D. o. 2005. Findings from the National Agricultural Workers Survey (NAWS) 
2001-2002. Research Report No. 9. 
 
LeGassick, J. 2013. Liners should not be the first to blame. Pages 72-73 in Progressive 
Dairyman. Progressive Dairyman. 
 
Loeffler, S. H., M. J. de Vries, and Y. H. Schukken. 1999. The Effects of Time of Disease 
Occurrence, Milk Yield, and Body Condition on Fertility of Dairy Cows. J. Dairy Sci. 82:2589-
2604. 
 
Loh, K. and S. Richardson. 2004. Foreign-born workers: trends in fatal occupational injuries, 
1996-2001. Mon. Labor Rev. 127:42-53. 
 
Losinger, W. C. 2005. Economic impacts of reduced milk production associated with an increase 
in bulk-tank somatic cell count on US dairies. J. Am. Vet. Med. Assoc. 226:1652-1658. 
 
Ma, Y., C. Ryan, D. M. Barbano, D. M. Galton, M. A. Rudan, and K. J. Boor. 2000. Effects of  
Somatic Cell Count on Quality and Shelf-Life of Pasteurized Fluid Milk1. J. Dairy Sci. 83:264-
274. 
 
MacDonald, J. and D. Newton. 2014. Milk Production Continues Shifting to Large-Scale Farms. 
Amber Waves:7-1E,2E,3E,4E,5E,6E,7E. 
 
Mein, G. A., Neijenhuis, F., Morgan, W.F., Reinemann, D.J., Hillerton, J.E, Baines, J.R., 
Ohnstad, I., Timms, L., Britt, J.S., Farnsworth, R., Cook, N., Hemling, T. 2001. Evaluation of 
bovine teat condition in commercial dairy herds: 1. Non-infectious factors. in Proc. 2nd 
International Symposium on Mastitis and Milk Quality. 
 
Mitchell, R. J., Williamson, A.M. 2000. Evaluation of an 8  hour versus a 12 hour shift roster on 
employees at a power station. Appl. Ergon. 31:83-93. 
 
Neijenhuis, F., H. W. Barkema, H. Hogeveen, and J. P. T. M. Noordhuizen. 2000. Classification 
and longitudinal examination of callused teat ends in dairy cows. J. Dairy Sci. 83:2795-2804. 
 
Neijenhuis, F., H. W. Barkema, H. Hogeveen, and J. P. T. M. Noordhuizen. 2001. Relationship 
between teat-end callosity and occurrence of clinical mastitis. J. Dairy Sci. 84:2664-2672. 

110 

 

 

 
NMC. 1999. Teat Lesions Can Lead to Milking Problems, Mastitis. NMC. 
NMC. 2012. Procedures for Evaluating Vacuum Levels and Air Flow in Mlking Systems. 2004 
Revision ed. N. M. Council, ed. National Mastisis Council, Verona, WI. 
 
NMC. 2013. Reccomended milking protocols. in www.nmconline.org. NMC, ed. NMC, Verona, 
WI. 
 
Norman, H. D., Walton, L.M., Durr, J. 2015. Somatic cell counts of milk from Dairy Herd 
Improvement herds during 2015. Council on Dairy Cattle Breeding. 
 
Paduch, J.-H., E. Mohr, and V. Krömker. 2012. The association between teat end hyperkeratosis 
and teat canal microbial load in lactating dairy cattle. Vet. Microbiol. 158:353-359. 
 
Paduch, J. H., Mohr, E., Kromker, V. 2012. The association between teat end hyperkeratosis and 
teat canal microbial load in lactating dairy cattle. Vet. Microbiol. 158:353-359. 
 
Rasmussen, M. D. 2004. Overmilking and teat condition. Pages 169-175 in National Mastitis 
Council Annual Meeting. NMC. 
 
Raubertas, R. F. and G. E. Shook. 1980. Relationship Between Lactation Measures of Somatic 
Cell Concentration and Milk Yield<sup>1</sup>. J. Dairy Sci. 65:419-425. 
 
Román-Muñiz, I. N., Van Metre, D. C., Garry, F. B., & Smith, R. A. 2007. Dairy Worker 
Training Experiences. Pages 20-22 in Proc. Fortieth Annual Conference, American Association 
of Bovine Practitioners, Vancouver, British Columbia, Canada. 
 
Royster, E. and S. Wagner. 2015. Treatment of Mastitis in Cattle. Vet. Clin. North Am. Food 
Anim. Pract. 31:17-46. 
 
Samoré, A. B., S. I. Román-Ponce, F. Vacirca, E. Frigo, F. Canavesi, A. Bagnato, and C. 
Maltecca. 2011. Bimodality and the genetics of milk flow traits in the Italian Holstein-Friesian 
breed. J. Dairy Sci. 94:4081-4089. 
 
Sandrucci, A., A. Tamburini, L. Bava, and M. Zucali. 2007. Factors Affecting Milk Flow Traits 
in Dairy Cows: Results of a Field Study. J. Dairy Sci. 90:1159-1167. 
 
Schepers, A. J., T. J. G. M. Lam, Y. H. Schukken, J. B. M. Wilmink, and W. J. A. Hanekamp. 
1997. Estimation of Variance Components for Somatic Cell Counts to Determine Thresholds for 
Uninfected Quarters. J. Dairy Sci. 80:1833-1840. 
 
Schewe, R. L., J. Kayitsinga, G. A. Contreras, C. Odom, W. A. Coats, P. Durst, E. P. Hovingh,  
R. O. Martinez, R. Mobley, S. Moore, and R. J. Erskine. 2015. Herd management and social 
variables associated with bulk tank somatic cell count in dairy herds in the eastern United States. 
J. Dairy Sci 98:7650-7665. 
 

111 

 

 

Schukken, Y. H. P., L.G., Nydam, D., Baker, D.E. 2005. Using miIlk flow curves to evaluate 
milking procedures and milk equipment. Pages 139-146 in Proc. 44th Annual National Mastitis 
Council Meeting, Orlando, FL. 
 
Seegers, H., C. Fourichon, and F. Beaudeau. 2003. Production effects related to mastitis and 
mastitis economics in dairy cattle herds. Vet. Res. 34:475-491. 
 
Smith, S. M., T. Perry, and D. Moyer. 2006. Creating a Safer Workforce. Prof. Saf. 51:20-25. 
 
Stup, R. E., J. Hyde, and L. A. Holden. 2006. Relationships between selected human resource 
management practices and dairy farm performance. J. Dairy Sci. 89:1116-1120. 
 
Swinkels, J. M., A. Hilkens, V. Zoche-Golob, V. Krömker, M. Buddiger, J. Jansen, and T. J. G. 
M. Lam. 2015. Social influences on the duration of antibiotic treatment of clinical mastitis in 
dairy cows. J. Dairy Sci 98:2369-2380. 
 
Tančin, V., A. H. Ipema, and P. Hogewerf. 2007. Interaction of Somatic Cell Count and Quarter 
Milk Flow Patterns. J. Dairy Sci 90:2223-2228. 
 
USDA, A., VS. 2013. Determining US MIlk Quality Using Bulk-tank Somatic Cell Counts. A. 
United States Department of Agriculture, VS, , ed. 
 
USDA, A., VS, CEAH. 2008. Antibiotic Use on U.S. Dairy Operations, 2002 and 2007. Page 5. 
A. USDA, VS, CEAH, ed. USDA, Fort Collins, CO. 
 
USDA, A., VS, NAHMS. 2014. Dairy 2014 milk quality, mIlking procedures and mastitis on 
U.S. dairies, 2014. Fort Collins, CO. 
 
Vaarst, M., B. Paarup-Laursen, H. Houe, C. Fossing, H.J. Andersen. 2002. Farmers' Choice of 
Medical Treatment of Mastitis in Danish Dairy Herds Based on Qualitative Research Interviews. 
J. Dairy Sci. 85:992-1001. 
 
Vila, B., G. B. Morrison, and D. J. Kenney. 2002. Improving Shift Schedule and Work-Hour 
Policies and Practices to Increase Police Officer Performance, Health, and Safety. Police 
Quarterly 5:4-24. 
 
von Keyserlingk, M. A., N. P. Martin, E. Kebreab, K. F. Knowlton, R. J. Grant, M. Stephenson, 
C. J. Sniffen, J. P. Harner, 3rd, A. D. Wright, and S. I. Smith. 2013. Invited review: 
Sustainability of the US dairy industry. J. Dairy Sci. 96:5405-5425. 
 
Wagner, A. M. and P. L. Ruegg. 2002. The effect of manual forestripping on milking 
performance of Holstein dairy cows. J. Dairy Sci. 85:804-809. 
 
Wagner, S. A. and R. J. Erskine. 2009. CHAPTER 101 - Decision Making in Mastitis Therapy. 
Pages 502-509 in Food Animal Practice (Fifth Edition). D. E. A. M. Rings, ed. W.B. Saunders, 
Saint Louis. 

112 

 

 

Watters, R. D., N. Schuring, H. N. Erb, Y. H. Schukken, and D. M. Galton. 2011. The effect of 
premilking udder preparation on Holstein cows milked 3 times daily. J. Dairy Sci. 95:1170-1176. 
 
Weiss, D. and R. M. Bruckmaier. 2005. Optimization of individual prestimulation in dairy cows. 
J. Dairy Sci. 88:137-147. 
 

Wenz, J. R., S. M. Jensen, J. E. Lombard, B. A. Wagner, and R. P. Dinsmore. 2007. Herd 
Management Practices and Their Association with Bulk Tank Somatic Cell Count on United 
States Dairy Operations. J. Dairy Sci. 90:3652-3659. 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

113 

 

 

CHAPTER 5 CONCLUSION 

The dairy industry is changing and with those changes there is an increasing reliance on non-

family labor. The growing employee force desire training and educational opportunities, 

however more importantly, they are wanting the scientific knowledge explaining why protocols 

are important. The actions of employees have a direct impact on milking outcomes including 

delayed milk ejection, stimulation time during the pre-milking routine and overmilking. This 

dissertation identified some of the employee, management and parlor variables that explained 

these outcomes.  

.................. In chapter 2 stimulation time during the pre-milking routine proved to be a significant 

independent variable effecting the herd-level prevalence of delayed milk ejection. However, little 

research has shown what employee factors affect stimulation time. Chapter 3 investigated this 

question with the multivariable model showing that stimulation time was associated with number 

of passes in the pre-milking routine. Increasing herd size being associated with decreasing 

stimulation time and because of this relationship, increased risk for delayed milk ejection. Both 

of these results suggest that larger herds, which tend to emphasize parlor efficiency may have 

less focus on ensuring an adequate parlor routine, thus driving employees to milk cows faster at 

the expense of stimulation.   

 Chapter 4 investigated what independent factors were associated with overmilking. Multivariable 

analysis showed that shorter shift lengths were associated with increased overmilking time. This 

characterizes issues that are typical of smaller herds, with shorter shift lengths, employees feel 

less ‘parlor push’ pressure and therefore have more time to subjectively decide if cows have 

completed milking. These employees may be more prone to reattaching milking units or setting 

them to manual in an attempt to harvest more milk.  

114 

 

 

In both of these scenarios, employees play a crucial role in a cow’s ability to have a quick, 

efficient, and comfortable milking experience. With the current changes in the dairy industry, 

those employees deserve opportunities for training as well as pathways for communication with 

management and especially herd veterinarians, which may retain educated, engaged employees 

on farms. However, this is only the start and further research is needed to investigate what 

management characteristics may be associated with such variables as employee turnover and 

performance independent of herd size.  

 

 

115