mt:- ltllll'llllllllllIllllllllllllllllllll

31293 017141403

This is to certify that the

dissertation entitled

WORKPLACE DISABILITY MANAGEMENT INVENTORY:
DEVELOPMENT OF A SCREENING INSTRUMENT

presented by

Brett Cornell VanTol

has been accepted towards fulfillment
of the requirements for

Ph.D. (kgmfin Counseling Psychology

 

 

ow

Major professor /

Date 51/3’98

MSU i: an Affirmative Action/Equal Opponunity Institution 0-12771

 

 

 

LIBRARY

Mlchlgan State
University

 

 

 

PLACE IN RETURN BOX
to remove this checkout from your record.
TO AVOID FINES return on or before date due.

 

MTE DUE I DATE DUE DATE DUE

 

0:752 STZJJS

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1/98 Mpg-p14

WORKPLACE DISABILITY MANAGEMENT INVENTORY:
DEVELOPMENT OF A SCREENING INSTRUMENT

By

Brett Cornell VanTol

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Counseling, Educational Psychology,
and Special Education

1998

ABSTRACT
WORKPLACE DISABILITY MANAGEMENT INVENTORY:
DEVELOPMENT OF A SCREENING INSTRUMENT
By

Brett Cornell VanTol

Work disability is understood as a complex process resulting
from the interaction of many factors, of which one important realm is
those factors related to the employer. There is a growing
recognition of the need to study work disability from a broader
perspective, however measures are needed which identify salient
constructs from each realm in a parsimonious manner. The purpose
of this study was to develop an empirically based refinement of the
research instrument of employer factors reported by Hunt, Habeck,
VanTol, & Scully (1993). The Objective was to produce a brief
survey instrument while maintaining the integrity of the theoretical
model. The original instrument contained 95 likert items which were
reorganized in an eight factor solution. In this study, a 44 item,
four factor solution was produced using factor analysis and
systematic item reduction procedures. The factors were interpreted

and labeled as follows: Safety Diligence and Hazard Prevention,

Valuing Human Resources, Reducing Work Disability and Promoting
Return to Work, and Top Management Support and Continuous
Improvement of Safety. Multiple regression reanalyses demonstrated
the comparative effectiveness of the refined instrument with the
original solution in accounting for the variance of employer work
disability outcomes. A regression model incorporating all four
elements of the model as reflected in the four factor solution
demonstrated the importance of prevention on reducing work
disability outcomes. Cluster analysis was utilized to determine
employer subgroups based on factor score profiles. Discriminant
analysis was used to confirm and interpret the clusters. Two
performance subgroups were identified with one containing employers
with high factor scores and the other with significantly lower scores.

MANOVA” demonstrated group differences between employer

,J
,/

subgroups on two summary measures of work disability outcomes
(Lost Work Day Case Rate and Lost Work Day Rate). Logistic
regression was utilized to determine if employer demographic
variables demonstrated a relationship with the performance subgroups.
Prediction rates were not high enough to demonstrate the utility of
using employer demographic variables as screening indicators to

predict disability prevention and management behaviors.

Copyright by
Brett Cornell VanTol
1998

DEDICATION
This is dedicated to my family who made possible the completion of

this work with their patience and enduring love. Thank you, Lyn,

Adelyn, & Kylee.

ACKNOWLEDGMENTS

A special thank you is due to my dissertation chairperson and
advisor, Rochelle Habeck, who has worked patiently with me
throughout my graduate studies. Her instruction and professional
standards have encouraged me to think critically and precisely, and
her support. encouraged me to complete this goal. In addition, I
wish to express appreciation to Michael Leahy, my academic advisor,
whose support and direction has greatly facilitated my professional
development. I would also like to thank the members of my
committee, William Mehrens, Ph.D., Nancy Crewe, Ph.D., Michael
Leahy, Ph.D., and Rochelle Habeck, Ph.D.

This study was supported in many ways by Allan Hunt, Ph.D.
and‘The Upjohn Institute for Employment Research. I am grateful
for the continued expertise and administrative support provided. I
am also thankful to Robert Fabiano, Ph.D., for his support and
encouragement.

Thank you to Cornelius and Trudi VanTol for giving me a
desire to learn and to Dean and Marcia Boot for their on-going
support and encouragement. I am especially grateful to my wife,
Lyn. Without her love, support, and prayers this dissertation would

not have been possible.

vi

TABLE OF CONTENTS

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

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

CHAPTER 1

INTRODUCTION ................................................................. 1
Statement and Significance of the Problem ......................... 2
Statement of the Purpose ................................................. 6
Definition of Terms ......................................................... 7

CHAPTER 2

REVIEW OF THE LITERATURE .......................................... 11
Prevention of Workplace Injuries and Illness ...................... 11
Response to Potential Workplace Disability ........................ 15
Company Culture and Management Commitment .................. 17

"Disability Prevention Among Michigan Employers, 1988-1993" 18

CHAPTER 3
METHODOLOGY ................................................................ 34
Instrument Refinement Procedures ..................................... 35
Factor Analysis ........................................................... 35
Reliability Analysis and Item Reduction ......................... 38
Testing the Refined Instrument in Regression Reanalyses 39
Exploring Patterns of Workplace Behavior ..................... 40
Post Hoc Analysis ........................................................ 42

CHAPTER 4
RESULTS ........................................................................... 44
Item Analysis .................................................................. 44
Dependent Variables .................................................... 44
Covariates .................................................................. 44
Independent Variables .................................................. 45
Structural Analysis .......................................................... 45
Extraction .................................................................. 45
Rotation ..................................................................... 48
Factor Refinement ....................................................... 49
Factor Interpretation ................................................... 52

vii

Table of Contents, Cont'd

Psychometric Properties ................................................... 65
Multiple Regression Reanalyses ........................................ 67
Classification Analysis ..................................................... 78
Post Hoc Analysis ........................................................... 85
CHAPTER 5
DISCUSSION ...................................................................... 88
Content Coverage and Reliability ...................................... 88
Comparative Associations ................................................. 92
Employer Subgroups ........................................................ 97
Implications for Further Research ..................................... 99
APPENDICES
Appendix A: Disability Prevention Among Michigan
Employers ................................................. 103
Appendix B: Eight Factors from the Hunt, Habeck,
VanTol, & Scully Study (1993) ................... 115
Appendix C: Item Means and Standard Deviations ............ 121
Appendix D: Rotated Factor Matrix ................................ 125
Appendix E: Regression Models ...................................... 129
REFERENCES .................................................................... 141

viii

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

10:

11:

12:

13:

14:

15:

LIST OF TABLES

Final Sampling Proportions and Ratios ...................
Reliability Coefficients for the Eight Factors ..........
Intercorrelation Matrix for the Eight Factor Solution
Items and Factor Loadings for Factor 1 ................
Items and Factor Loadings for Factor 2 ................
Items and Factor Loadings for Factor 3 ................
Items and Factor Loadings for Factor 4 ................

Observed Correlation, Alpha Coefficients, and
Corrected Correlations Among the Four Factors .......

Prevention Model - Summary of Selected Partial
Regression Coefficients and Model Performance ........

Disability Management Model - Summary of Selected
Partial Regression Coefficients and Model

Performance ..........................................................

Managerial Model - Summary of Selected Partial
Regression Coefficients and Model Performance .......

Summary Model - Summary of Selected Partial
Regression Coefficients ..........................................

Canonical Discriminant Function Coefficients ...........

Factor Score Group Means and Standard
Deviations for Two Clusters ..................................

MANOVA Testing for Differences Between Cluster
Means on Six Outcome Variables ............................

ix

22

29

30

54

56

58

61

65

71

72

73

74

81

83

84

LIST OF FIGURES

Figure 1: Conceptual Model of the Role of The Workplace

in Work Disability ................................................. 12
Figure 2: Scree Plot of the Eigenvalues ................................. 47
Figure 3: Plot of Distance Coefficients by Cluster Stage ......... 80

CHAPTER 1
INTRODUCTION

Work disability is impairment in work due to a work related
injury or illness. Workplace disability prevention and management is
a proactive employer-centered process of coordinating the activities
of preventing "accidents and impairments from incurring in the first
place, and an effective internal system for responding to injuries that
do occur" (Hunt, Habeck, VanTol, & Scully, 1993, p.1-13).

The problem of workplace disability remains a central concern
for business in terms of providing employees with work related
impairments the opportunity to continue to be productive and in
terms of managing the increasing costs of disability in income
maintenance, medical care, and related expenditures. In order to
prevent and manage work disability a greater understanding is needed
of the multifaceted process of work disability which incorporates
variables associated with the person and the injury or illness, as
well as the larger work environment and social system.

To guide improved efforts in the future, the proposed study
attempts to contribute to this important social problem by developing
a research tool measuring employer factors of work disability that
will facilitate an integrated understanding of the work disability
process. A brief, survey instrument corresponding with the

theoretical organization of employer influences on work disability will

be valuable to researchers who wish to study employer influences
within a more comprehensive model of work disability.
5 1 5° '6 E l E ll

Research findings from various disciplines have identified a
wide range of variables that are associated with work disability,
including: demographic characteristics (e.g., age and gender as found
in Tate, 1992), psychological traits (e.g., helplessness as found in
Walker, 1992), injury characteristics (e.g., type and severity of the
injury as found in Bigos, Spengler, Martin, Zeh, Fishers, &
Nachemson, 1986), employer policies and behaviors which impact the
process of work disability (e.g., implementing managed care for
workers' compensation injuries to control costs and retain quality
health care as found in Wiesel, Boden, & Feffer, 1994) and
community economic conditions (e.g., general inflation and wage
growth as found in Victor, Gardner, Sweeney, & Telles, 1992).

Thus, in 1988 Berkowitz described work disability as a
socioeconomic phenomenon and "a complex phenomenon that does not
lend itself to simple solutions" (p. 51). In 1991, Battie and Bigos
deduced from medical research that work disability is a
biopsychosocial process resulting from far more than physical factors.
As a result of this growing body of diverse research over the past
decade, the phenomenon of work disability is no longer considered to

be adequately studied within a single disciplinary perspective. There

appears to be a growing demand for research that is able to study
work disability from a broader perspective, incorporating the
biopsychosocial realm, as well as the work environment realm and the
policy systems realm which are now clearly influential (Fordyce,
1995').

In order to develop more complete models of work disability,
better tools are needed to identify and measure the salient constructs
from each of the realms. Comprehensive research models that seek
to understand these diverse constructs and the interactions between
them will become more feasible when the salient constructs from
theory within each realm have been identified and parsimonious
measures with adequate validity and reliability for measuring these
realms have been developed.

This study proposes to refine such a measure to more
efficiently assess the realm of the workplace and its impact on work
disability. The study by Hunt, Habeck, VanTol, and Scully (1993)
titled "Disability Management Among Michigan Employers, 1988-
1993," identified and quantified a range of workplace factors
described as employer policies and practices associated with the
incidence and outcomes of work disability. The literature review that
guided this work drew information from the fields of safety
engineering, prevention and wellness, business and human resource

management, medical and vocational rehabilitation, and psychology.

The review formed the basis of a theoretical model of work disability
from which a research instrument was developed and on which the
design for analyzing the data was based. A follow-up qualitative
study was developed which corresponded to the empirical model and
helped provide practical understanding of the processes involved,
ultimately providing further validity to the findings and their
interpretation. The results of the Hunt, et a1. (1993) study further
substantiated the impact of workplace factors and work disability, in
that employer procedures aimed at preventing and resolving work
disability were associated with lower rates of work disability.
Subsequent to this study, a number of requests have been made
of the authors by other researchers to identify the most salient and
powerful predictors (i.e., factor scores, items) of workplace factors
from the variables included in the original study instrument. These
researchers have recognized the implications of these findings and
seek to incorporate the workplace realm in their studies of work
disability. They include: The Rand Corporation, The Health
Institute of the New England Medical Center, The Workers'
Compensation Board of British Columbia, the Accident Rehabilitation
and Compensation Insurance Corporation of New Zealand, the
Workers' Compensation Research Institute, the California State
Workers' Compensation Fund, and individual dissertators from the

fields of business and nursing. These researchers are interested in

including employer factors in their models of work disability to
address a broader perspective, but are constrained from using the
entire research survey instrument due to the number of other
variables that must be included in the data collection and data
analysis. Until now, the researchers have used various intuitive
methods to select a reduced set of variables from the instrument,
without knowing how well these selected items represent the
sequence of related behaviors and policies postulated by the theory
or modeled in the analyses.
In discussing abbreviated measures, Smith and McCarthy (1995)
note:
The availability of multivariate data-analytic techniques has
perhaps encouraged investigators to study models involving
relationships among numerous variables simultaneously;
unfortunately, this has frequently led to practices such as
eliminating items from each scale in a protocol. We advise
against such practices: the psychometric properties of a
measure cannot be imputed to a short form without empirical
testing. Often, use of abbreviated measures attenuates
reliability. Even more frequently, internal consistency is
preserved but validity is attenuated because of reduced
coverage of the target construct. Systematic measure
refinement analyses should demonstrate retained content
coverage, maintained reliability, and maintained validity prior to
use of abbreviated measures. (p. 306)
What is needed to guide this body of research is a systematic,
empirical refinement of the research instrument with the objective of
producing a brief survey instrument that corresponds with the

theoretical organization developed from the literature review and

original study findings.

WW

Motivated by the need for continued research in the area of
work disability, the objective of the proposed study is to develop an
instrument which is more parsimonious and continues to correspond
with the theoretical organization of the constructs of employer
influence. The study by Hunt, et al. (1993) attempted to provide a
high degree of specificity and practical information in the areas of
safety and disability prevention for the target audience of employers
and safety policy officials. Thus, the factor solution and data
analysis focused on a more specific elaboration of the constructs
involved.

This study proposes to develop a simpler representation of the
underlying theoretical model in a refinement of the factor analysis
process and to attempt further item reduction until the most
parsimonious, yet reliable, set of items is selected to represent each
factor. The study then proposes to empirically validate the refined
instrument by testing the fit of the simpler model to the original data
and comparing its ability to account for work disability outcomes as
compared to the more specific model used by Hunt, et al. (1993).
Finally, the study also proposes to explain the utility of the final
factor solution for identifying subgroups of workplaces that have
similar profiles of disability prevention behavior, to evaluate how

these profiles are related to various disability outcomes, and whether

performance subgroups can be predicted from demographic
characteristics of the firm.

The specific research questions to be addressed in the study
include:

1. Can the reanalysis procedures produce a refined
version of the original instrument that would be more
parsimonious, yet equally effective in capturing the
hypothesized theoretical structure of the role of the work place
in preventing and managing disabilities?

2. Can the refined, brief instrument adequately capture
the variance associated with employer policies and practices
that contribute to the incidence and outcomes of work
disability?

3. Can discernable subgroups of employers be identified
on the basis of similar profiles of disability prevention and
management behavior that have implications for prediction and
screening?

Within this area of investigation, certain definitions were
important in order to maintain clarity in describing certain phenomena
and in operationalizing variables. The following terms and their
definitions are offered as they pertain to this study:

1. Work disability, as operationalized in this study, is

impairment in an employee's ability to perform work arising from a
work related injury or illness. From an organization's perspective,
work disability progresses in stages. The first stage is the
occurrence of injury and symptom incidents. It is logically deduced
that greater the frequency of exposure to injury or illness, the
greater the likelihood that the organization will have injuries that
result in lost work time. MIOSHA Recordable Rate is a measure of
the average number of injuries and symptom incidents that require
more than first aid for every 100 employees.

The second stage is an injury or illness that prevents an
employee from participating in work for one day which results in a
lost work day case (Restricted work, which results when an injured
employee can not perform his or her typically assigned job duties but
remains at work with an altered work assignment, was not measured
in this study). This represents disability incidence. Lost Work Day
Case Rate is a measure of the average number Of lost work day cases
for every 100 employees. In general, when an employee has been
absent from work for seven days due to an injury or illness, a
workers' compensation wage-loss claim is filed. Workers'
Compensation Wage—Loss Claim Rate is the average number of wage-
loss claims per 100 employees and also represents disability
incidence, but at a more severe level than lost work day cases.

The third stage is a measure of the duration of the disability.

Lost Work DastPer Case is the average number of lost work days
per lost work day case. In addition, there are two measures which
summarize an employer's experience of work disability. Lost Work
Day Rate is the average number of lost work days per 100 employees
and is a function of both the incidence of lost work day cases and
the duration of each case. Workers' Compensation Payment Rate is
the average workers' compensation medical and wage loss expenses
incurred per 100 employees.

2.- Workplace disability prevention is the policies and practices
an employer utilizes to prevent the occurrence of work disability.
This is largely comprised of traditional safety practices such as safety
training and guarding equipment to reduce exposure to hazards.
However, it also includes ergonomic initiatives and strategies to
improve employee diligence and consistency in the performance of
these policies and practices.

3. Workplace disability management is the policies and practices
an employer utilizes to respond to a work disability case. An
employer responds by assessing the parameters of the case so that an
appropriate response can be initiated. In addition, an employer
responds by providing appropriate medical care to resolve the injury
or illness. Finally, by altering the work demands, process, or
environment to accommodate or compensate for impairments in such a

manner as to eliminate or reduce the restriction from work

participation.

4. Corporate culture is the work environment that results from
human resource policies and practices an employer utilizes to involve,
cultivate, and develop its workforce. A positive work environment
that values employees is consciously developed in order to attract

and facilitate the retention of employees.

10

CHAPTER 2
REVIEW OF THE LITERATURE

This review of the literature is organized to first address the
independent variables used in the study by capturing relevant
information from the realms of employer prevention Of injury and
illness, employer response to injury and illness, and employer culture.
Then study "Disability Prevention Among Michigan Employers, 1988-
1993" (Hunt, et al. 1993) is reviewed.

The conceptual model which guided the Hunt, et. a1, (1993)
study is reproduced in Figure 1. The model conceptualizes the
employer level factors as: a) organizational and business
characteristics that are structural and demographic (i.e., size and
industry), b) cultural and managerial characteristics which reflect the
human resource philosophy of the company, c) before injury behaviors
which are interventions aimed at preventing injuries or illnesses, and
d) after injury responses to minimize the disability consequences of a
given injury. These four main areas guided the literature review for
the Hunt, et al. (1993) study and the subsequent development of the
instrument.

Safety is comprised of physical aspects such as protective
personal equipment, machine guards, and housekeeping (Chaffin,

1987; Michigan Department of Labor, 1990; Susser, 1989); as well as

11

Organizational and

Business Characteristics

 

    

    

 
  
    
 

COMPANY
ENVIRONMENT

Managerial Style

and Corporate Culture

DISABILITY .
PREVENTION 33‘? WW ‘ 3:333:28
AND e avuors p
MANAGEMENT

INTERVENTIONS ‘ I

I INJURY INCIDENCE.

 

 

 

 

DISABILITY INCIDENCE

 

 

OUTCOMES

 

V DISABILITY DURATION

 

 

I OVERALL OUTCOMES AND COSTS

 

EjguLLL. Conceptual Model

12

behavioral aspects, including motivation, knowledge acquisition, and
attitudes (McAfee & Winn, 1989; Smith, 1988; Victor, 1985;
Vojtecky, 1988). The safety literature demonstrates a long, well-
established history of the importance of safety equipment, training,
and individual accountability.

Safety programs (policies and practices) are widely and readily
available, and in many instances their presence is legislated.
Therefore, companies are distinguished from each other not by the
presence or absence of a program, but by the level of efficiency and
effectiveness achieved. Management commitment to safety training
and practice, accountability, and innovations, therefore, is an
important factor in explaining successful outcomes.

Management commitment is defined as the extent to which
managers participate in safety by attending safety trainings, wearing
safety gear, and following safety rules (Kavianian, Rao, & Sanchez,
1989; Young, 1989). Additionally, the extent to which safety is
supported by management in the verbalization of safety goals which
are considered equally with production and quality goals (Gaertner,
Newman, Perry, Fisher, & Whitehead, 1987), and in behavioral
confirmation of this stated direction via resource allocation decisions
(Sims, 1988).

Accountability authorizes employees and supervisors to

implement immediate, corrective action with the work process and

I3

with each other (Marcus, 1988). Additionally, data are gathered,
analyzed, and disseminated with performance incentives,
consequences, and corrective actions. This process is delineated both
at the incident level (accident investigation) (Jacobs & Nieburg,
1989; Webb, Redman, Wilkinson, & Sanson-Fisher, 1989), and at a
systems level (Elkiss, 1987; Gaertner et al., 1987; Gross, 1988,
Lutness, 1987; Phillips, 1989; Rappaport, 1988). Innovations in the
area of safety equipment and practice receiving recent attention have
been in the area of ergonomics (Chaffin, 1987; OSHA, 1990; Susser,
1989)

There are other programs companies use to reduce injury
incidence, injury severity, and disability which do not fall in the
traditional category of safety. These include a company’s
commitment to maintain and improve employees’ health through
health promotion and the provision of wellness resources (Galvin,
1986; Prevention Leadership Forum, 1987). Second, the use of
active Employee Assistance Programs which provide employees and
supervisors with mental health education and easily accessed
treatment resources (Stern, 1990). Third, the use of testing for
illegal substance use in job applicants and employees as a means to
prevent accidents (Prevention Leadership Forum, 1987) and with
employees when substance use is suspected of causing an accident

(Michigan Department of Labor, 1990). Finally, the use of

14

systematic screening of employees for health or disability risks
(LeClair & Mitchell, 1989; Prevention Leadership Forum, 1987), and
the systematic screening of employees in high-risk jobs for early
physical symptoms (Bigos et al., 1986; Galvin, 1986).

Disability management emerged in the 1980's in response to
rapidly rising medical costs (Tate, Habeck, & Galvin, 1986). As with
safety, management commitment is no less important for the success
and effectiveness of disability management (Schwartz, Watson, Galvin,
& Lipoff, 1989); however, disability management is relatively newer
and the techniques and methods much less tested and refined.
Management commitment to disability management is evident, in large
part, by the very presence of policies and techniques. In the original
instrument developed for the Hunt, et al. (1993) study, this is
reflected by the item content of the Management Commitment Scale
which contains nine items measuring commitment to safety, one item
for commitment to disability management, and one item covering both
areas. There is also one scale measuring Safety Accountability, but
just one item measuring disability management accountability.

Disability management must have both structure and process.
With regards to structure, it is important that disability management
have a place and an initial access point with an identified individual

who is knowledgeable of companies policies and workers'

15

compensation structure. Furthermore, it is important that the entire
disability management process from injury to resolution is coordinated
from this place (Askey, 1988; Koch, 1988; Schwartz & Beggelman,
1986)

The first step in the process of disability management is to
determine the validity of 3 workers' compensation claim (Hill &
Gipson, 1990). Once the validity has been established, it is
important to proceed with a fair and caring approach with techniques
'as simple as ensuring that workers' compensation wage replacement
checks are mailed on a timely basis (Shrey, 1988) to having the
supervisor maintain communication with the employee (Schwartz et
al., 1989). A more time consuming strategy is to evaluate and
monitor the medical services available in the community in order to
provide employees with the best quality care (Askey, 1988; Carbine
& Schwartz, 1987; Schwartz et al., 1989).

It is important from the very beginning that the process of
medical care be directed toward return to work. Thus, it is
important that companies educate employees and the external medical
and case management providers about their disability management
process (Carbine & Schwartz, 1987; Carbine, Schwartz, & Watson,
1989; Gapen, 1990; Schwartz et al., 1989). Early return to work is
the process of identifying ways to bring employees back to work in

productive roles at the earliest point in time. That may be while

16

they are still healing, but medically stable enough to perform some
meaningful task (Boschen, 1989). To accomplish this, a return-to-
work program should be based on a systematic analysis of all the
tasks within the company broken down by skills, physical demands,
and the potential for accommodations (Gice & Tompkins, 1987;
Kasdan & McElwain, 1989; McDonald, 1990; Schwartz et al., 1989).
Systematic placement, in cooperation with the medical community,
protects the healing process and is fair to every employee’s
commitment to company productivity. Accountability is also
important within the disability management and return to work
process. This has been implemented in the forms of charging wage
loss back to an employee’s department until a return-to-work
placement is found and including a department’s lost work day rate
on a supervisor’s evaluation form (Askey, 1988).
WWW

Leadership has emerged as a critical factor in recognizing,
understanding, and accepting the responsibility of changing realities
for business (Frost, 1989). Leadership is expressed both in the
formal policies and practices adopted by a company and in the
context or culture within which those policies are practices are
implemented. Therefore, it is believed that a measure of a company's
culture provides an important indicator of subsequent effectiveness of

safety and disability policies and practices (Kavianian, et al. 1989;

17

Lewin & Schecter, 1990; Schwartz, et al. 1989)

The literature revealed several important indicators of culture:
management decision making styles which are characterized by
employee participation (Kavianian et al., 1989; Lewin & Schecter,
1991; Marcus, 1988); support and commitment based on resource
allocation and upper management visibility (Gaertner, et al. 1987;
Schwartz et al., 1989); and open communication (Akabas & Gates,
1990; Lewin & Schecter, 1991). Additionally, the literature revealed
the importance of more concrete aspects of environment and culture
as reflected in cleanliness and housekeeping (Labor, 1990; Lutness,
1987), noise, air quality, light controls (Prevention Leadership
Forum, 1987), and the provision of resources to promote employee
growth through education and opportunities to learn new skills

(Rosen, 1986).
......W‘.” "L. _i 9.::°-
W The study selected seven specific

industries to include in the sample. The industries were chosen with
the following three objectives: (a) that they be among the top
MIOSHA hazard rated industries, (b) that they have substantial
employment numbers in Michigan, and (c) that they provide for some
diversity among the industries to maximize the generalizability of the

findings. The final selection included: Food Production, Standard

Industrial Classification (SIC) 20; Furniture and Fixtures, SIC 25;

18

Rubber and Miscellaneous Plastics, SIC 30; Fabricated Metals, SIC
34; Machinery, except Electrical, SIC 35; Transportation Equipment,
SIC 37; and Health Services, SIC 80. The sample includes six of the
eight most hazardous industries and Health Services (SIC 80) which
provides some diversity to the sample by contrasting with the
manufacturing environment and yet is a service industry with
significant work disability incidence. Employers with less than 100
employees were omitted from the sample due to the infrequent
occurrence of workers' compensation claims, making it difficult to
establish a reliable performance level using data that spanned four
years.

W. A cross-sectional sampling design was
utilized due to constraints of time and resources. The Michigan
Employment Security Commission, which covers establishments for
unemployment insurance purposes, provided data which identified the
industry (SIC classification), employment level, and total payroll of
establishments within the seven industries. The universe of
establishments for the second quarter of 1988 was used as the
sampling frame from which to draw a random sample. In order to
fully represent the sample, from the perspective of an establishment's
risk for experiencing work disability, a random sampling design was
developed which would sample from each industry proportional to the

expected hazard rate as compared to the other industries in the

19

sample. Thus, in industries where work disability was distributed
with greater variance, more sample points were allocated so that the
efficiency of each sample point was equalized across all of the entire
sample.

To accomplish this task, the theoretical assumption was made
that the variance of work disability would be roughly proportional to
the mean of the hazard rate chosen to represent work disability; so
that allocation of sample points according to the proportions between
the industry means for the expected hazard rate would be roughly the
same as allocating them according to the variance. The hazard rate
chosen was the average number of lost work day cases per 100
employees for that industry. The sampling ratios for each industry
were calculated so that the proportional distance between them
corresponded with the proportional distances between the hazard
rates.

A target sample size of 500 employers (establishments) was
chosen, and a target sample size for each industry was calculated.
Dividing this industry sample size by three resulted in a target
sample number for each of the three industry/size stratum chosen:
small, 100-249 employees; medium, 250 - 499 employees; and large,
500 plus employees. These target sample numbers for each of the 21
industry/size stratum were then further modified based on the

following rules: when not enough establishments were available, all

20

were chosen; an upper limit of 60 was imposed for the small firms in
SIC 34 stratum; and an arbitrary minimum sample size of 20
establishments per stratum was adopted for SIC 80. The actual
sample produced by the above decisions was 517 establishments.
Table 1 outlines the population available, the sample number selected,
and the final ratio of establishments for each industry.

W. The survey questionnaire from the

Hunt, et al. (1993) study was titled "Disability Prevention Among

Michigan Employers,’ and contained 95 items in eight scales. An
extensive literature review was conducted to identify and record
behaviors and policies that were empirically or theoretically
associated with prevention and reduction of work disability. In
addition, highly visible, common practices identified by experts and
believed to be associated with prevention and reduction of work
disability were included. The constructs that resulted were distilled
into 228 statements that were validated and prioritized by expert
reviewers.

Utilizing this information and independent sorting on the part
of the researchers and expert reviewers, the items were sorted and a
conceptual structure of eight theme areas for scale development was
identified. These eight theme areas and subsequent scales were

titled: Management Commitment, Safety Accountability, Safety

Intervention, Physical Work Environment, Disability Claims

21

domain!“ ~23 coax—E .5383“ .583.qu .8 82sz SEE: .m .3 3 32 :33qu ...—888M 208338.".—

..E oases 56.3: .m 2 E2 .ooaeaax .398 .oz :89; noise 23:2; 56.3: .82 5:3 .2 .m as ._8.§> .O .m

 

 

 

 

 

 

£8an .> .m Scam .< .m .3 .. .30 _ -wwS Ego—new sawing: meoE< 5:535 tax—585.. .88.“ Bang 33 038 2:. ”202
. . . . 22 ESE -
and m _ S SS 52 2: «3 was .992:
88 of .o 22 23 owed 80.0 $3 $3
83 5 m8 5 NAN 5 mm a
:m 8 . a 8 a. S a a 235
83 and 28 83 8o. _. 83 80.. 83
SN 8 8 R «N 2 e 2 $8 +83
5 .8 z 8 mm 2 a 2 093
Sec 33 X3 33 83 83 33 as as.
SN 3 mm 2 a. S. 2 a 8933
m: om :.. mm a. 2 _ _ 2 5282
:3 E; and $3 88 v93 38 3,3 98
wow 8m 3 E w: a a S $30:
8N om :. a 8 2. _ _ 2 =25
atom EoEascm 332m 83m
5.8m .25 E8282 was: .3 e spasm 8.55.5 RE 8cm 2 83....an
298. 8 2m 2“ 2m 2 2m 3 2m 2 2m 3 2m on 2m 2295

 

 

 

 

 

 

 

 

 

 

 

_ 033.

22

Management, Disability Intervention, Employee Risk Prevention, and
Company Environment.

Individual items were written so that the respondent rated the
employer's actual achievement on a scale from one to five (1 =
never, 0%; 2 = occasionally, 25%; 3 = sometimes, 50%; 4 = usually,
75%; 5 = always, 100%), indicating the percentage of time the policy
is implemented or the behavior achieved.

A similar process was utilized to capture and sort variables
from related research and constructs identified in the literature which
comprised organizational characteristics. These organizational and
workforce characteristics that have demonstrated an association with
work disability outcomes need to be measured and controlled for in
the data analysis. These control variables included: type of workers'
compensation insurance administration, past history of some form of
safety or disability management consultation, industry regulation
requirements, number of employees in various categories (e.g., part-
time versus full time, salaried versus hourly), average hourly wage,
workforce tenure, union representation, and employee benefits and
programs.

To model the work disability process, the outcome measures
ranged from measures of incidence (i.e., number of MIOSHA
recordable incidents) to measures of duration (i.e., total number of

lost work days, not restricted days). These self-reported performance

23

outcomes were taken from specific columns of data located on the
company's MIOSHA Log and Summary of Occupational Injuries and
Illnesses, Form 200. Specific reference to the column and location
of the data needed was included to reduce confusion and facilitate
consistency in reporting. In addition, companies were asked to
report specific workers' compensation data for 1989.

.Principles outlined by Dillman (1978) in his book, Emma
jljelephgne Suryeys; The flfgtal Design Method, were utilized to
construct the physical aspects of the instrument so as to facilitate the
ease of completion. Furthermore, the instrument was pretested with
individual experts, potential consumers of the study results, and
companies not in the sample, but drawn from the population. This
feedback was utilized to make a final edit of the instrument. A copy
of "Disability Prevention Among Michigan Employers" can be found
in Appendix A.

.The survey was typically addressed to the president or CEO,
except for the larger firms where it was addressed to the Director of
Human Resources. Again, procedures recommended by Dillman
(1978) were utilized to mail and follow-up with the survey. A
number of employers were identified as out-of-business, otherwise not
reachable, or not appropriate for the survey making the total number
of potential respondents 477. A total of 220 employers responded,

for an aggregate response rate of 46 percent.

24

To assess for response bias, it was possible to merge the
MESC data on each employer with data available from the Bureau of
Workers' Disability Compensation regarding each employer's work
compensation experience to date. Comparisons made with a probit
regression analysis of respondents and non-respondents revealed that
respondents were likely to be larger, have fewer workers'
compensation claims, and show some geographical tendency. Size
was controlled for in the data analysis and having employers with
better work disability outcomes more likely to be represented in the
sample would tend to make the research conclusions conservative.

The 95 items which comprised the eight a priori scales were
subjected to a factor analysis to validate the theoretical framework,
refine the conceptual structure of the scales, and evaluate the
potential for reducing the number of items. The factor analysis
utilized principle component analysis with placement of ones on the
diagonal. No values were imputed for missing data. A scree
analysis of the eigenvalues of the factor solution was used to
determine the optimum number of factors to retain. The 8-factor
solution was chosen for its greater specificity for the study purpose.
Oblique rotation was utilized to interpret the factor pattern loadings.
Within this 8-factor solution, items with simple structure were
analyzed first to interpret the factor. In addition, items were

retained that met an empirical cutoff and added to the reliability and

25

 

 

W;
to
in
WI
m.
Th
lIlI
Ca

fat

pa

en
0p
or;
lea

me

Val

as

interpretation of the scale, even though they also loaded highly with
other factors. Seventeen items were deleted because they did not
meet any of the criteria for inclusion.

A confirmatory factor analysis of the items within each factor
was performed using principle components analysis with a prior
commonality estimate equal to 1 and no rotation. The coefficients
from the first factor were analyzed to determine their correlation
with the factor. Eight items which were found to detract from the
overall internal consistency of their respective factor were dropped.
Thus the final factors were comprised of 73 items, which included
the repetition of three items shared between two factors (Disability
Case Monitoring and Proactive Return to Work). The resulting eight
factors were labeled and interpreted as follows in the eight
paragraphs below.

Factor one, People Oriented Culture, reflects a workplace
environment that achieves employee participation in decision making,
open communication, and the sharing of information. In smaller
organizations it may be the result of the personality of the workplace
leaders, whereas in larger corporations it more likely is expressed by
means of a formal structure.

Factor two, Active Safety Leadership, operationalizes the stated
value of placing safety on par with other workplace objectives such

as quality and production. It includes systems for tracking data so

26

that accountability can be implemented, risks identified, and resources
committed to correct hazards. The leaders personally model safe
behavior and are knowledgeable about the safety risks.

Factor three, Wellness Orientation, reflects the extent to which
a company promotes the overall health of employees through
education and the promotion of healthy lifestyle activities.

Factor four, Safety Training, assesses the timely provision of
complete safety training to all key employees, including new and
temporary employees.

Factor five, Safety Diligence, describes the extent to which
safety measures are practiced by the employees on a consistent basis.
This includes correcting unsafe behavior, maintaining safety
equipment, housekeeping, and timely investigation of all unsafe
incidents.

Factor six, Ergonomic Solutions, assesses the implementation of
solutions to minimize repetitive motion and stress and strain injuries.

Factor seven, Disability Case Monitoring, describes a process
for validating the medical needs of an employee and ensuring that
proper medical care is provided. Furthermore, it directs the focus of
medical treatment towards returning the employee back to work.

Factor eight, Proactive Return-to-Work Program,
operationalizes the extent to which the company has implemented a

systematic procedure for enabling employees to return to the

27

 

workplace at the earliest possible time, even before they are
medically released as able to resume their regular duties.

The eight revised factors corresponded with the theoretical
model of the work disability management process and retained the
conceptual structure of the eight scales. Table 2 provides the
reliability coefficients for each factor, which ranged from Safety
Training (.72) to People Oriented Culture (.96). Table 3 displays the
intercorrelation matrix of the eight factors, which ranged from the
lowest correlation between Ergonomic Solutions and Wellness
Orientation (.17) to the highest correlation between Disability Case
Monitoring and Proactive Return to Work (.72). However, recall
that these last two factors share three items. Appendix B contains
the item content of each factor. These eight factors were used as
the independent variables representing employers policies and
practices.

Wan

People Oriented Culture and Active Safety Leadership both
reflect the employer's approach to the care and management of the
employees. Safety Training, Safety Diligence, and Ergonomic
Solutions all represent specific safety behaviors. Disability Case
Monitoring and Proactive Return to Work represent the employer's
response to an employee who has been harmed or injured at work.

The multivariate analysis utilized these three groups of factors in

28

Table 2

El'l'l' L‘Efi' EIE'IE

 

 

Factor # of items Cronbach's Alpha
Active Safety Leadership 13 .88
Safety Training 4 .72
Safety Diligence 13 .89
Ergonomic Solutions 4 .87
Disability Case Monitoring 10 .93
Proactive Return to Work 14 .92
Wellness Orientation 3 .87
People Orientated Culture 12 .96

 

Note. From "Disability Prevention Among Michigan Employers,
1988-1993," by H. A. Hunt, R. V. Habeck, B. C. VanTol, and S. M.
Scully, 1993, Upjohn Institute Technical Report No. 93-004,
Kalamazoo, MI: W. E. Upjohn Institute for Employment Research.
Copyright 1993 by W. E. Upjohn Institute for Employment Research.
Reprinted with Permission.

29

E0333— EoEtnoEEm L8 33:35 SEED .m .3 .3 32 £33qu hanged EoEmoEEm
a8 25sz 5.083 .m .3 A2 .oonaESaM .3vo .oz toned 3253:. 33:2: 5383 .33
.238 .2 .m €2.32; .0 .m £88: .> .m .23: .< .m 3 .. .32-me .Eioaem 53222

368.4 cot—85:. 3:33.52 Sec 533.53 :33 “.3832
coinage .3536 of :o 3:333 o: 3565000 232

4283:. 2: 30.3 ”cote—9:60
SS 3 EN 52.. 8?: .....z 5%

 

 

ES 2. 8. 2. ON. on. 3. 3.. 3522530533 .m
£3 E. 3.. 2.. 2. mm. 2.. 3838-53; 3:895 s
33 :u. 2. 2. mm. mm. macozaoz 35 3:235 8

ES 2. Z. 2.. 3.. Seesaw £5235. .m

a: 3. 9.. 3.. maze; 53am .4

a: an. 3.. 883.5 baam .m

33 3. 35333 383 2:3. .N

83 23:6 32.25 0:8; ._

w n c n v m N _ 33am
.. . . .. . . .. . .... ... . ., ....

m 03:.

30

four separate models titled respectively: Managerial Model,
Prevention Model, Disability Management Model, and Summary
Mode].

The Managerial Model regressed two dependent variables which
represent broad, summary measures of disability outcome, namely
Lost Work Day Rate and Workers' Compensation Payment Rate for
1989 on Active Safety Leadership and People Oriented Culture.
Greater achievement of Active Safety Leadership was associated with
a IOWer Lost Work Day Rate (-1.338, p<.lO) and greater
achievement of People Oriented Culture was associated with a lower
Workers' Compensation Payment Rate for 1989 (-l.852, p<.05).

The Prevention Model included three dependent variables:
Recordable Rate, Lost Work Day Case Rate, and Workers'
Compensation Claim Rate. Each of these variables were regressed on
the independent variables of Safety Diligence, Safety Training, and
Ergonomics. This model represents the influence of traditional safety
and accident prevention on the process of work disability from injury
or illness occurrence (MIOSHA Recordable Rate) to work disability
occurrence (Lost Work Day Case Rate and Workers' Compensation
Wage-Loss Claim Rate). The results demonstrated that a greater
achievement in Safety Diligence (-l.941, p<.05) and Safety Training
(-l.652, p<.05) were associated with reduced Lost Work Day Claim

Rate. Greater achievement of Safety Diligence was also associated

31

with reduced Workers' Compensation Wage-Loss Claim Rate (-3.103,
p<.01).

The Disability Management Model regressed dependent variables
that represented the occurrence or duration of work disability (Lost
Work Day Case Rate, Workers' Compensation Claim Rate, and
Average Lost Work Days per Case) on Disability Case Monitoring,
Proactive Return-to-Work Program, and Wellness. Increased
achievement in Proactive Return-to-Work Programs was associated
with reduced Lost Work Day Case Rate (-3.235, p<.01) and reduced
Workers' Compensation Claim Rate (-1.978, p<.05).
Counterintuitively, the results demonstrated that greater achievement
of Disability Case Monitoring was associated with a higher a higher
Lost Work Day Case Rate (1.581, p<.10). This may be attributed to
the cross sectional design of the data gathering in that measures of
the independent and dependent (outcome) variables were taken at the
same time and thus may not indicate a causal relationship. For
example, this finding could be due to an increase in case monitoring
in response to an increase in lost work day cases.

A Summary Model was reported to provide an overview of the
most significant variables from the previous analyses. In the-
Summary Model, the broad, summary outcome variable of Lost Work
Day Rate was regressed on Safety Diligence, representing work

disability prevention, and Disability Case Monitoring, representing

32

response to work disability. Greater achievement of Safety Diligence
(-2.575, p<.01) and Proactive Return-to-Work Program (-2.l34,

p<.05) were associated with a lower Lost Work Day Rate.

33

CHAPTER 3
METHODOLOGY

This study proposed to reanalyze the original data to produce a
refined version of the original instrument that would be parsimonious,
yet equally effective in capturing the hypothesized theoretical
structure of the role of the work place in preventing and managing
disability. The methodology set forth below sought to accomplish
that goal and address the performance of the refined instrument
regarding its comparative ability to capture the variance associated
with employer policies and practices that contribute to the incidence
and outcomes of work disability as measured in the original study.

To further explore how the theoretical constructs of workplace
policies and practices that contribute to the incidence and outcomes
of work disability operate together in practice, this study also
attempted to identify subgroups of employers with similar self-
reported levels of behavior on the factors describing the employer

l

policies and practices. Comparative statistics of the dependent

v
_.4-.

variables were computed and examined for these subgroups; including
the workplace disability incidence and outcome measures, to explore
how the work disability process may differ for workplaces with
different profiles of behavior. A post hoc analysis examined the
extent to which employer characteristics (i.e., workplace demographic

variables) could be used to identify the subgroups of employers with

34

 

similar work place prevention and response behaviors.
WWW:

This study sought to create a refined, briefer version of the
original instrument for broader research efforts. The purpose of this
study, was to identify a simpler factor structure that corresponded to
. the theory of work disability prevention and management and to build
from it a refined instrument with adequate psychometric properties.
The conceptual model (see Figure 1 in Chapter 2) .which guided the
development of the original survey instrument delineated four main
construct areas which .were believed to be associated with an
employer's work disability outcomes. These constructs were
managerial style, corporate culture, before injury behaviors, and after
injury responses. The model also outlined the specificity level of the
dependent variables. This study sought to operationalize the
conceptual formulations set forth in this model in a simpler manner,
yet consistent with level of specificity found in the outcome
measures, so that the potential to capture an association was
maximized (DeVellis, 1991).

The sample, sampling procedure, and data collected were
retained from the original research which was reviewed in Chapter
Two.

W. First, a factor analysis using a principal-axis

factoring procedure was performed. Pedhazur and Schmelkin (1991)

35

report that principal-axis factoring is a method of identifying the
"unobserved variables" or constructs, where the extracted factors are
treated as independent variables and the items are treated as
dependent variables. [Principal-axis factoring attempts to explain the
common variance without including the unique variance associated
with each item and the error variance." .‘Thus, the requirements for
identifying the constructs are restrictedto the common item variance
producing, in effect, purer constructs as compared to those produced
from principle components analysis which includes unique and error
variance in the components which are extracted.

Factor analysis procedures included examination of the Kaiser-
Meyer-Olkin measure of sampling adequacy, the Bartlett Test of
Sphericity, and the anti-image correlation to assess the extent to
which the factor analysis assumptions were met. Principle-axis
factoring was used to extract the factors. A scree plot was produced
and analyzed for determining the best factor solution with a minimum
cutoff of eigenvalues greater than one. The theoretical model
allowed for the likelihood that correlations between the factors
exists. Therefore, both orthogonal and oblique rotations were used
to identify the factors. If on the basis of the oblique rotation, "it is
concluded that the correlations among the factors are negligible, the

interpretation of the simpler orthogonal solution becomes tenable"

(Pedhazur & Schmelkin, 1991, p. 615).

36

Next, because the goal of this study was to identify a simple
factor structure to best explain the data, an principle components
analysis was also performed with the resulting solution compared to
the principle-axis solution reported above in order to assess the
extent to which the objective of reduction had been obtained.
Principal components analysis is most appropriately used as a data
reduction method (Pedhazur & Schmelkin, 1991). It extracts error
and item unique variance, as well as common variance, and therefore,
assigns the maximum amount of variance to the components
extracted. Thus, it has the potential for producing a simple structure
which accounts for the greatest amount of variance. As a rule of
thumb, principal components analysis should extract over 50 percent
of the variance with the first two or three components (Pedhazur &
Schmelkin, 1991). Including the total variance, however, does not
allow for a meaningful interpretation of the extracted components.
Thus, the solution identified by the principal components analysis was
used as a standard of efficiency and effectiveness with which the
principal-axis factor solution was compared. If the principal
components analysis extracted components which differed from the
principal-axis extracted factors, the factors extracted from the
principal-axis factoring solution would have been used to develop
the refined instrument because correspondence with the theory and

interpretability of the factors were of greater priority than reduction

37

alone.

Simple structure was identified with both orthogonal rotation
using the Varimax procedure and oblique rotation using the Oblimin
procedure. Both rotations were compared for the degree of simple
structure achieved, if orthogonal rotation achieved comparable simple
structure it would be chosen for the simplicity it affords and the
creation of distinct factors that result.

WWW. Constructs exist at
varying levels of hierarchy or aggregation and understanding their
dimensionality is important in instrument development. Broad or
aggregate constructs cannot be analyzed on an aggregate level
without understanding how the multidimensional facets correlate and
interact with each other (Smith & McCarthy, 1975). In this study,
the objective of parsimony while maintaining theoretical
representation was facilitated, not only in assessing for a simpler
factor solution, but also by attempting to distill the measure of each
factor to the fewest prototypical items reflective of that factor
(Smith & McCarthy, 1995, p.305) while maintaining a target level of
reliability. Thus, to achieve parsimony the following rules were
adopted to facilitate a systematic retention of the items effected to
the greatest extent by the latent variable represented by the factor on
which it loaded the highest, and having minimal effects from the

other latent variables comprising the process of work disability

38

 

prex

fact
wou
noti
to d
rem;
An :
thro
85

alph
final

factc

prevention and management.

To remain in the scale the item had to: a) achieve an initial
factor loading of .40, b) needed to demonstrate that it's removal
would lower alpha, c) needed to have an item-total correlation not
noticeably less than the other item-total correlations, and d) needed
to demonstrate a level of communality when regressed on the
remaining items which was not noticeably less than the other items.
An iterative process was undertaken until each of the three rules, b)
through d), no longer applied. Cronbach's Alpha was targeted for
.85 in order to provide room for the achievement of an acceptable
alpha level with the application of the scales on another sample. The
final factor solution was interpreted based on simple structure in the
factor matrix.

W In
order to assess how well the refined factors explained the outcome
data, the multiple regressions from the prior study were repeated.
The factor scores developed from this study were used in place of
the original factor scores where they corresponded to the appropriate
regression model. The multiple regression analyses included the
dependent variables measured in log transformation. The log
transformation was chosen as the distributions of the variables in log
form demonstrated greater regularity and symmetry. Furthermore, the

impact of outliers on the parameter estimates were minimized. The

39

independent variables were measured in Z-score standardized form.
The covariates had various measurement properties; either
dichotomous, categorical, or continuous (these were also measured in
a log transformation to reduce the influence of outliers). The same
statistical procedures were used in the regression models, including
simultaneous entering of all the variables. Previous research
supported the a priori prediction of the direction of influence of most
variables on the outcome measures, therefore the regression results
were subjected to a one-tailed test of significance.

WW Next, a cluster
analysis was used to determine if similar subgroups of employers
could identified based on their behaviors (policies and practices) to
prevent and respond to work disability. To accomplish this, an
agglomerative hierarchical cluster analysis was performed using the
SPSS Hierarchical Cluster Analysis procedure. The variables used to
cluster the cases were the factor scores from the refined factor
solution identified in this study.

The squared Euclidean distance measure was utilized to measure
the distance between variables. The squared Euclidean distance is
the most commonly used distance measure (Norusis, 1994). The
method used for combining clusters was Ward's Method, as this uses
information from all the variables and calculates clusters with the

property of having the smallest squared within-cluster distances.

40

Ward's Method is useful when the variables are measured with the
same unit, as is the case with this study (Sneath & Sokal, 1973, as
cited in Norusis, 1994).

Determining the optimal number of clusters to retain involves a
compromise between parsimony and homogeneity. Distance
coefficients were analyzed to assist in determining the number of
clusters to retain. Small coefficients indicate that fairly
homogeneous clusters are being combined. A large increase in the
coefficient between two adjacent steps in the agglomeration schedule
indicates a point at which the clusters contain relatively dissimilar
members (Norusis, 1994).

Cluster analysis can produce statistical structure in a sample
where no theoretically logical structure exists. Therefore, to test the
structure produced by the cluster analysis, the group membership
from the clusters was used in a Discriminant Analysis to assess the
extent to which the variables used to determine the clusters could be
combined as independent variables to predict group membership. The
SPSS Discriminant Analysis procedure was used to assess the extent
to which the factors as independent variables correctly predicted
group membership. The proportion of correct classification of
employers into the previously assigned subgroups (clusters) was
computed and examined. The discriminant analysis also provided a

discriminant function which was used to provide an interpretation of

41

the clusters that were developed with the cluster analysis.

The resulting clusters theoretically provided discrete groups of
employers found to be statistically similar based on their self-
reported performance in accomplishing prevention and response
policies and procedures associated with a reduction in work disability
outcomes. To test the practical meaning of the performance
subgroups, the next step was to describe these cluster groups in
terms of the disability outcome measures. A MANOVA procedure
was used to assess whether there was a statistically significant
difference between the performance subgroups on the disability
outcome measures. If patterns of work disability outcomes are
discernable among the performance subgroups, further validity may be
attributed to the refined'instrument and additional understanding
gained about the ways these employer policies and practices function
in relation to the progression of disability outcomes.
W

In a post hoc analysis, an attempt was made to classify the
performance subgroups with the covariate variables. In the Hunt, et
al. (1993) study, covariates which demonstrated significant partial
regression coefficients included: average workers' tenure, industry
establishments that are a part of multiple plant firms, size,
unionization, wages, and type of insurance. A logistic regression was

conducted to determine the extent to which the covariates could

42

 

classify the performance subgroups. Comparatively accurate
classification rates would substantiate the screening value of these

demographic variables.

43

 

CHAPTER 4
RESULTS
Item Analysis

W; In the Hunt, et. al. (1993) study the
dependent variables were first transformed into rates so that the
variables would be in a form that could be compared across different
size employers. The measures were then reviewed for normality and
a natural log transformation was much preferred for regularity and
symmetry, and it minimized the influence of outliers. The log
transformation used was log(1 + r) where r represented the rate of
the dependent variable. This eliminated the log of 0, a legitimate
value, which is undefined and it made the regression coefficients of
the dependent variables interpretable as a percentage change
associated with a one unit change in an independent variable.

Covariates. The covariates included both dichotomous and
continuous variables. For the dichotomous variables, one category
was dropped from the regression equations to avoid
overdetermination. The mean of a dichotomous variable is the
proportion of the sample possessing the particular characteristic. The
continuous variables were measured in natural log transformation to
reduce the influence of outliers and to facilitate their reporting in the
regression equations such that a one unit change in the covariate

would have a percentage change on the dependent variable.

44

 

 

mm. The first step was to examine the

response distributions of the individual items used in the survey
instrument. This was done to identify items that had highly skewed
and unbalanced distributions and limited variability, although no items
were eliminated at this point.

The items were measured on a five point Likert rating format.
The item means and standard deviations are listed in Appendix C.
There were two items with means less than two, 11 items with means
between two and three, 48 items had means between three and four,
and 34 items had means greater than four. Thus, in general, the
instrument contains mostly items that describe policies and behaviors
which were reported by the respondents as being achieved more than
half the time.

Structural Analysis

The first research question of this study asked if a refined
version of the original instrument could be produced which captured
the hypothesized theoretical structure of the role of the work place
in preventing and managing work disability in a more parsimonious
manner, and thus be less of a burden in future research. Exploratory
factor analysis was utilized to address this question.

Extraction. Exploratory factor analysis was utilized to assess
the structure within the 95 items of the survey with the purpose of

identifying the presence of a simpler structural solution than the

45

eight-factor solution attempted by Hunt, et al. (1993). The
theoretical model that was set forth in the development of the study,
which had become apparent in the pilot study (Habeck, et al. 1991)
was based on four main constructs considered to be important in the
reduction of work disability (see Figure 1 in Chapter 1). Thus, the
aim was to validate the theoretical model in the factor analysis and
assess the adequacy of the refined factor solution for defining these
constructs within the item pool.

The Bartlett Test of Sphericity indicated that the correlation
matrix was not an identity matrix. The Kaiser-Meyer-Olkin measure
of sampling adequacy and inspection of the anti-image correlation
matrix indicated that the linear effects of the other variables in the
matrix explained much of the correlations between pairs of variables
and thus the matrix met the assumptions for reduction by factor
analysis.

A factor analysis was conducted using principal-axis factor
procedure which initially replaces the diagonals of the correlation
matrix with squared multiple correlation coefficient estimates of the
communalities and iterates through the process of extracting factors.
Eigenvalues were analyzed and a scree plot was produced to assess
the four factor solution. The scree plot is shown in Figure 2. There
are 23 factors with eigenvalues greater than one.

Visual inspection of the "elbow" shows a few potential cutoff

46

points for the number of factors to retain. Retaining four factors is
the first potential cutoff point and is also the point at which the
difference in variance accounted for by each subsequent factor most
noticeably levels off. Thus, from using the scree plot, it appears
that retaining four factors may provide a latent variable solution that
represents the variance in the correlation matrix in a simpler

solution.

 

Factor S cree Plot

 

 

 

 

 

 

 

z
0
mi
5.
m.
! 5. \I
50 , , ,vv ' vv"7:::27-x:;::‘7:::w7““7““7vittt“:::=‘--:::::::::
‘lOfl‘lOZlZSS‘QflSUQTIDUQO‘
FmMntar

 

 

Elm. Scree Plot of Eigenvalues.

The first unrotated factor accounts for 23.9 percent of the
variance, the second for 7.0 percent, the third for 4.5 percent, and
the fourth for 3.4 percent which means that the first four factors
cumulatively account for 38.8 percent of the variance in the

correlation matrix. Initial communality estimates in the factor

47

procedure ranged from .60 to .91, thus the correlation matrix
contained highly related items, showing good potential for reduction
by factor analysis.

Rotation. Following extraction, the retained factors were
rotated to achieve simple structure and enhance interpretability
(Thurstone, 1947 and Cattell, 1978 as referenced in Kline, 1994).
Orthogonal and oblique rotations were compared for simple structure.
Varimax procedure for orthogonal rotation produced simple structure
which was very comparable to that achieved with oblique rotation.
Thus the orthogonal rotation was chosen for the simplicity of
developing latent factors which are more distinct from one another
(Kline, 1994).

Guadagnoli & Velicer (1988) reported that when components
contained four items with factor loadings in the range of .60+, they
were stable with sample sizes of 150. These four resulting factors
approach that criterion and thus, with a sample size of 200, it is
believed that an appropriate level of stability has been established.
The rotated factor matrix is provided in Appendix D. Factor 1
contains 29 items, Factor 2 contains 22 items, Factor 3 contains 25
items, and Factor 4 contains 19 items. The rotated factors account
for 36.3 percent of the variance in the correlation matrix.

The four factor solution was then compared with extraction

using the Principal Components procedure, which places ones on the

48

diagonal of the correlation matrix and thus has the potential of
accounting for greater variance with fewer factors. Extraction and
Varimax rotation of four components resulted in components which
were almost exactly similar to the four factors extracted with
Principal-axis factoring procedure. These four rotated components
accounted for 38.8 percent of the variance. The solution using
Principal-axis factoring was retained as this procedure is more
appropriately used for the purpose of identifying latent variables
corresponding with theory. The Principal Component procedure
provided confirmation that the four factors extracted were accounting
for a significant amount of the potential common variance.

antoLRofinomonL Conservative rules were followed to
achieve parsimony. All items loading less than .40 on each factor
were eliminated, resulting in the deletion of six items from Factor 1
(DCM13, DCMOS, MCll, SA03, DCMOl, and SA02); five items from
Factor 2 (ERP09, ERP02, ERPlO, CE09, and ERP04); two items from
Factor 3 (D101 and DCM03); and six items from Factor 4 (ERPOI,
ERPOS, SA08, MCO3, CE08, and $101). Seventy-six items were
retained.

Next, to further promote parsimony and construction of scales
which clearly distinguished between factors, items were deleted when
they loaded greater than .30 on another construct. This eliminated

five items from Factor 1 (8112, S108, 8103, MClO, and 8110); three

49

items from Factor 2 (CE07, CE06, and CEOS); five items from Factor
3 (D102, D109, D110, D104, and D113), and seven items from Factor
4 (SA04, MC07, SAIO, MC06, MC08, S102, and MCOl). Thus, 56
items were retained.

The next level of refinement was done within each factor using
internal consistency reliability procedures. The Cronbach's alpha for
Factors 1, 2, and 3 was .89 and for Factor 4 was .79, thus there was
believed to be some room for greater parsimony while attempting to
achieve a final alpha of greater than .85. The "Alpha If Item
Deleted" statistic was assessed to determine if any items were
. detracting from internal consistency. One item was deleted in Factor
3 (D107). Next, "Corrected Item-Total Correlation" and "Squared
Multiple Correlation" statistics were assessed, respectively, to
identify items that correlated noticeably less than the others with the
factor and items with noticeably less shared communality. No items
were deleted based on the item-total correlation, but two items from
Factor 1 (SI09 and $111) and one item each from Factors 2 (SA01)
and 4 (SA05) were deleted based on noticeably lower shared
communality.

Finally, the correlation matrix of the items was assessed for
pairs of items with noticeably high correlations. These items were
then judged to assess whether their content overlapped. In Factor 1,

items PWE08 and PWE09 correlated .69 and items PWE07 and

50

PWE08 correlated .68. The content of these items was judged to
overlap and PWE08 was retained as it correlated the highest with the
factor and had the highest shared communality. Items PWE07 and
PWE09 were deleted. 8106 and 8104 also correlated at the .68 level,
but their content was not judged to overlap. Within Factor 2, items
ERP06 and ERP07 correlated .88 with each other and their content
was judged to overlap. ERP06 was deleted as its item-total
correlation and shared communality was the lowest. In Factor 3,
items DCM07 and DCM06 correlated at .74 and their item content
was judged to overlap. Item DCM06 was deleted based on its lower
item-total correlation and communality. DCMll and DCM12
correlated at .63, but their content was not judged to overlap. Thus
an additional nine items were deleted from the factors.

The Reliability procedure was run again on the remaining items
to assess Cronbach's alpha and "Alpha if Item Deleted”, "Corrected
Item-Total Correlation", and "Squared Multiple Correlation" statistics.
Two items, one each in Factors 1 (PWElO) and 2 (ERP08), detracted
from alpha and were deleted. This procedure was repeated a third
time and one item from Factor 2 (ERP07), which now showed
noticeably lower communality, was deleted. The final Reliability
procedure demonstrated completion of the decision making rules used
above to refine each factor. Forty-four total items were retained.

Cronbach's alpha for Factor 1 (n = 13) was .87; for Factor 2 (n =

5]

10), .89; for Factor 3 (n = 16), .89; and for Factor 4 (n = 5), .78.

W. The four factors as refined and used in
the subsequent data analysis are reported in Tables 4 - 7. The
Tables list the factor loadings for each item and its original scale
membership on the questionnaire as indicated by the item number
(i.e., MC = Management Commitment, SA = Safety Accountability, S1
= Safety Intervention, PWE = Physical Work Environment, DCM =
Disability Claims Management, D1 = Disability Intervention, ERP =
Employee Risk Prevention, and CE = Company Environment). The
items contained in each factor'are presented in tabled format.
Interpretation of the factors is presented in narrative format in the
following paragraphs.

Factor 1 is titled "Safety Diligence and Hazard Prevention."
The items retained in this factor appear to reflect a high level of
commitment, attention, and involvement in preventing risks to safety.
They reflect the role that supervisors play in the practice of safety
and the conscientious and meticulous application of safety practices
throughout the company. Knowledge provided through safety training
is specifically targeted in regard to specific audiences within the
company, specific hazards, specific jobs, and specific safe work
practices. Additionally, this factor reflects the use of hazard
prevention techniques both in terms of purchasing or modifying

equipment and work procedures to reduce hazards and risks for

52

injury, and techniques incorporated into employee behavior in the
form of using personal protective equipment where indicated.

This factor, then, is best understood not as a simple
representation of the breadth of safe behavior but as a measure of
the quality to which targeted safety practices are diligently and
consistently achieved. For example, the item regarding safety
training for employees, a more common behavior, dropped out; but
the items regarding specific training for new employees, temporary
employees, and supervisors were retained because they represent a
higher level of diligence and attention to detail. This diligent and
conscientious attention to detail is noted in the other items, too.
For example, "Supervisors document W accidents and
violations for review and consideration" and "The company achieves
oxoeflom, housekeeping" (underlined for emphasis).

Factor 2 is titled, "Valuing Human Resources." This factor
represents an effort on the part of management to cultivate and use
employee input in meaningful and important ways in company decision
making. This factor reflects the extent to which the company
facilitates employee involvement in various decision making
processes, such as aspects of the immediate work process as well as
long term planning; the extent and directions in which information is
communicated throughout the company; the achievement of a

collaborative climate in which employees are empowered to raise

53

Table 4

 

 

finalization

Factor

Loading Item# Item

.66 $107 Supervisors confront and correct unsafe behavior
and hazards when they occur.

.60 SA07 Supervisors document even minor accidents and
violations for review and consideration.

.57 PWE02 Equipment is well maintained.

.57 8106 Supervisors are informed about possible hazards and
trained in safe work practices for the jobs they
supervise.

.56 PWEO6 Existing equipment and tools at this plant have been
modified to minimize safety hazards.

.54 SA06 Supervisors complete accident records promptly.

.54 PWE08 Strategies are used to reduced repetitive
movements.

.53 S105 Temporary or temporarily assigned employees are

given training on-site before being placed on a job

or working with new equipment.

54

Table 4 (cont'd)

 

Factor
Loading Item # Item

 

.53 PWE03 Workers use personal protective equipment where
indicated.

.51 PWE04 Safety guards and equipment are used in hazardous
operations.

.5] PWEOS Safety and health issues are considered in the

acquisition of new machinery, equipment and tools.
.49 PWEOI The company achieves excellent housekeeping.
.46 8104 New and transferred employees are given training
regarding specific hazards for their particular job

before being placed on the job.

 

55

Table 5

 

 

Factor

Loading Item # Item

.67 CE13 Management seeks and considers employee input in
company decisions.

.66 CE15 Workers have some control over the work process
and productivity demands.

.62 CEIO Employees are formally included in the company's
goal setting and planning process.

.62 CE12 The company shares information with employees
about the financial status and productivity needs of
the company.

.61 CEll The company achieves open communications where
employees feel free to raise issues and concerns, or
make suggestions.

.60 CE03 Job satisfaction among employees at this company is
high.

.60 CE04 Working relationships are collaborative and

cooperative in this company.

56

Table 5 (cont'd)

 

Factor
Loading Item # Item

 

.56 CE14 Employee involvement programs, such as quality
circles and labor-management participation teams,
are used to generate employee participation in
company operations.

.51 CEO] Ownership and accountability are pushed to the
lowest levels of the organization.

.50 CE02 The company demonstrates concern about retaining
and developing personnel through its human

resources policies and programs.

 

57

Table 6

 

Factor
Loading

Item #

Item

 

.70

.70

.68

.68

.66

.62

DCM07

DCM12

DCMO4

DCMO9

D106

DCMll

Duration of disability is evaluated to identify claims
needing case management and rehabilitation services.
Claim management is well coordinated from initial
injury to claim resolution.

The company monitors employees off work due to
disability and their projected return-to-work date.
When the company refers for professional case
management or rehabilitation services, they still
maintain contact with the employee and monitor the
return-to-work process.

The company maintains regular communication with
the injured employee's attending physician.
Responsibility for disability claim management and
return-to-work coordination is assigned to a specific

individual in the company.

58

Table 6 (cont'd)

 

Factor

Loading Item #

Item

 

.58

.57

.56

.53

.51

.50

D103

D112

D105

D114

D108

D111

Injured employees are contacted by a designated
person within the company immediately following
medical treatment.

Follow-up contact is made with the employee and
supervisor after successful return-to-work to deal
with any needed adjustments.

The treating physician is asked to identify worker
restrictions and capacities as well as a target date
for return-to-work.

There is cooperation and coordination among
departments in efforts to return injured employees
to work.

The company develops alternative placement options
and modified job duties to return disabled
employees to work.

When an injured worker is unable to resume prior
duties the company provides job retraining for

reassignment in a productive capacity.

59

Table 6 (cont'd)

 

Factor

Loading Item #

Item

 

.48

.47

.46

.42

DCM08

DCMIO

DCM02

ERPO3

Rehabilitation professionals are used to evaluate
work capacity and develop individualized
rehabilitation plans when injured workers are unable
to resume employment.

The company conducts audits to evaluate the quality
and effectiveness of medical and rehabilitation care
provided to its injured employees.

Disability claims are evaluated early and accurately
to determine their validity.

Employees are encouraged to promptly report

physical symptoms arising from job tasks.

6O

I
'll 1.9 . 9 90.9.... 9 e 9 ‘ .99M!.!°."'l 999

 

Factor
Loading Item# Item

 

.66 MC04 Management allocates staff time of specific
individual(s) for safety responsibilities.
.63 MC09 The company strives for continuous improvement in

safety performance.

.56 MCOS The safety manager receives support from top
management.
.55 MC02 Top management supports the safety program by

attending safety meetings and training sessions.
.44 SA09 The company identifies specific jobs and
departments with high accident incidence and lost

work time.

 

61

issues and concerns; and the value of employees as demonstrated by
support for personnel development.

Factor 3 is titled, "Reducing Work Disability and Promoting
Return to Work." This factor is comprised of policies and practices
employers utilize to identify and prevent potential work disability
through effective early intervention for return to work and to address
disabilities of longer durations that require more extensive medical
and rehabilitative intervention to resolve. An important aspect of the
factor is that communication is maintained with all the parties
involved, including the employee, during critical points in the process
such as immediately following a medical procedure. The focus of
these actions is coordinated communication and policy to facilitate
early return to work. Within the company, resources and strategies
for accommodation are provided in an organized and cooperative
manner. The motivational manner and interactive quality with which
all these actions are carried out can vary from care and concern to
control and cost containment.

There are two facets of responding contained within this
factor. The first facet is the length of responding across time such
that there continues to be interaction with an employee with extended
work disability duration even to a point after referral for professional
rehabilitation involvement and job retraining. The second facet of

responding is the extent to which the employer responds with a

62

variety of flexible, alternative placement options to promote return to
work opportunities. With some exceptions, the items of this factor
tend not to reflect the quality of the behavior or policy as much as
the existence of the behavior or policy. For example, "Duration of
disability is evaluated to identify claims needing case management and
rehabilitation services" does not address quality aspects of the
evaluation nor how or when need is ascertained. It simply measures
the presence of some form of duration monitoring. This reflects the
relative newness of the literature and corresponding practices in the
discipline of disability management where solutions are often achieved
through exploration and innovation as opposed to refinement.

Factor 4 is titled, "Top Management Support and Continuous
Improvement of Safety." The items in this factor reflect the level to
which top management supports safety as evident in resource
allocation and top management's visibility in safety programs; and the
proactive use of data to reduce risk and measure improvement. The
items-reflect the importance of having individuals who are responsible
for safety acting with authority and support provided by top
management. The importance of accountability is demonstrated in the
measurement of safety and disability management performance, and
management principles of continuous improvement are utilized.

The factor inter—correlation matrix is reported in Table 8 along

with the correlation correction for attenuation due to unreliability

63

(Schmitt, 1996). From the corrected correlations it is observed that
Factor 1 and Factor 4 correlate the highest (.62, n = 220). This
supports the expected relationship that a company's diligence in the
practice of safety is associated with their top management support
for safety. Factor 2 and Factor 4 are moderately correlated (.53, n
= 219), demonstrating the association between valuing of human
resources and top management support for safety. Additionally,
Factor I and Factor 2 are moderately correlated (.52, n = 219)
demonstrating the relationship between valuing of human resources
and diligent practice of safety.

Factor 3 correlates moderately with both Factor 1 (.44 n =
219) and Factor 4 (.44, n = 219) demonstrating the relationship
between preventive efforts to reduce work disability and promote
return to work with diligence in safety practice and top management
support for safety. Factor 3 and Factor 2 are moderately correlated
(.40, n = 219) demonstrating the general link between valuing of
human resources and efforts to reduce work disability and promote
return to work.

To further assess for multidimensional representation within
each refined factor, each factor was subjected to a principle
components analysis with forced extraction of three components.
Next, the loadings of the items on the first component were examined

to determine how well each item loaded on the first component in

64

Table 8

.0“ 'c o 5.01 501.. 0' '1 ,1. o ' '0 I '._09

 

 

WW3

Factor 1 2 3 4
Safety Diligence and Hazard Prevention (.87) .52 .44 .62
Valuing Human Resources .46 (.89) .40 .53
Reducing Work Disability .39 .36 (.89) .44
Top Management Support .51 .44 .37 (.78)

 

Note. Alpha coefficients are presented on the diagonal, observed
correlations below the diagonal, and correlations corrected for
attenuation above the diagonal.
comparison to the other components extracted. In all cases, the
items loaded highly on the first component. The lowest loading
within each of the four factors was .51, and in no case did an item
load higher on another component. This provides further evidence
for unidimensionality with each of the four factors (Clark & Watson,
1995)

Woo The resulting four scales were
assessed for normality using the Lilliefors test of hypothesis that the
data are from a normal distribution (Norusis, 1994). With Factors 3
and 4 the hypothesis of normality was rejected, however, it is noted

that with large sample‘sizes small differences can be significant. The

65

factors were observed in graphic form and a negative skew was much
more pronounced in Factor 4 than in Factor 3. Thus, for Factor 4
the property of having a normal distribution has not been met. The
items measuring this factor have a low ceiling and may not capture
the full variance associated with top management support. The
potential .for Factor 4 to perform in the analyses is limited by its
psychometric properties.

The Lavene Statistic was computed to assess the equality of
each factor's variance across the SIC and size categories which were
used as structural covariates in the regression models (Norusis,
1994). All group variances were equal across size categories and all
group variances, except for Factor 4, were equal across SIC code
categories. Visual analysis of box plots for Factor 4 by SIC Code
revealed a noticeably large variance for SIC 20 as compared with
each of the other SIC categories. Thus, as noted above, the non-
normal distribution of Factor 4 may represent a low measurement
ceiling in terms of the items selected for inclusion. Factor 4 does
not demonstrate adequate psychometric properties to facilitate its
performance in the analyses. New items for Factor 4 will need to be
written and tested in future research. As this study is constrained to
the items available, Factor 4 will be used in the analyses, but
interpretation of the results made in light of this finding.

In summary, this study has developed a four factor, 44 item

66

solution with conservative rules for item inclusion while maintaining
alphas in the upper end of the acceptable range, with one exception,
for each factor. These four factors correspond very highly with the
conceptual, theoretical model (see Figure I found in Chapter 1) used
in the Hunt, et al. (1993) study. The content of Factor 2, Valuing
Human Resources, and Factor 4, Top Management Support and
Continuous Improvement of Safety correspond with the theorized
factors of Managerial Style and Corporate Culture from the
conceptual model. Factor 1, Safety Diligence and Hazard Prevention
corresponds with the theorized factor of Disability Prevention
describing behaviors implemented to prevent injury. Factor 3,
Reducing Work Disability and Promoting Return to Work,
corresponds with the theorized factor of Disability Management
describing behaviors implemented after an injury to prevent work
disability. Thus, the four factor solution provides validity to the
theoretical model used to conceptualize the study. Furthermore, a
four factor solution provides a more parsimonious structure for a
refined version of the instrument.
Multiple Regression Reanalyses

The second research question asked in this study was whether
the refined, parsimonious instrument would adequately capture the
variance associated with employer policies and practices that

contribute to the incidence and outcomes of work disability.

67

 

To address this question, the study compared the empirical
performance of the refined research instrument with the performance
of the eight factor solution, 73 item version used in the Hunt, et al.
(1993) study. The validity of the refined instrument was assessed by
replicating the original regression models as closely as possible and
comparing the results with the earlier version. The series of multiple
linear regression analyses performed in the Hunt, et al. (1993) study
was replicated in this study. The predictive variables were entered
using forced entry in a single step. Complete description of the
regression models for each of the nine reanalyzed regressions are
presented in Appendix E. However, the SPSS regression program is
not able to apply a one-tailed test to the t-statistic of the partial
regression coefficient. Thus, the one-tailed significance tests are
reported in the Hunt, et al. (1993) study and two-tailed t-tests are
reported in the descriptions contained in Appendix E of this study.
Two-tailed t tests present a more stringent requirement for obtaining
significance as the confidence interval is split between the two tails
of the distribution. Partial regression coefficients reported as
significant in the Hunt, et al. (1993) study may not appear as
significant in the output generated by the SPSS program and a t-
distribution table is needed for comparative purposes (Glass &
Hopkins, 1984).

In the Hunt, et al. (1993) study three multiple regression

68

analyses were performed within the Prevention Model to explore the
relationship between Disability Prevention (before injury behaviors)
and outcomes of injury incidence (MIOSHA Recordables) and
disability incidence (Lost Work Day Cases and Worker's
Compensation Wage—Loss Claim Rate). In the Hunt, et al. (1993)
study, before—injury prevention behaviors were represented by three
factors: Safety Diligence, Ergonomic Solutions, and Safety Training.
In the four factor solution developed in this study, before-injury
prevention behavior is represented by Factor 1, Safety Diligence and
Hazard Prevention. Reported in Table 9 are Partial Regression
Coefficients for each of the factors for the three multiple regressions
and the three re-analyses using the Prevention Model. Also reported
is the t-statistic, with an indication of its one-tailed significance level
taken from a t-distribution table. In this manner, Table 9 is designed
to facilitate comparison of the three factors from the Hunt, et al.
(1993) study with the single, corresponding factor from this study.
In addition, the R2 and Adjusted R2 for each regression is presented
so that a comparison can be made between the variance accounted for
by regressions from the Hunt, et al. (1993) study versus this study.
As with the eight factor solution, the model based on the four
factor solution, using Safety Diligence and Hazard Prevention to
predict MIOSHA recordable incidents, demonstrated no significant

relationship. Safety Diligence and Hazard Prevention demonstrated

69

signi
ll’orl
In b<
Prev
et al

ICSpI

Mb
6pr
(aft:
C as<
dun
11m
Ma
Pr.

CC

In

significant relationships with both Lost Work Day Case Rate and
Workers' Compensation Wage-Loss Claims in the expected direction.
In both of these regressions, using Safety Diligence and Hazard
Prevention accounted for just slightly less variance than in the Hunt,
et al. (1993) study; Adjusted R2 =.13 to .14 and .11 to .12,
respectively.

Reported in Table 10 are three regression analyses and their re-
analyses which were used with the Disability Management Model to
explore relationships between Disability Management Interventions
(after injury responses) and disability incidence (Lost Work Day
Cases and Workers' Compensation Wage-Loss Claims) and disability
duration (average number of Lost Work Days Per Case). Once again,
three factors in the Hunt, et al. (1993) study reflected Disability
Management Interventions, namely, Disability Case Monitoring,
Proactive Return-to-Work Programs, and Wellness Orientation. The
corresponding factor from the four factor solution in this study is
Factor 3, Reducing Work Disability and Promoting Return to Work.
The second regression attempted to explore the relationship between
these factors and the duration of disability, measured as the average
number of Lost Work Days per Case. Neither the regression from
the Hunt, et al. (1993) study nor its reanalysis in this study produced
a significant F statistic and both had Adjusted R2 values near zero.

In the other two regressions Factor 3 from this study demonstrated

70

 

 

 

 

 

Solution - Dependent Variable R2 Adj. R2
Factor B t

From 8-Factor - Recordable Incidents 0.30 0.21
Safety Diligence -0.066 -0.865
Ergonomic Solutions 0.003 0.049
Safety Training -0.006 -0.082

From 4-Factor 0.30 0.22
Safety Diligence & Hazard Prev. -0.064 -l.127

From 8-Factor - Lost Work Day Case Rate 0.24 0.14
Safety Diligence -0.l66 -l.941*
Ergonomic Solutions 0.031 0.413
Safety Training -0.l30 -l.652*

From 4-Factor 0.21 0.13
Safety Diligence & Hazard Prev. -0.219 -3.298**

From 8-Factor - Workers' Compensation Wage-Loss Claim Rate 0.21 0.12
Safety Diligence -0.213 -3.103"
Ergonomic Solutions 0.044 0.741
Safety Training -0.031 -0.522

From 4-Factor 0.19 0.11
Safety Diligence & Hazard Prev. -0.179 -3.38l‘”

p=significant at .10 level, I"=significant at .05 level, *"'=significant at .01 level

71

Table 10

I'll ll llll-S EEI ll? '1
WW

 

 

 

 

Solution - Dependent Variable R2 Adj. R2
Factors 8 t

From 8-Factor - Lost Work Day Case Rate 0.24 0.16
Disability Case Monitoring 0.147 1.581p
Proactive RTW Program -0.295 -3.235**
Wellness Orientation 0.071 1.056

From 4-Factor 0.21 0.13
Reducing Work Disability -0.130 -2.124‘I

From 8-Factor - Lost Work Days Per Case 0.07 -0.02
Disability Case Monitoring 0.057 0.856
Proactive RTW Program -0.019 ~0.290
Wellness Orientation -0.060 -1.240

From 4-Factor 0.05 -0.03
Reducing Work Disability 0.006 0.142

From 8-Factor - Workers' Compensation Wage-Loss Claim Rate 0.15 0.07
Disability Case Monitoring 0.034 0.496
Proactive RTW Program -0.l45 -1.978*
Wellness Orientation 0.038 0.703

From 4-Factor 0.14 0.07
Reducing Work Disability -0.091 -l.893*

 

p=significant at .10 level, *=significant at .05 level, **=significant at .01 level

72

Table 11

 

 

 

l - 1' I 1'
M l rf r n
Solution - Dependent Variable R2 Adj. R2
Factor B t
From 8-Factor - Lost Work Day Rate 0.18 0.10
Active Safety Leadership -0.113 -l.338p
People Oriented Culture -0.081 -0.893
From 4-Factor 0.17 0.09
Top Management Support -0.051 -0.656
Valuing Human Resources -0.112 -1.3I7p
From 8-Factor - Workers' Compensation Payment Rate 0.16 0.07
People Oriented Culture -0.212 -l.852*
Active Safety Leadership 0.040 0.380
From 4-Factor 0.16 0.08

Valuing Human Resources «0237 -2.224*

Top Management Support 0.062 0.612

 

p=significant at .10 level, *=significant at .05 level, M=significant at .01 level

73

Soluno

From 8

From 4-

Full 4.}

p=51gnl

Table 12

 

 

 

..Hlle ,UO" '0. ' L‘° in 0‘ 1‘1. 2911,99'1

We:

Solution - Dependent Variable R2 Adj. R2
Factor B t

From 8-Factor - Lost Work Day Rate 0.25 0.17
Safety Diligence -0.211 -2.575**
Proactive RTW Program -0.160 -2.134*
Wellness Orientation -0.039 -0.501

From 4-Factor 0.22 0.14
Safety Diligence & Hazard Prev. -0.207 -2.482**
Reducing Work Disability -0.09l -l.194

Full 4-Factor Solution - Lost Work Day Rate 0.22 0.13

Safety Diligence & Hazard Prev. -0.225 -2.392**
Top Mngt. Support 0.059 0.698
Reducing Work Disability -0.097 -l.l98
Valuing Human Resources -0.031 -0.357

g

P=Significant at .10 level, *=significant at .05 level, ‘”=significant at .01 level

74

significant relationships with the dependent variables at the .05 level.
The overall model explained 13 percent of the variance in Lost Work
Day Case Rate compared with 16 percent for the Hunt, et al. (1993)
study, and seven percent of the variance in Workers' Compensation
Wage-Loss Claim Rate which was equivalent to the Hunt, et al.
(1993) study.

Reported in Table 11 are two regressions and their re-analyses
using the Managerial Model. In the Hunt, et al. (1993). study, the
two factors of People Oriented Culture and Active Safety Leadership
represent the Managerial Model. From the four factor solution,
Factor 2, Valuing Human Resources, and Factor 4, Top Management
Support, correspond and are used in the reanalyses. Factor 2,
Valuing Human Resources demonstrated an association at the .10
level with Lost Work Day Rate and Top Management Support showed
no relationship. In the Hunt, et al. (1993) study, People Oriented
Culture showed no relationship and Active Safety leadership was
associated at the .10 level. Using Factors 2 and 4 accounted for just
slightly less variance compared with the Hunt, et al. (1993) study
(Adj. R2 = .09 vs. .10, respectively). Factor 2, Valuing Human
Resources was also related to Workers' Compensation Payment Rate
at the .05 level of significance, corresponding to the finding in the
Hunt, et al. (1993) study where People Oriented Culture was also

equivalently related. Using Factors 2 and 4 accounted for 8 percent

75

of the variance compared with 7 percent for the Hunt, et al. (1993)
study.

Reported in Table 12 is the Summary Model from the Hunt, et
al. (1993) study and its corresponding reanalysis. The Summary
Model was reported to provide an overview of the most significant
factors from the previous analyses. The two factors chosen from the
Hunt, et al. (1993) study's eight factor solution were Safety
Diligence representing disability prevention and Proactive Return-to-
Work Programs representing disability management. Wellness
Orientation was also included to see if a relationship would emerge
in a simultaneous analysis when the elements of prevention and
management were controlled. In the reanalysis there was no way to
represent wellness as none of the items reflecting wellness made it
into the four factor solution. The corresponding factors representing
disability prevention and disability management from the four factor
solution are Factor 1, Safety Diligence and Hazard Prevention and
Factor 3, Reducing Work Disability and Promoting Return-to-Work.
In the reanalysis, Factor 1 performs nearly equivalent in comparison
and Factor 3 falls just short of a significant relationship at the .10
level. The reanalysis accounts for 14 percent of the variance
compared to 17 percent for the Hunt, et al. (1993) study regression.

In addition, Table 12 reports the findings for a Full Summary

Model using all four the factors from this study. This full summary

76

prc

ele
lir
wa

su;

f0
pr

CC

S1;

81

Ct

31

provides an exploration of the relationships for all four primary
elements of the theoretical model in a simultaneous analysis. The
first step in the analysis was to determine whether multicollinearity
was a problem in the Full Summary Model. Lewis-Beck (1980)
suggests that the best way to determine whether multicollinearity
presents a problem is to regress each independent variable on all
other independent variables. If any R2 is near 1.0, there is high
multicollinearity. When this procedure was followed, the highest R2
found was .35, indicating that multicollinearity does not present a
problem in this study. Safety Diligence and Hazard Prevention
continues to demonstrate a significant relationship while Reducing
Work Disability and Promoting Return-to-Work falls just short of
significance at the .10 level. Neither Top Management Support or
Valuing Human Resources demonstrate a significant relationship. The
regression accounts for 13 percent of the variance in Lost Work Day
Rate.

The regression reanalyses demonstrated that the psychometric or
empirical properties of the four factor, 44-item solution are nearly
equivalent to the eight-factor, 73-item solution, but with considerable
efficiency gained in the number of independent variables used in the

analytical model.

77

di

of

cl:

Cl‘.

31'.

61

S:

cl

SC

Classification Analysis

The third research question asked in this study was whether
discernable subgroups of employers could be identified on the basis
of similar profiles of disability prevention and management behavior.

Clusters Based on anto[ Profile. To address this question
cluster analysis was used to determine if meaningful subgroups of
employers could be determined by their factor scores. A cluster
analysis procedure was used to identify homogeneous groups of
employers based on their scores from the four factor solution.
Squared Euclidean distance, which is the sum of the squared distance
over all the variables, was computed for each pair of employers in
terms of their four factor scores. The factor scores are z-scores and
thus are measured on the same scale equalizing each variable's
contribution to the distance measure.

Agglomerative hierarchical cluster analysis with Ward's method
for cluster formation was used to form clusters. In this technique,
clusters are formed by grouping cases, based on the distance
measures for each case to each cluster, into bigger and bigger
clusters until all cases are members of a single cluster. Clusters are
represented by the overall sum of the squared within-cluster
distances, thus each case contributes to the cluster score, and
clusters are formed based on homogenous or "nearest alike" factor

scores..

78

During the clustering process, as the clusters get bigger so do
the distance scores, indicating that the clusters being formed are less
homogeneous. Identifying an "optimal" number of subgroups of
employers involves finding an adequate compromise between
parsimony (a manageable number of subgroups) and homogeneity
(high similarity in the factor scores for the employers in each
subgroup). Plotting the distance coefficient used in each
agglomerative step provides a visual method of observing the
decrease in homogeneity as the number of clusters gets smaller. A
large jump in the sequence of distance coefficients indicates the
grouping of two dissimilar clusters and thus indicates a stopping
point. Figure 3 shows the plot of the distance coefficients and the
clustering stage.

In the clustering procedure, 219 cases were used with one case
eliminated due to missing variables. At cluster stage 218 a
noticeable increase in the distance coefficient in noted, indicating
that relatively homogeneous clusters were formed until the last
possible cluster when two were formed into one. Thus two
subgroups of employers are identified based on similarity among their
four factor score profile. The number of employers in each group
were 107 and 112 respectively.

Winn A discriminant analysis was

performed to validate and help interpret the cluster solution. A

79

 

 

 

 

new Coaldam
a

 

masq-

 

 

 

EjgoxLl. Plot of Distance Coefficients by Cluster Stage
discriminant analysis procedure was used with the four factor scores
re-entered as the independent variables for predicting employer
subgroup membership. The dependent variable was the two
subgroups created from the cluster analysis. One function was
created by the analysis to discriminate among the two employer
subgroups, based on the factor score profiles. The canonical
correlation between the discriminate scores and the groups was .81 (n
= 219). The square of the canonical correlation is the ratio of the
between-groups sum of squares and the total sum of squares and
represents the proportion of the total variance attributable to
differences among the groups. Thus, the discrimination function

Produced was able to account for 66 percent of the total variance

80

attributable to differences among the employer subgroups.
Interpretation of the function proceeds similar to that in
multiple regression since the coefficients for a particular variable are
dependent on the other variables in the solution. Since the variables
in this function are measured in standardized form, their relative
contributions to the function can be observed in the function
coefficients. Further understanding of the factor can be derived from
examining the correlations between the values of the function and the
values of the variables. Table 13 reports the function coefficients
and the pooled within-groups correlations between the variables and

the canonical discriminant functions.

 

 

Table 13
no 9 "it t H 0‘ I it '00 0 A II". o
0 'ou. 0'1“. -. t 01' -l°l«, 9' urn I H
Function Pooled Within-
Factor Coefficient Group Correlation
Safety Diligence/Hazard Prev. .7903 .77
Valuing Human Resources .0212 .34
Reducing Work Disability .5994 .54
TOp Management Support .1460 .40

‘

Factor 1 demonstrated highest coefficient within the function

81

and high correlation with the discriminant function scores (.77, n =
219). Factor 3 demonstrated an important contribution to the
discriminant function with its coefficient level and it correlated
moderately with the discriminant function scores (.54, n = 219). The
remaining factors, Factor 2 and Factor 4 correlated moderately with
the discriminant function scores (.34 and .40 respectively, n = 219),
but were of comparatively limited contribution in the discriminant
function. Thus, the discriminating function relied to a large extent
on the performance of an employer for Factor 1, Safety Diligence and
Hazard Prevention, and Factor 3, Reducing Work Disability and
Promoting Return to Work, compared with the remaining two factors
to distinguish between the two groups which were created in the
cluster analysis using all four factors.

Finally, the discriminant function was used to assign employers
to the two cluster groups to determine its ability to correctly classify
the employers. The proportion of the 219 employers correctly
classified was 94.06 percent, supporting the validity of the clustering
solution. Therefore, the two groups of employers are presented and
described based on the four factor scores. Table 14 reports the
group means and standard deviations for the two clusters on each
factor.

The discriminant function provides a straight forward

eXplanation of the two cluster groups. The employers comprising

82

cluster 1 demonstrated, as a group, higher achievement of work place

policies and practices as measured by all four factors, than did the

employers comprising cluster 2.

Validation of this cluster grouping

would be further supported by comparing the cluster means on the

outcome variables for significant differences.

It was expected that

cluster 1 would have significantly lower group means on each of the

six dependent variables used to measure work disability.

 

 

Table 14
:a - a‘ 0,0u‘,. :30 to 09‘ a a t .0
Clnsim
Safety Valuing Reducing T.M.
Diligence HR Work Dis Support
mean mean mean mean
Cluster s.d. s.d. s.d. s.d.
1 0.7451 0.4343 0.6166 0.5072
0.577 0.973 0.564 0.636
2 -0.6809 -0.4159 -0.5898 -0.4673
0.973 0.839 0.972 1.041

 

Multivariate analysis of variance was used to test for

differences between cluster means for each of the six outcome

variables. Hotellings Test of Significance with 6 D.F. was significant

(p < .05). Therefore, Table 15 reports the results of univariate F-

tests. The outcome variables were entered in log form.

83

There is no difference between the performance subgroups
(Cluster 1 and Cluster 2) for incidence rate of MIOSHA recordables
(MIOSHA Recordable Rate), average duration of a lost work day
case (Lost Work Days Per Case), number of workers' compensation
wage loss claims per 100 employees (W.C. Wage Loss Claim Rate),
and the average workers' compensation losses per employee (Workers'
Compensation Losses Per Employee). The number of lost work day
cases per 100 employees (Lost Work Day Case Rate) and the number

of lost work days per 100 employees (Lost Work Day Case Rate) do

show a significant difference at the .05 level.

 

 

 

Table 15
u a 9 a ' .l t o 9 I o l u l H
QuimchLaiiables

Source Hyp. MS Error MS F Ratio
MIOSHA Recordable Rate 0.0026 0.5648 0.0046
Lost Work Day Case Rate 2.8279 0.5734 4.9317*
Lost Work Days Per Case 0.0949 0.2470 0.3845
Lost Work Day Rate 3.1210 0.6684 4.6688"
W. C. Wage-Loss Claim Rate 0.8540 0.4230 2.0190
W. C. Losses Per Employee 0.0895 1.2664 0.0706
Note. Univariate F-tests with (l, 140) D.F. *o < .05

84

Thus, creating performance subgroups by differentiating between
employers at two levels, based on their achievement of policies and
practices as measured by a profile of the four factors developed in
this study, is validated by a significant difference in the work
disability variables for disability incidence (Lost Work Day Case
Rate) and a broad, summary measure of work disability (Lost Work
Day Rate).

Post Hoc Analysis
W. The final step in this
study was to assess the extent to which these performance subgroups

could be predicted from demographic variables. If demographic
variables could be used to predict high or low disability outcomes,
then information which is easily available could be used in a
screening instrument to initially classify employers at some known
level of probability. Logistic regression analysis, which can be used
with dichotomous dependent variables and either dichotomous or
continuous predictor variables, produces a linear regression equation
which selects coefficients that maximize the likelihood of classifying
a case correctly. Therefore, logistic regression was used to explore
the relationships between covariates and high and low performance
subgroups of employers.

In the Hunt, et al. (1993) study, there were 11 covariates used

in the regression analyses. These included whether or not a labor

85

union represented some of the employees, percent of employees on
rotating shifts, percent of employees with tenure less than one year,
percent of the workforce salaried, and the average hourly wage paid
to employees. Three covariates describe characteristics of an
employer related to risk management. These include the type of
insurance (individual self-insurance, group insurance, and commercial
insurance); safety standards required for accreditation or affiliation;
and the use of risk management services that are offered by insurance
companies, industry affiliated groups, or outside consultants. The
remaining covariates are structural in nature and included the size of
the employer, if the company was comprised of more than one work-
site location, and the industry classification.

While there is some inherent structure to these covariates, there
is no theory to determine a priori which variables should precede
others in a hierarchical analysis. The covariates were entered
simultaneously into a logistic regression for purely technical
prediction purposes, to see what their performance would be in
predicting an employer's classification based on the four factor
performance profile. Thirty-four cases were excluded due to missing
data, 186 cases were entered into the analysis. There were 91 cases
in the high performance group (cluster 1) and 95 cases in the low
performance group (cluster 2).

The regression function from the logistic analysis correctly

86

classified 59 percent of the high performance employers and 68
percent of the low performance employers, for an overall correct
classification rate of 64 percent (correct classification, n = 119;
misclassified, n = 67). Having multiple site locations (WCCIR) was
significant at the .01 level and belonging to SIC 25, as compared
with the other industrial classifications included in the regression,
was significant at the .05 level. Both variables demonstrated a
coefficient which was associated with greater likelihood of being a
low performance employer. Having individual self-insurance was
significant at the .10 level and was also associated with a greater
likelihood of being a low performance employer.

Thus the variables of multiple site locations and SIC 25 were
the only variables to be assigned significant beta coefficients. The
regression equation combined for a correct classification of 64
percent of the employers, which is above the expected chance

classification of 50 percent.

87

CHAPTER 5
Discussion

The purpose of this study was to develop a research instrument
which would correspond with the theoretical organization developed
in the Hunt, et al. (1993) study but achieve greater parsimony than
the eight factor solution utilized in the Hunt, et al. (1993) study.
This study was organized around three research questions. First,
could reanalysis procedures produce a refined version of the original
instrument that would be more parsimonious, yet equally effective in
capturing the hypothesized theoretical structure of the role of the
workplace in preventing and managing disability? Second, could the
refined, briefer instrument adequately capture the variance associated
with employer policies and practices that contribute to work
disability? Third, could subgroups of employers be identified on the
basis of similar profiles of disability prevention and management
behavior that have implications for prediction and screening?

[2 C l E l' 1.].

Refinement of an instrument is important to on-going research
for the purposes of demonstrating construct consistency and for
raising questions that motivate theory building and improvement
(Smith & McCarthy, I985). The challenge of the first research
question in this study was to determine whether a refined

parsimonious instrument could retain content coverage of the targeted

88

constructs as outlined by the theoretical model of prevention and
managing workplace disability. The refinement sought to promote
item reduction, unidimensionality, and maintenance of internal
consistency in a manner that would facilitate stability and enhance
the potential for replication of the factor structure on an independent
sample.

From the results presented in Chapter 4, one can conclude that
a more parsimonious model for representing the variance structure
within the "Disability Prevention Among Michigan Employers" survey
questionnaire does exist. Principal-axis factoring with orthogonal
rotation produced four unidimensional factors which correspond in
content with the four main areas of employer controlled influence on
work disability described in the conceptual model of work disability
outlined in the Hunt, et al. (1993) study.

The four factors that were extracted demonstrated a range of
factor loadings consistent with the findings reported by Guadagnoli &
Velicer (1988) for factor stability with sample sizes of 200. In
initial scale development a benchmark internal consistency level of
.80 is recommended (Nunnally, 1978; Clark & Watson, 1995). In
balancing parsimony with internal consistency, a target Cronbach's
alpha was set at .85, where possible, to provide a cushion of internal
consistency for maintaining the benchmark level when replicating the

scale on an independent sample. Factor 1 (.87, n = 13), Factor 2

89

(.89, n = 10), and Factor 3 (.89, n = 16) achieved the target level of
alpha following complete iteration of the chosen conservative rules
for item retainment. The internal consistency for Factor 4 was
acceptable (.78, n = 5).

In the Hunt, et al. (1993) study, factors from the eight factor
solution corresponding to before injury behaviors aimed at preventing
the occurrence of a work-related injury of illness were Safety
Diligence, Ergonomic Solutions, and Safety Training. In the Hunt, et
al. (1993) study these three scales were comprised of 21 items. In
this study, Factor 1, Safety Diligence and Hazard Prevention
corresponds to this area and contains 13 items. The focus of Factor
1 is diligent and meticulous application of safety behavior, and its
items represent active attention and thorough involvement in hazard
reduction and risk prevention.

In the Hunt, et al. (1993) study, factors from the eight factor
solution corresponding to after injury responses aimed at monitoring
work disability and promoting return to work were Disability Case
Monitoring, Proactive Return-to-Work Program, and Wellness
Orientation. The three factors in the Hunt, et al. (1993) study were
comprised of 24 items. In this study, Factor 3, Reducing Work
Disability and Promoting Return to Work, corresponds to this area
and contains 16 items. The focus of Factor 3 is continued

monitoring of work disability of longer duration to ensure appropriate

90

treatment and services for resolution and identifying and preventing
potential work disability through effective early intervention for
return to work.

In the Hunt, et al. (1993) study, the factor from the eight
factor solution corresponding to cultural characteristics and human
resource orientation was People Oriented Culture, which contained 13
items. In this study, Factor 2, Valuing Human Resources
corresponds to this area and contains 10 items. The focus of Factor
2 is effort on the part of management to use employee input in
meaningful ways in company decision making and to cultivate
employee growth.

In the Hunt, et al. (1993) study, the factor from the eight
factor solution corresponding to managerial style was Active Safety
Leadership, which contained 13 items. In this study, Factor 4, Top
Management Support and Continuous Improvement of Safety,
corresponds to this area and contained 5 items. The focus of Factor
4 is support of top management to the safety program through
resources and visibility, accountability through measurement, and
continuous improvement.

Thus, the Hunt, et al. (1993) study utilized 73 items across
eight factors. This study achieved reduction to 44 items across four
factors. It was important in the systematic refinement procedures

utilized in this study to retain, in balance with parsimony, an

91

adequate level of content coverage. Constrained to the items
available in the original instrument, it is believed that content
coverage for the targeted constructs was adequate in that for the
three broader constructs ten or more items were retained with
moderate inter-item correlations and significant item overlap was
eliminated. For researchers interested in including employer factors
in their models of work disability, they now have a theory based,
refined instrument. In some instances, this level of parsimony is not
adequate for the constraints with which the researchers are faced.
Further research is needed to determine, separate from the theory of
work disability, which specific items may be highly predictive of
outcomes.

In the multiple regression analyses, Safety Diligence and Safety
Training from the Hunt, et al. (1993) study were associated with
Lost Work Day Case Rate (p < .05), and Safety Diligence was
associated with Workers' Compensation Wage-Loss Claim Rate (p <
.01). In this study, Factor 1, Safety Diligence and Hazard
Prevention was associated with both of these outcome measures at
the .01 level.

Furthermore, the discriminant function which confirmed the
performance subgroups contained a comparatively large coefficient for

Safety Diligence and Hazard Prevention compared with the other

92

three providing further evidence for the importance of prevention in
achieving low work disability. Safety Diligence and Hazard
Prevention demonstrates a prominent place in the model of work
disability.

In the multiple regression analyses, Disability Case Monitoring
was associated with Lost Work Day Case Rate (p < .10). Proactive
Return-to-Work Program was associated with Lost Work Day Case
Rate (p < .01) and with Workers' Compensation Wage-Loss Claim
Rate (p < .05). In this study, Factor 3, Reducing Work Disability
and Promoting Return to Work, was associated with both Lost Work
Day Case Rate and Workers' Compensation Wage-Loss Claim Rate (p
< .05). The refined factor did not demonstrate as strong of a
relationship with Lost Work Day Case Rate as did the more specific
Proactive Return-to—Work Program from the Hunt, et al. (1993)
study. This may suggest that in reducing items and factors in favor
of parsimony, the ability of the factor to measure an employers'
provision of appropriate, alternative return-to-work options was
compromised. The ability to measure an association with a
dependent variable is maximized by independent variables that
corresponds in level of specificity DeVellis, 1991). Factor 3,
Reducing Work Disability and Promoting Return-to-Work achieves
parsimony, but perhaps at the expense of specificity .

However, the ability of Factor 3 to demonstrate a relationship

93

with both Lost Work Day Case Rate and Workers' Compensation
Wage-Loss Claim Rate demonstrates the general and significant
relationship between after injury response behaviors and the incidence
of lost work day and workers' compensation cases and the relative
importance of this factor in the conceptual theory of work disability.
In the discriminant function which was produced, Factor 3 had the
second largest coefficient and provided an important contribution to
the equation.

In the original multiple regression analyses, People Oriented
Culture was significantly associated with Workers' Compensation
Payment Rate (p < .05). In this study, The performance of Factor 2,
Valuing Human Resources, revealed an association with Lost Work
Day Rate (p < .10) and an association with Workers' Compensation
Payment Rate (p < .05). These associations provide validity towards
the inclusion of workplace culture in a model of work disability.

In the multiple regression analyses, Active Safety Leadership
from the Hunt, et al. (1993) study demonstrated an association with
Lost Work Day Rate (p < .10). In this study, Factor 4, Top
Management Support and Continuous Improvement, did not
demonstrate a significant relationship. This area of employer
influence, as represented in both the Hunt, et al. (1993) study and
this study, has not demonstrated a level of association comparative to

the other three areas. The lack of comparative association in these

94

studies raised several questions.

First, does this factor belong in the model as a distinct and
separate area of employer influence? This question is supported by
the fact that a number of items were eliminated from this factor due
to the fact that these items also loaded at a significant level on
Factor 1, Safety Diligence and Hazard Prevention, and Factor 2,
Valuing Human Resources. In fact, a eight of the items loaded
significantly on either Factor 1, Factor 2, or both. The existence of
relationships between Factor 2 and Factor 4, and Factor 1 and Factor
4 suggests that Factor 4 may be describing a link between Factor 2
and Factor 1, and not a distinct construct area.

Retaining the belief that Factor 4 is a distinct construct area
begs the question of how the scale fell short in measuring that
construct area. The analysis of psychometric properties demonstrated
that the distribution of the scale, Factor 4, in this study negatively
skewed and significantly different from normal, thus violating the
assumptions for multivariate analyses. This scale had a low ceiling
in that most employers reported achieving a high level of top
management support as measured by this scale. It may be that the
scale failed to capture the true variance existing within the sample
and thus was less likely to reflect the association that may exist in
the sample between top management support and outcomes of work

disability.

95

,The parsimony achieved in modeling employer factors
influencing work disability with a four factor solution allowed the
full model to be entered into regression analysis. This full regression
model, incorporating all components of the theory, provides
additional information on the inter-relationships of the components
within the theoretical model. Simultaneous analysis of these
components demonstrated the relative weight and importance of
preventative behaviors in the relationship of disability prevention and
management behaviors to overall work disability performance.

However, while the achievement of parsimony provides the
opportunity to represent the full model of employer factors, the
results suggest the more general construct measures may be less
effective in demonstrating the relationships between the employer
factors and measures of work disability. Controlling for all other
factors present in the model, Safety Diligence and Hazard Prevention
demonstrated its association with the dependent variable of Lost
Work Day Rate, the summary dependent variable of work disability
reflecting both the incidence and duration of lost work day cases.
However, the fact that Factor 3, Reducing Work Disability and
Promoting Return-to-Work, did not demonstrate an association with
workplace disability in the summary model, despite doing so in the
more narrowly modeled regressions of Lost Work Day Case Rate and

Workers’ Compensation Wage-Loss Claim Rate, suggests that more

96

narrowly defined constructs and associated measures of employer
behavior may be needed to demonstrate the specific influence of
employer behaviors on work disability. For example, in Factor 3, a
few items were dropped from the Proactive Return-to-Work Program
factor used in the Hunt, et al. (1993). It may be that the associative
power of certain specific behaviors (e.g., return to work) are muted
when they are contained with other specific behaviors (e.g., claim
management) to represent a broader construct (e.g., post injury
response). -
EmnlmLSuhannns

The third question raised in this study was whether employers
could be classified into subgroups based on their performance as
profiled with the four factors. If achievement levels on the. four
factors resulted in employer subgroups that could be classified, then
the outcome and covariate levels associated with those subgroups
could be examined to learn more about how the process of work
disability may differ for different types of employers.
Classification of the employers using their scores on the four factors
resulted in two clearly distinguishable groups. A discriminant
analysis demonstrated that these two employer groups differed on
their achievement of work disability prevention and management
across all four factors. Thus, if employers reported high achievement

on one factor, they were likely to report comparatively higher

97

achievement on all of the factors. In a MANOVA analysis the
performance subgroups created by the discriminant analysis also
demonstrated significant differences in Lost Work Day Case Rate and
Lost Work Day Rate. In particular, this finding provides further
support for the important role of Safety Diligence and Hazard
Prevention and Reducing Work Disability and Promoting Return to
Work in modeling the influence of employer factors on work
disability.

Using covariates to predict membership in high and low
performance subgroups demonstrated that the presence of multiple
plants is associated with the lower performing subgroup. The
interpretation derived from this finding is that multiple location
plants are an impediment to the achievement of higher levels of
workplace disability prevention and management.

Another finding is that SIC 25 (Furniture Manufacturing) is
associated with the lower performing subgroup. This finding is
probably best explained in light of the findings from the qualitative
portion of the Hunt, et al. (1993) study which reported that those
employers who were engaged in and knowledgeable of innovative
disability management solutions, rated their achievement on the
disability management scales with a tougher standard of reference.
Within Michigan, SIC 25 is known to have companies who have been

recognized for their innovation, leadership, and achievements in the

98

area of disability prevention and management (Wasserman, 1993;
Habeck, Williams, Dugan, & Ewing, 1989). This finding and its
interpretation in light of the qualitative findings of the Hunt, et al.
(1993) study serves to highlight the importance of replicating these
findings on an independent sample.

Implications for further research include the need to perform a
confirmatory factor analysis. A confirmatory factor analysis would
assess for the goodness-of-fit of a four factor model in explaining
the variance. Furthermore, a better exploration of the strengths of
the relationships between the four factors could be modeled and
analyzed. In particular, the theory needs to operationalize the path
of influence of workplace culture on work disability outcomes, and
the extent to which Top Management Support and Continuous
Improvement of Safety is a distinct construct.

Furthermore, this study is limited by the possibility of reverse
causation. The employers were surveyed in 1991 and asked to self
report their disability prevention and management behaviors.
Outcome measures were provided from the MIOSHA log data for the
years of 1987 - 1989. The study correlates prior outcomes with
later behaviors thought to be associated with those outcomes in a
proactive, not reactive, manner. Further research with a prospective

design is required to establish an appropriate sequential relationship

99

between behaviors and outcomes.

Additionally, the study demonstrates the achievement of
parsimony in the number of factors used to model the realm of
workplace constructs associated with work disability outcomes.
However, in achieving parsimony specificity was compromised. The
results suggest the need for research that attempts more parsimony,
yet with better specification. This might be achieved with the
identification of hierarchical levels of constructs.

Workplace disability is a multifaceted phenomenon.
Increasingly, employers have realized that disability incidence and
duration can, to some degree, be mediated by workplace factors.
Researchers attempting to model and understand the influence of the
workplace on disability, particularly within a broad context, need
empirically sound instruments which provide specified coverage of the
target constructs and appropriate reliability so that informed
decisions can be made regarding the need for specificity versus
parsimony. Further research that explores greater levels of
specificity, particularly in relation to more specific outcome measures
is warranted. For example, neither the eight factor nor the four
factor solutions demonstrated theorized relationships with recordable
incidents or work disability duration . Systematic error in the
measurement of these outcomes may have contributed to the lack of

expected findings. Further research is needed to explore the

100

relationship between employer factors and accident incidence and

disability duration.

101

APPENDICES

102

APPENDIX A

Disability Prevention Among Michigan Employers

103

W. E. UPJOHN INSTITUTE
for Employment Research

 

 

mmmhummmmmummommm
dwnmomummmummum havoc-chm
coca.

Ymmnmmahéuuwuuany.mmdum
ththWdhdeamwflaflhMbqo-edvm
uumuwmdwm.

lfyoabvenymaouIIeNerm'aanfludmphneanmo) 343-5541.“
mkmMPh-amfiamhbwwm»:

nan-n-
wanna-humus
”HWWAv—e
mung-m4“

 

104

Part1. OrganhaflonalSelf-Aaaeeunent

mmmmmofpdkbndWMempImmymmmmedabofmjufiam
diaabilhy. Weundumndthatnoeompanyisinvolvedinallmeeeaetivhiea. andmathtrealhytheeeau'aegieaalelnrd
naehieve. Therefore. hishnpornntthatyoueflh‘edym, fi'omyourperapective. meexaenttowhichymrorganintion
MMMMhachm. Memeevelyhemusingtheaealepmvided bycirclhigthebeatreeponae
Ioreach'un.themhnaqplieabkwyoorm.plaaedmh[ll.indieafingdmhmm.

we“

Phaeteflobyeomiduhgtheamflmlehmww
MMHMWMmdfiMWMMM
munch”)

.Topmprovidealeadenhipandaefivelymin

”sumptuous.

.Topmmmenferymbymm

manila-lining“

. MWMpruapprwI-hmflmwymb.

4. mammmamws)mm

10.

ll.

.‘I'heaafluywreedvaamanm
.mumwaummuu

Wane.

.Topmregub-lymuemy'aaeddemnd

.‘l'hmpanyemhafundabadtuanaafeeoodhiouand

ctr-inal-

.Meoqanyauiveaforeommmbafiaty

pater-mace.
mamwmmmmmh
Topmheom’medlommhufloy-

”My

Nwmwomwmmfimmaoevalmm
rehaforeeaafetyperfomneeJIe-emememaovhiehyouoaeeaeh
oflhemelhodadmaibedbelow. (Chelethebeetreapomeforaehhem)

mmuwmmwmm
andlorwrhunpnhe.

. Violating safety mleareanha indiaciplinryaedoa.
. Themaarelifleaymiormeeaumm

eeadifiooswimomfearofreprhal.

. wmwmmmummm-
perform.

beekmlheh

105

New
t0!)

the
(0%)

w
(“25$
elk“)

My
(he. 25%
ethane)

I“
meant
and.)

3*
(aha-amt
ethane)

I)"
(“15!
due“)

0"
(m 15!
attain)

Alum
(IQ!)

Mm
(loos)

ll.

.Safuyperformehevahmdaspanofmperviaon‘performee

appraisal.

.Squviaors mmmwy.

andooluidelm.

adamaeondnaedateguhm.

. Themideotifieaapeeificjobaandmwhhhiflac-

mmmoeenpafieullneahhndaeeidadnbmyu
mmwmmmmmmuu
woman-um.

my“

MMhammembmm.
MahmuwdwhWyI-edxcm
hemmebrmhhn.)

I.

Safetygealaaledevelepedndeommimdaoevelyue.

2. m-aymamuuwary.maym

3.

10.
ll.

12.

m bidentify andaddreas aafety problems.
Wmhflamedabahpoaiblehamdsoflhehjobaandm
trained in aafeworkpnetieea for theirjoba.

. mmmwummwwfic

mummmjobwmbemmouujos.

. Tmormuflyanimdawloyeumgiveoniniuon-

abbefuebeingplaeedonajoborworkhgwhhaewequipm.
Warehfamedfiootpoasibleham'dsatflminediaaafe

.mmmmmmmmm

theyoeenr.

. Wmminedhownmndmmfewm

doe-waters.

. Wmnmapdnahudovnmmfemormp

hemkpmeeaawhenanonaafeeondidonuiaea.
WMmeonecaedonadmelyhaah.
Macadam-em. idemifyhngenneaandineluliog
MWWan-lyahofinjurymdiflmtnarein-

106

New
(0‘)

(to-25%
cliche)

'0

I”
In“!
ofht'm)

M
(“15!
dulled-Ia)

(lQ‘)

m Wet-hm

mmwmnmumnvbflmfimm
riabbyamdingtolhephyaialenviminwhiehworkhpafum-

d.

99w»:-

0

(Chelelllebeatreqonaeforaehium

Whadlnfim'nad.

.Worluamepenooalplouaivemmeindhmd.

mmummmhmm
mandhealdiiamueeoosidendiameaeqohin‘ooofm

.mmuwammmmmn

WWW.

7. Jobaaremodifiedwhgheavyandrepedfivelimqaoaannim.

8. mebwrepeddvem.

9. wwmunmmwdm
flow.

10.

WMcpbmh-admejwambefar-
“MN.

MW”.

mmmmh'smnmm’m
Wehhnawheomqm.1‘omqananaehofu
WWI-edinmeChdemebeumfor
.ehiun.)

. mmdmmmmdmaw

aibleloenployeeaw'ingan vorldnghowa.
‘ ' mewumomw

3. Dinbilhybaefitdlaekaareianedhialinelym.
4. mmmmofimunwum

“Wm-Wk”.

. Wmmmdfirmwkaymanflfimwfie

objeefivealoachieve.

.Wmmwmymmmm

maurmmqummmn
work.

.MJWEMPMMWCO:

. “WMmmedaoevah-nwkapaehyand

MW " MM" ' plmwheninjwedworkenare
unableuorenuneemploym.

107

Norms-I-

(01)

—-~

New
0!)

(“23!
often)

NNNN

(~23!
oft“)

(mu!
dubs)

UWWU

I“
til—ht!
dud-e)

I”:
(about 75$
datum)

b“b

I”,
(ha 73%)
o! I: the)

(lQ’)

MIRIAM

Alma
(1m!)

immeeemyrefenforpmfesionaleaaemor

“mama-workplace“.

10.1bemymmbmdnqmlhymefiecdm
ofmedicalandrehabiliratiooeareprovidedtohahijm'edanployeea.

ll.lapoaibilhyfordiaabilityclaimmanagememndm~to-work
mumbndwaapecifieindividmliameeompany.

[Icahnwhwelleoordinmedbminhhlinfilytoelaim
molded.

My“

Mumm,mummmuh
phendamflyueaioeaaeaofhy‘mynddinbilhyflovham
mumwmwhmwummm
mama-n)

Lumeheaanm-pamwlhy
haddnhmrduiomwlhymm
LAWN” Mbealphyam" abomyourjoba
dmmhnflym' WM
3.13juedqnployeeaareeollaccdbyadeaigmtedpenonwhhialhe
4.Fonow-tpeomwhhdnabled' worker'saremadeatlegularw-
valabyaeompanyrepreeerm'veaeeordmg' mammalian.
5.1‘heueningphyaie‘nniaaakedtoidemifymmiedeoaand
Wanfluammfwm.
6.1'heeompanym' ' regnhreommeanon“ whhmemfimd'
l 's l' I .. -
7. mMMawmuq-nauuw
dualism.
tuneup-hymen!“ plenum“ andmodrfied’
jobdonea' toremndudaled' unployeestowork.
9..Theeoupanyaaeamaidiasaasiativedevieesandflexible
Maehedflingmfacilimpmofmwm.
10AM ispuov-ded' towel-mm tomhejobaeeemmodnons'
- orpuehaeapeenl‘ aervneea' neededtoaaaiaremm-to-worlt.
ILWhuuinjmedvorbrismbletoreeumepriordodeameeom-
mmhpbmfihhmhamm.
12.Follow-opeomactiamadewithdieemployeeandmperviaorafler
MWwd-lwhhanyneededm.
lime“ hcleuiyelpmaed' whhanngned'
1.1.. ‘
14.1MehW-dewrdimmdepumindfom
nominjmadamloyeeanwork.

108

Neva-MW”

0S)

(cl—25$
died-a)

tit—hi!
«math-e)

(den 79!
attain)

(1&5)

mum

Someeoq-ieeuytoidendfyorpreveuvariomriakfacmdmmay
Ieadwemployeediaabiliryflowlmexnnthasyourfirmbeoomeinflw Never My Sen-h- ood, Ale-ye

ed'mtheriakprm‘ smegm' liaaedbelow? Circledaebeetreeponee (OS) (m3! (mu-Ir clears: (loos)
bra-china.) ( omens-c) erase-m done-n

l. wmhmbmmmmmw-

fumieranredlaabofpardenlarjobantely. l 2 3 4 5
2. Wmmeenedforjobreluedhealdiordiaabilhyriahon

aoomhndnghaeia. l 2 3 4 5
lawloyeeaaremonn’dnplonpdyrwonphyehlsymuh-

inghomjobtaeb. l 2 3 4 5

4. Wmnnhndnmgdnjobwmu
.yhdiaeeoployeediffialna(mehaantmeabue.m.
pennalprouema). l 2 3 4 5

5. Theeompaoyaedvelyprmdlenaeofancnfloyaeaaeiannee
mmnwmmnmmdm
mmimerferevhhwort(mehasntmeeabnee.aneea.per-

aornlproblelns). l 2 3 4 5
6.1'heeomanyoenlnhameeenompponhealhpromodonor
velhesprograns. l 2 3 4 5
7.1'opnnngememmonandpmieipnainheahhpromodon
(wellneeshotivhiea. l 2 3 4 5

8. memwmmmww
riakheeen.

9.Theeoomanyaeraensjobqlieamsforillegalmme. l 2 3 4 5
lO.Theeooyanyeonbeta“foreauae"mhatanoeabmeuetingofhs

airplanes. l 2 3 4 5
mm

Floally.holybelhtwatyleandorganinioml“enlmre"
relmhnaomewaytoaaferypefiom-Ieeanddhabilhyeoatajheeeon-
aideryouroooyanyenvhonmeotaodcritieallymediemwwhich Never anon-lg Send- H Ahoy:

thee: mm your organintieo. (Circlethe beet reapome 105) It!" 25% (about Hf (thou 75$ (100$)
fore-damn.) Michel and.) annual

I. Owneralipandaeembilityarepmhedtodleloweetlevelaofdle

W l 2 3 4 5
2. Mmmmnaboutmininganddeveloping

pcamneldu'oughhehunnnreaomeepoliciesaodprograrm. I 2 3 4 5
3.10baatiafaetionamongeuployeeeatdlheawanyhhigh. l 2 3 4 5
4. Working relationflnpe u'e colhborative and W in this

company. I 2 3 4 5
5. Thereiaahighlevelofuuatinmeenployee/employerrelanonahip

athiseowany. l 2 3 4 5

109

10.

ll.

13.

14.

15.

. Sflnshmhflding.enecbing.prebhnsolving.andm

nonmembemrsmmeseleodonofmvisusandm
atdn'seooyeoy.

. Wadwmnehedinwafllsmas

. Anorguiaed. eaeeriveprooeesisesedforpievanoesndeonfliet

mmum.

. WNMWMWWMW

asareslinebeds.
WmhnHyhehdsdiniem'sgoelaenhgnd
phingproeess.
mmmmmmwu
mummdm. «destinations.

. mmmmwwmum

mdMeeeboflbeeoqny.
w aeeksnd col-flees We: but in my
«Incident

WWW. nebugnlhyeirdesendbor-
mum-Ins. arenaedsogmeluenployeeper-
mumm.
wmmmeeeuoloverwkploeessndprohedvhy
dun-Is.

110

Partll. Organizatioml Summary
mmwsforspxifichmmnmmlwmhowmebehavbumedmhflrdmm
mm.hmm.hmybemaryfmywwcomuhwhbofiersinywmindmm
obnmmismformdon.WhaeemMaannmavdhble,mmabkeahmtesmwfficiau.Youreffortto
mflMhafiaflmefierhMJndwfllemblemwmmem
puefiwnlyshdymrfirm'swddemafidhafihqperfmmdmwhveofkredwwidemyw.We
meyoutbddseeonfidemhlhyofyuhmeswinbewoseeedatdnUpjohlnsdnm.

“new
Ql.Wharhyowworhs'eonpenlionhurmeemoe?(Cirelel.2.or3)
Lluivihalsdf-inseruoe
DoyouuaeaIbhd-persymmhdew)
No
Y8...Am-Il
lamest-henna:
Grown-no

 

 

lilo-loam
Carrier-In

02.Huyomemynodvedhdmolservbawoouemmmepedmvyeus?(firdemba)
l.No(GosoQ3)
2.Yes...Fromwbom?(Ch-elelasersofdiosedsatapply)
A.Privuseenraonreeseehasinsuraneeearrierornadeusoeinion.
B.hNhnmmum¢uMkhideLaba(Sfl).CmeMkm

 

 

Deparm
rommmmmimpnmmmmwucmm)
W
(on) (IOS) (255) (>505)
Privec SeesorServiees l 2 3 4
Public Sector Services I 2 3 4

Q3. Bmmyrqdmbmnhymmmwdbyanfiwmwmymkguw
aeeredhadon)? (Circlennrnber)
LNo
2. Yes

WWW-d”
For me medians in this section please fill III the meals at pen-eats. using minutes when necessary.
Ql. Does this firm have mltiple plants or facilities? (Circle umber)
I. No
2. Yu.. .Plaeeindieasewhiehspeeifiephnds)orfeeility(s)yuurespaeesreferm.

 

 

I986 1987 1988 1989
Q2. Numberofeoployeesarthisheilhy:
Falldrne
Pandrne
TarporaryorContract

 

 

 

 

 

 

lll

Q3.Approa'nmapmofeurerhwfloreewhoarez
Sdariedteserwt)
”(non-w)

a.

 

Q4.Approahmamemavengebonrlywagefornen-¢xenpsworkers S—
inmlramofyowjobmficamsdoyongenerahylire?

06.Wperesmofworflereehidnefouowingjobenegoriesz

 

__S

B.W.nelnioalndapponsrafi __S

cmmorrlreaueprwidan _5

rural 100a

Q‘LWp-umofmwbowkrodngm ___$

«.mpmdmtbhyumwm _S
Q9.Wpuouhofeumlryloyeesubo:

Flavebe-vhbdleeenmelyhuMQe(l)y-r .__S

Bavebesnwhbteeoqaoymoredlnunumyars __S

Havereedveddgnifiennewmormindselastyeu __S

crewman-vmmmm ___uyloyees

Qll.Wu¢l-obadalrployeeshving(m)iolm W

QIZ.WyvhrpueadWI-vingmhy-ofisormdne
nbednassooedin‘orl?

QI3.IsuyolyouwfloreeartisfaeilhyMbyaudon?(Cirelemer)
1.No Wyndham)
2.x....wmmdmmaw_s

Whatudonsarerepreaemsd?

 

 

 

Mohdoudmdwafibnammfiqmbipmnfidem)
Never Oeeadoully Sol-nines Usually Always
l 2 3 4 5

Approdonesenlmotcofminlm—

112

mun-n

mmwquI-mmquummmmawm
ummm.mmmummuumm.

01.

02.

Q3.
Qt.

Tmmdwm-whmmnm
(Ml.2.6+colm8.9.l3fmumm

19“ I987 1988

1909

 

Tommdmdabbmumhinginlwwkdnys
(M3+cdum10)

 

 

Toulmhuoflouwkhyucolm4+cohmln

 

 

ToulmbaoflMMc-Iuwhichm
MMGMwu-m(m7(m

Wat-3’ Gnu-Dan

mummummmhumam.mmmw.
01.

02.
Q3.

Ammwmm'WMh 199.
Chin“ with medical mm mlv

durumbnhalfin(mnm7bnm)
Wmdmmhlwvhichmmw
WWW'WMp-fihlm:

Medic-lean

Wapbabacfilm

Mum-drug-

Mmumammmmmmummumwywm
my.Unwinhbaefitwmbmofiudmhnmkmhfiafiwduwmmm.(Cinch

tuna-chum.)
N. I. "my I.
(0!) (aha-:23! (mm nouns
eta-pupa) am; of”)

hmmm .................. 1 2 3 4
2.?1idfidlave ........................... l 2 3 4
3.81mmdiabilitybaefiu ............... l 2 3 4
«Uganda-hummus ............... l 2 3 4
S.Peanonorruimbacfn .............. l 2 3 4
6. Wopymwm

wmfibmmm l 2 3 4
lacuna-commuting” _

mm'wm

tsunami-7) ........................ l 2 3 4
tawnyeanhacem ............... l 2 3 4
1me ................. I 2 3 4
lawn-veerdfldmm ......... l 2 3 4
11.3mb-workm ................... l 2 3 4
11me ................. l 2 3 4
l3.fiofixdmhgotmmmgphn .......... 1 2 3 4
“.mpyforindiwm ....... l 2 3 4

113

Al
(my
nos
5

MUM“

M

MMMMMMMM

Hmfiomu-ymapflyywmfluuhbbnm.m.
mmmfiomumfifivfllwunwmmmfim
man.uummmwmuw Yumdhamaham
h..wilhcmd-duhlhom

 

Mmhmm

0|”lhutmfihWI-d

114

APPENDIX B

Eight Factors from the Hunt, Habeck, VanTol, & Scully (1993) Study

 

115

Table B1

 

l.e - 0 «.10 ‘11 r h M. to. U ui' .’ ' ‘ 0.1.110.
Mammalian
Factor

Items

 

Factor 1: People Oriented Culture

1. The company demonstrates concern about retaining and
developing personnel through its human resources policies and
programs. (CE02)

2. Job satisfaction among employees at this company is high.
(CEO3)

3. Working relationships are collaborative and cooperative in
this company. (CEO4)

4. There is a high level of trust in the employee/employer
relationship at this company. (CEOS)

5. Skills in team building, coaching, problem solving, and
communication are important factors in the selection of
supervisors and managers at this company. (CEO6)

6. Supervisors and managers are trained in interpersonal skills
such as effective communication and conflict management
(CE07)

7. Employees are formally included in the company's goal
setting and planning process. (CElO)

8. The company achieves open communications where
employees feel free to raise issues and concerns, or make
suggestions. (CEll)

9. The company shares information with employees about the
financial status and productivity needs of the company. (CE12)
10. Management seeks and considers employee input in
company decisions. (CE13)

11. Employee involvement programs, such as quality circles
and labor-management participation teams, are used to generate
employee participation in company operations. (CE14)

12. Workers have some control over the work process and
productivity demands. (CEIS)

Factor 2: Active Safety Leadership

1. Top management provides leadership and actively
participates in managing the safety process. (MCOl)

116

Table Bl (cont'd)

 

Factor
Items

2. Top management supports the safety program by attending
safety meetings and training sessions. (MC02)

3. Management allocates staff time of specific individual(s) for
safety responsibilities. (MCO4)

4. The safety manager receives support from top management.
(MCOS)

5. Management has direct knowledge of the potential hazards
in the workplace. (MCO6)

6. The company commits funds to address unsafe conditions
and equipment. (MC08)

7. The company strives for continuous improvement in safety
performance. (MCO9)

8. Supervisors have established goals for safety and received
regular feedback on their performance. (SAO4)

9. Safety performance is evaluated as part of supervisor's
performance appraisal. (SA05)

10. Meaningful safety audits involving supervisors, line
employees, and senior management are conducted at regular
intervals. (SAO8)

11. The company identifies specific jobs and departments with
high accident incidence and lost work time. (SAO9)

12. The company uses occupational health and accident data to
analyze patterns and trends that indicate risk situations. (SAIO)
13. The safety program or committee has the responsibility,
authority and resources to identify and address safety
problems. (8102)

Factor 3: Safety Diligence

l. Violating safety rules results in disciplinary action. (SAOZ)
2. Supervisors complete accident records promptly. (SA06)

3. Supervisors document even minor accidents and violations
for review and consideration. (SAO7)

4. Supervisors confront and correct unsafe behavior and
hazards when they occur. (SIO7)

5. Identified hazards are corrected on a timely basis. ($110)
6. Accident records are complete, identifying causes and
including recommendations for corrective action. ($111)

7. The company achieves excellent housekeeping. (PWEOI)

117

Table Bl (cont'd)

 

Factor
Items

8. Equipment is well maintained. (PWE02)

9. Workers use personal protective equipment where indicated.
(PWEO3)

10. Safety guards and equipment are used in hazardous
operations. (PWEO4)

11. Existing equipment and tools at this plant have been
modified to minimize safety hazards. (PWE06)

12. Safety is considered equally with production and quality F
goals in management thinking and plant operations. (MCIO)
13. Someone capable of handling work related disability claims
is accessible to employees during all working hours. (DCMOl)

Factor 4: Disability Case Monitoring

 

1. Disability claims are evaluated early and accurately to
determine their validity. (DCM02)

2. Employees with continuing disability are reevaluated
through an assessment of their medical recovery and potential
for returning to work. (DCMO6)

3. Duration of disability is evaluated to identify claims needing
case management and rehabilitation services. (DCMO7)

4. Rehabilitation professionals are used to evaluate work
capacity and develop individualized rehabilitation plans when
injured workers are unable to resume employment. (DCM08)

5. Responsibility for disability claim management and return-
to-work coordination is assigned to a specific individual in the
company. (DCMll)

6. The treating physician is asked to identify worker
restrictions and capacities as well as a target date for return-
to-work. (D105)

7. The company monitors employees off work due to disability
and their projected return-to-work date. (DCMO4)

8. When the company refers for professional case management
or rehabilitation services, they still maintain contact with the
employee and monitor the return-to-work process. (DCMO9)

9. Claim management is well coordinated from initial injury to
claim resolution. (DCM12)

10. The company maintains regular communication with the
injured employee's attending physician. (D106)

118

Table Bl (cont'd)

 

Factor
Items

Factor 5: Proactive Return-to-Work Program

1. Follow-up contacts with disabled workers are made at

regular intervals by a company representative according to a

predetermined plan. (D104)

2. The company maintains a detailed inventory that quantifies

the physical demands of its jobs. (D107) '—
3. The company develops alternative placement options and
modified job duties to return disabled employees to work.
(D108)

4. The company uses resources such as assistive devices and
flexible work scheduling to facilitate placement of restricted
workers. (D109)

5. Assistance is provided to supervisors to make job
accommodations or purchase special services needed to assist
return-to-work. (D110)

6. Follow-up contact is made with the employee and
supervisor after successful return-to-work to deal with any
needed adjustments. (D112)

7. Return—to-work assistance is clearly organized with assigned
responsibilities. (D113)

8. There is cooperation and coordination among departments in
efforts to return injured employees to work. (D114)

9. Top management is committed to maintaining workers in
employment when injuries or disabilities occur. (MCll)

10. The company monitors employees off work due to
disability and their projected return-to-work date. (DCM04)

11. When the company refers for professional case management
or rehabilitation services, they still maintain contact with the
employee and monitor the return-to-work process. (DCM09)

12. Injured employees are contacted by a designated person
within the company immediately following medical treatment.
(D103)

13. The company maintains regular communication with the
injured employee's attending physician. (D106)

14. When an injured worker is unable to resume prior duties
the company provides job retraining for reassignment in a
productive capacity. (D111)

"L'li‘. ~."l

_-_IM‘K-.' -‘l‘.a
‘.

119

Table Bl (cont'd)

 

Factor
Items

Factor 6: Wellness Orientation

1. The company commits resources to support health
promotion or wellness programs. (ERP06)

2. Top management supports and participates in health
promotion (wellness) activities. (ERP07)

3. Employees are provided with personal data about their
specific health risk factors. (ERP08)

Factor 7: Ergonomic Solutions

1. Job are modified to keep heavy and repetitive lifting to a
minimum. (PWE07)

2. Strategies are used to reduced repetitive movements.
(PWE08)

3. Ergonomic strategies are used to improve workstation
design and work flow. (PWE09)

4. Position rotation or job enlargement is used where jobs
cannot be further ergonomically corrected. (PWEIO)

Factor 8: Safety Training

1. Employees are informed about possible hazards of their jobs
and are trained in safe work practices for their jobs. (S103)

2. New and transferred employees are given training regarding
specific hazards for their particular job before being placed on
the job. (8104)

3. Temporary or temporarily assigned employees are given
training on-site before being placed on a job or working with
new equipment. ($105)

4. Supervisors are informed about possible hazards and trained
in safe work practices for the jobs they supervise. (8106)

120

APPENDIX C

Item Means and Standard Deviations

121

 

Table C1

hmMmundfitaudarLDmm

 

 

Variable Mean Std Dev N

MC01 3.69 1.11 220
MC02 3.35 1.32 220
MC03 4.28 .83 220
MC04 4.22 .94 219
MC05 4.27 1.03 220
MC06 4.12 .73 220
MC07 4.08 1.09 220
MC08 4.17 .92 220
MC09 4.35 .85 220
MCIO 3.90 1.01 220
MC11 4.16 .89 220
SA01 2.84 1.07 220
SA02 3.28 1.14 219
SA03 4.39 .80 220
SA04 2.95 1.31 220
SAOS 2.95 1.41 216
SA06 3.91 .81 220
SA07 3.37 1.12 219
SA08 3.35 1.32 220
SA09 3.75 1.22 220
SA10 3.56 1.30 219
S101 3.66 1.17 216
S102 4.15 1.10 220
8103 4.10 .78 219
8104 3.79 .93 220
8105 3.62 1.05 218
S106 3.89 .93 218
8107 3.67 .85 219
8108 2.67 1.24 219
S109 4.00 1.20 218
8110 4.11 .72 219
S111 4.22 .80 219
8112 3.58 1.11 217
PWEOI 4.02 .79 218
PWE02 4.13 .59 218
PWE03 4.35 .62 218

Table C1 (cont'd)

 

Variable Mean Std Dev N

PWE04 4.58 .58 216
PWE05 4.21 .93 217
PWE06 4.23 .74 216
PWE07 3.67 .93 218
PWE08 3.38 1.04 216
PWE09 3.25 1.07 216
PWEIO 2.91 1.29 215
DCMOI 4.00 1.07 219
DCM02 4.44 .68 218
DCM03 4.41 .63 218
DCM04 4.44 .86 218
DCM05 1.83 1.25 219
DCM06 4.33 .86 219
DCM07 4.32 .91 219
DCM08 4.06 1.10 215
DCM09 4.13 1.01 216
DCM10 3.19 1.35 217
DCMll 4.47 .97 218
DCM12 4.20 .82 219
DCM13 1.64 1.30 217
D101 3.24 1.15 218
D102 3.41 1.28 218
D103 3.64 1.25 219
D104 3.36 1.34 219
D105 4.67 .58 219
D106 4.02 .97 219
D107 2.97 1.39 218
D108 3.87 1.09 219
D109 3.21 1.35 219
D110 3.03 1.31 215
D111 3.28 1.24 214
D112 3.48 1.19 219
D113 3.46 1.21 217
D114 3.61 1.10 219
ERPOI 2.76 1.65 218
ERP02 2.64 1.50 219
ERP03 4.58 .74 217
ERP04 3.25 1.18 218

123

.r—n—n- .1“1)"-Ir-“4-

Table C1 (cont'd)

 

Variable Mean Std Dev N

ERPOS 3.51 1.56 218
ERP06 3.24 1.51 219
ERP07 3.18 1.45 219
ERP08 2.88 1.52 218
ERP09 3.92 1.73 218
ERP10 2.72 1.77 218
CE01 3.11 1.11 216
CE02 3.83 .90 218
CE03 3.53 .79 219
CE04 3.68 .70 219
CE05 3.40 .84 219
CE06 3.85 .91 219
CE07 3.50 1.02 219
CE08 4.15 .88 219
CE09 3.74 1.08 219
CE10 2.72 1.17 219
CE11 3.80 1.00 219
CE12 3.63 1.13 219
CE13 3.22 1.04 218
CE14 3.20 1.29 219
CE15 3.10 1.08 219

 

Note. The authors changed item SA11 to DCM13 to more accurately
reflect the content of the item. Number of valid observations
(listwise) = 183.

124

APPENDIX D

Rotated Factor Matrix

125

Tabk:1)1

 

 

3 .. " '10151101°u, 1'20 .0“. ' 0.‘ 011v
Iteni Factor 1 Factor 2 Factor 3 Factor 4
S107 .65540 .22893 .11685 .05052
SA07 .59885 .05715 .06329 .16018
PVVEOZ .57367 .1555] .08255 .16934
$106 .56928 .24889 .23198 .14259
$110 .56632 .17712 .0588] .31126
PVVE07 .56488 .19128 .13576 .02515
PVVE06 .55696 .03036 .06323 .19153
SA06 .54137 'a02155 .06284 .11408
PVVE08 .53740 .25899 .23020 .11500
5105 .53409 .19670 .06998 .00668
PVVE03 .53070 .01892 .14777 .13039
thIO .51896 .26277 -a08135 .44207
$111 .51735 .08348 .18013 .08248
PVVE09 .51478 .23059 .23979 .18235
PVVE04 .51194 .0412] .07983 .20106
PVVEOS .50855 .27157 .03214 .18696
PVVEOI .48515 .20376 .01794 .12243
$103 .46924 .32779 .15413 .13334
8109 .46015 .24839 .06732 a00284
S104 .45920 .18779 .19620 .13240'
PVVEIO .4497] .10515 .2855] .09679
$108 .43199 .35525 .08532 .09125
$112 ..41738 .30177 .19429 .27903
SA02 .38927 .04422 «a09100 .07224
DCMO] .36699 -.08716 .27499 -.00413
SA03 .36623 .06648 .32384 .2467]
thII .32103 .28428 .22885 .16739
DCMOS .25036 .24197 .11514 .18434
I)Chdl3 .0817] .05943 .01175 .0435]
CEH3 .18723 .67282 .00103 .2013]
CEJS .27245 .65732 .08170 .06794
CEJO .25389 .62462 .04674 .13522

126

 

Table D1 (cont'd)

 

 

Item Factor 1 Factor 2 Factor 3 Factor 4
CE12 .10389 .61948 .07317 .18750
CE11 .27674 .60521 .12002 .25578
CE05 .34745 .60269 .01188 .12035
CE03 .20960 .60011 .10241 .09894
CE04 .29292 .59726 .03346 .04144
CE14 .10812 .56383 .14041 .12533
ERP07 -.04773 .55194 .15156 .19717
ERPO6 -.04081 .53648 .15786 .13711
CE06 .31542 .53219 .12565 .28439
CE01 .17782 .51319 .10645 .16023
CE02 .23933 .50040 .14305 .12319
ERP08 -.03805 .48492 .18030 .13146
CE07 .36079 .48342 .18068 .22355
SA01 .29848 .42457 .11341 .19540
ERP04 .16728 .38493 .27530 .15716
CE09 .21010 .36705 .11257 .35223
ERP10 .08440 .31391 .14135 .10614
ERP02 .11478 .26985 .18879 .20390
ERP09 -.08158 .17756 .03498 .14497
D113 .12382 .30698 .71311 -.02520
DCM07 -.06813 -.09713 .69831 .22861
DCM12 .18020 .08812 .69767 .09445
DCM04 -.02925 .04695 .68387 .02970
DCM09 .08727 .10596 .67558 -.01165
DCM06 -.02218 -.08353 .66405 .17822
D106 .01785 .16910 .66387 .14238
DCMll .19357 .04074 .62370 .01595
D103 .03557 .15210 .57702 .13934
D112 .20181 .19162 .56894 -.05634
D105 .11852 .04085 .55714 .13437
D104 .02356 .34095 .55483 .03908
D114 .28291 .28783 .53129 .19526
D110 .18239 .42635 .51354 -.03159
D109 .20485 .32889 .51219 .11743
D108 .10196 .23134 .50880 -.02065
D111 .19377 .24777 .50052 .01659
DCM08 -.02346 -.08028 .48437 .17868

Table D1 (cont'd)

 

 

Item Factor 1 Factor 2 Factor 3 Factor 4
DCMIO .12595 .14171 .47323 .07834
DCM02 .26337 -.10763 .46415 .15555
D102 .10913 .06291 .44193 .31667
ERP03 .19475 .19806 .41517 .1352]
D107 .01049 .27296 .40476 .02797
DIO] .30067 .29571 .34095 .24866
DCM03 .21145 -.05868 .27781 .04992
MC04 .16501 .13831 .09506 .66450
MC09 .28428 .24405 .03126 .63070
MC01 .32780 .30276 .01407 .62972
MC05 .28948 .18899 .08428 .55898
MC02 .24958 .26125 .05573 .54776
S102 .22593 .31717 .16442 .50050
MC08 .29727 .34708 .03628 .49312
SA09 .26925 .17397 .24224 .44438
MC06 .36939 .18900 .04199 .44072
SA10 .16376 .25972 .30521 .41424
MC07 .32521 .17172 .15253 .40871
SAOS .28748 .27991 .00498 .40720
SA04 .36665 .37448 .07267 .40315
S101 .1840] .38100 .08612 .38518
CE08 .13772 .18417 .35023 .37424
MC03 .29107 -.06216 .06823 .34069
SA08 .32886 .26668 .08902 .33643
ERPOS -.14772 .19003 .14377 .24387
ERP01 -.04724 .11889 .18612 .22565

128

APPENDIX E

Regression Models

129

Table E]

II"! 9 LI "\°' 015'. 0 . .0 l

 

 

Variable B SE B Beta
MW
Small (100 - 249) -.2091 .1305 -.1273
Large (over 500) -.2554 .1463 -.1470p
SIC 25 - Furniture -.2140 .2822 -.0628
SIC 30 - Rubber & Plastics -.1226 .2058 -.0595
SIC 34 - Fabricated Metals .1324 .1898 .0756
SIC 35 - Non-Electric. Mach. -.2402 .2311 -.0991
SIC 37 - Transport. Equip. .0000 .2164 .0000
SIC 80 - Health Services -1.2660 .2551 -.4975**
QQnILQLJLaLiaths;
Percent Salaried -.0988 .0960 -.0784
Tenure < 1 Year .1832 .0542 .2713"
Rotating Shifts -.l452 .2073 -.0506
Multiple Plants -.2354 .1185 -.1439*
Safety Standards -.0940 .1367 -.0555
Unionized .1774 .1270 .1101
Loss Control Consult. -.0817 .1339 -.0460
Independentlatiam;
Safety Diligence &

Hazard Reduction -.0643 .0570 -.0824

 

N91; R2 = .30; Adjusted R2 = .22; N = 163', p = significant at .10
level, * = significant at .05 level, ** = significant at .01 level.

130

Table E2

 

 

 

"it. o u. 0‘ "'1" ' ‘ « ' ..l-J‘ h
Variable SE B Beta
StructuraLlaLiables;
Small (100 - 249) .0790 .1507 .0444 _
Large (over 500) -.4957 .1680 -.2637**
SIC 25 - Furniture .2164 .3242 .0588
SIC 3O - Rubber & Plastics .1453 .2368 .0652
SIC 34 - Fabricated Metals .4073 .2187 .2152p
SIC 35 - Non-Electric. Mach. .1900 .2664 .0709
SIC 37 - Transport. Equip. .2895 .2486 .1221
SIC 80 - Health Services -.2002 .2930 -.0728
L

Percent Salaried .0071 .1116 .0051
Tenure < 1 Year .1531 .0624 2100*
Rotating Shifts .2007 .2382 .0647
Multiple Plants -.0527 .1376 -.0294
Safety Standards .1370 .1570 -.0748
Unionized .2176 .1465 .1249
Loss Control Consult. -.1079 .1523 -.0566
lndmndentJLaLiahm
Safety Diligence &

Hazard Reduction -.2195 .0665 -.2555**

 

121.019.. R2 = .21; Adjusted R2
level, "' = significant at .05 level, **

131

.13; N = 162; p = significant at .10

= significant at .01 level.

Table E3

 

 

1111- 0 u 0 1 ° N _-.l 1 __ iool

" ‘u oquoo ' o g A. ' out 0! -0
Km
Variable SE B Beta
StructleaLiahles;
Small (100 - 249) .1005 .1191 -.0699
Large (over 500) .0294 .1302 -.0192
SIC 25 - Furniture .3572 .2576 .1253
SIC 30 - Rubber & Plastics .1427 .2043 .0757
SIC 34 - Fabricated Metals .4960 .1869 .3166“
SIC 35 - Non-Electric. Mach. .2664 .2187 .1326
SIC 37 - Transport. Equip. .2142 .2047 .1137
SIC 80 - Health Services .1267 .2357 -.0608
Control Variables;
Percent Salaried .1094 .0857 -.1005
Tenure < 1 Year .1120 .0487 .1895*
Rotating Shifts .1827 .1841 .0740
Multiple Plants .1504 .1079 -.1033
Safety Standards .0194 .1259 .0132
Unionized .2447 .1144 .1753 *
Loss Control Consult. .1277 .1194 .0809
mm
Safety Diligence &

Hazard Reduction .1787 .0528 -.2532**

 

N911. R2 = .19; Adjusted R2 =

132

.11; N = 178', p = significant at .10
level, * = significant at .05 level, **

= significant at .01 level.

Table E4

 

 

"no. 0 U 0 1° ”.1 .,.. l‘

D 0. (1.1.1 It" you ' c .0 o D: {.
Variable B SE B Beta
81W
Small (100 - 249) -.0420 .1454 -.0239
Large (over 500) -.4661 .1654 -.2490**
SIC 25 - Furniture -.0054 .3156 -.0014
SIC 30 - Rubber & Plastics -.1009 .2281 -.0456
SIC 34 - Fabricated Metals .2777 .2069 .1493
SIC 35 - Non-Electric. Mach. .1878 .2559 .0694
SIC 37 - Transport. Equip. .1545 .2322 .0679
SIC 80 - Health Services -.3209 .2561 -.1215
mm
Multiple Plants -.0275 .1332 -.0156 .
Unionized .2378 .1340 .1377p
Group Self-Insurance .3493 .1456 .2101“
Individual Self-Insurance .2125 .1950 .0873
Hourly Wage -.8978 .2790 -.2830**
Winkle;
Reducing Work Disability

and Promoting RTW -.1308 .0616 -.1606*

 

18.0.1.2. R2 = .20; Adjusted R2 = .13; N = 170; p = significant at .10
level, * = significant at .05 level, ** = significant at .01 level.

133

Table E5

 

 

 

"no 0 u. 0 1° 01 -_ {0.1 1‘
D . u_.¢.° "a you ' ‘0 1° 0 D, "
Variable B SE B Beta
S_tr_u_c_tural_1ariahles;
Small (100 - 249) -.0109 .1032 -.0092
Large (over 500) -.0292 .1167 -.0236
SIC 25 - Furniture .0624 .2215 .0260
SIC 30 - Rubber & Plastics .0645 .1665 .0431
SIC 34 - Fabricated Metals .0716 .1522 .0568
SIC 35 - Non-Electric. Mach. -.0269 .1806 -.0157 -
SIC 37 - Transport. Equip. .0394 .1664 .0263 i
SIC 80 - Health Services -.0161 .1775 -.0098 i
Wabbit;
Multiple Plants .1559 .0943 .1335p
Unionized -.0442 .0939 -.0387
Group Self-Insurance .0528 .1040 .0478
Individual Self-Insurance -.1837 .1392 -.1115
Hourly Wage .1684 .2004 .0799
IndependenLlaLiahls;
Reducing Work Disability
and Promoting RTW .0062 .0437 .0114

 

N915; R2 = .05; Adjusted R2

= -.03; N = 178; p = significant at .10
level, " = significant at .05 level, ** = significant at .01 level.

134

Table E6

 

 

n". 0 u n‘ i" 'nai_._1 o 1 in h
U,sle°'ll'l UN. "1 0° 10 ' out H
W
Variable B SE B Beta
W
Small (100 - 249) -.2073 .1167 -.1467p
Large (over 500) -.0886 .1333 -.0588
SIC 25 - Furniture .1538 .2481 .0562
SIC 30 - Rubber & Plastics -.0185 .1985 -.0100
SIC 34 - Fabricated Metals .3654 .1804 .2381*
SIC 35 - Non-Electric. Mach. .2084 .2088 - .1021
SIC 37 - Transport. Equip. .1568 .1923 .0881
SIC 80 - Health Services -.1592 .2094 -.0794
Multiple Plants -.0810 .1062 -.0567
Unionized .2226 .1073 .1624“
Group Self-Insurance .2573 .1179 .1911*
Individual Self-Insurance .1113 .1589 .0545
Hourly Wage -.5929 .2228 -.2342**
Indemndentlariahlm
Reducing Work Disability

and Promoting RTW -.0914 .0483 .1388p

 

131521;. R2 = .14; Adjusted R2 = .07; N = 186; p = significant at .10
level, * = significant at .05 level, ** = significant at .01 level.

135

Table E7

un- o u o-‘ l'°' or 5.1-. ' 0 V .0‘ .I‘t

 

 

Variable B SE B Beta
WW

Small (100 - 249) -.0456 .1770 -.0222
Large (over 500) -.4562 .1973 -.2133*
SIC 25 - Furniture .3542 .3811 .0838
SIC 30 - Rubber & Plastics .1484 .2740 .0589
SIC 34 - Fabricated Metals .5288 .2498 .2421*
SIC 35 - Non-Electric. Mach. .3597 .3014 .1169
SIC 37 - Transport. Equip. .3313 .2768 .1258
SIC 80 - Health Services -.2425 .3052 -.0806
mum

Tenure < 1 Year .1344 .0761 .1609p
Multiple Plants .0163 .1598 .0080
Unionized .2089 .1663 .1049
Hourly Wage -.5849 .3600 -.1591
Individual Self-Insurance .4544 .1637 .2371”
We;

Valuing Human Resources -.1121 .0851 -.1118

Top Management Support and
Continuous Improvement
of Safety -.0512 .0780 -.0545

 

m R2 = .17; Adjusted R2 = .09, N = 165; p = significant at .10
level, * = significant at .05 level, ** = significant at .01 level.

136

Table E8

. A .. . I I I . O -
lllls . .U. ' \ ° .liI' 0 1.1.1! I

 

 

 

u=i-°'-. UH‘ "0. 0° 10." 0110‘le0l.ell‘l ia.‘
Variable B SE B Beta
WW1;

Small (100 - 249) -.4222 .2237 -.1660p
Large (over 500) -.3378 .2449 -.1304 ‘
SIC 25 - Furniture .1285 .4923 .0252
SIC 30 - Rubber & Plastics -.2680 .3715 -.0843
SIC 34 - Fabricated Metals .2398 .3333 .0886
SIC 35 - Non-Electric. Mach. -.1230 .3750 -.0354
SIC 37 - Transport. Equip. .2384 .3539 .0761
SIC 80 - Health Services -.4490 .3936 -.1265
Wilkins;

Tenure < 1 Year .1109 .0962 .1116
Multiple Plants -.1434 .1987 -.0581
Unionized .5819 .2091 .2433"
Hourly Wage .1180 .4371 .0266
Individual Self-Insurance .2409 .2052 .1029
W

Valuing Human Resources -.2366 .1064 -.1908p

Top Management Support and
Continuous Improvement
of Safety .0630 .1030 .0522

 

N919. R2 = .16; Adjusted R2 = .08; N = 160; p = significant at .10
level, * = significant at .05 level, ** = significant at .01 level.

137

Table E9

"no 0 .u. 0.‘ {‘9‘ 101A1,o__ 0 ._ oo‘. 1 1'

 

 

Variable B SE B Beta
81W
Small (100 - 249) -.0560 .1721 -.0273
Large (over 500) -.4898 .1929 -.2290*
SIC 25 - Furniture .4332 .3711 .1025
SIC 30 - Rubber & Plastics .0543 .2617 .0216
SIC 34 - Fabricated Metals .4582 .2425 .2098p
SIC 35 - Non-Electric. Mach. .3011 .2948 .0978
SIC 37 - Transport. Equip. .2403 .2694 .0912
SIC 80 - Health Services -.2802 .2975 -.0932
Wrinkles;
Tenure < 1 Year .1428 .0742 .1709p
Multiple Plants -.0361 .1562 -.0177
Unionized .1627 .1634 .0817
Hourly Wage -.6936 .3458 -.1887*
Individual Self-Insurance .5018 .1612 .2618”
Indmndentlatiahlc;
Safety Diligence &

Hazard Prevention -.2071 .0834 -.2047*
Reducing Work Disability &

Promoting RTW -.0917 .0768 -.0985

 

N91_e___ R2 = .22; Adjusted R2 = .14; N = 165; p = significant at .10
level, * = significant at .05 level, ** = significant at .01 level.

138

Table E10

1111-, 0. ' ‘ '1‘! ‘ 1H 5.1-. 1 Jr 11.1.. 'n .l'

 

 

Variable B SE B Beta
W

Small (100 - 249) -.0378 .1754 -.0184
Large (over 500) -.4971 .1941 -.2325*
SIC 25 - Furniture .4183 .3740 .0990
SIC 30 - Rubber & Plastics .0391 .2716 .0155
SIC 34 - Fabricated Metals .4575 .2474 .2094p
SIC 35 - Non-Electric. Mach. .3022 .2972 .0982
SIC 37 - Transport. Equip. .2350 .2746 .0893
SIC 80 - Health Services -.2940 .2998 -.0978
W

Tenure < 1 Year .1484 .0750 .1777“
Multiple Plants -.0302 .1575 -.0148
Unionized .1391 .1670 .0698
Hourly Wage -.6794 .3544 -.1848p
Individual Self-Insurance .5099 .1624 .2661"
Independexfliariahls;

Safety Diligence &

Hazard Prevention -.2248 .0940 -.2222*
Valuing Human Resources -.0315 .0882 -.0314
Reducing Work Disability &

Promoting RTW -.0972 .0812 -.1044
Top Management Support and

Continuous Improvement

of Safety .0596 .0854 .0634

 

N919. R2 = .22; Adjusted R2

139

.13; N = 165; p = significant at .10
level, * = significant at .05 level, **

= significant at .01 level.

REFERENCES

140

REFERENCES

Akabas, S. H, & Gates, L. B (1990). Organizational commitment:
The key to successful implementation of disability management.

Amencanjchalumanon, 1.6(3) 9-13, 32.

Askey, D. E. (1988). Effective worker compensation loss control7

strategies. WW 2(1)

Battle, M. C. & Bigos, S. J. (1991). Industrial back pain complaints:

A broader perspective WWW
22(2), 273- 282

Berkowitz, M. (1988). Forestalling disincentives to return to work.

BusinesLanLHealth, 2(4) 30- 32

Bigos, S. 1., Spengler, D. M., Martin, N. A., Zeh, J., Fishers, L., &
Nachemson, A. (1986). Back injuries in industry: A
retrospective study 111. Employee-related factors. $121119, 1_1_(3),
252-256.

Boschen, K. A. (1989). Early intervention in vocational rehabilitation.

Wm 3.2(3) 254- 265

Carbine, M. E., & Schwartz, G. E. (1987).S1_[919_g_e_s_9_£_m99_agm_g
W. Washington, D. C.: Washington Business Group

on Health/Institute for Rehabilitation and Disability
Management.

Carbine, M. E., Schwartz, G. E., & Watson, S. D. (1989). M1111
. . l . E I 19512 .
Washington, DC: Washington Business Group on
Health/Institute for Rehabilitation and Disability.

Chaffin, D. B. (1987) The role of biomechanics in preventing
occupational injury Hummus, 1.112(6) 599- 601

Clark, L. A., & Watson, D (1995). Constructing validity. Basic

issues in objective scale development. Emma
Assessment 1(3) 309- 319

141

DeVellis, R. F. (1991). WWW
Newbury Park, CA: Sage Publications, Inc.

Elkiss, H. (1987) Microcomputers in the workplace can improve

injury, illness prevention Q::11a1mna1_11:al111_&._8_a£e11,5_§.(5),
34- 37

Fordyce, W. 13., (Ed), (1995). W Seattle,

WA: International Association for the Study of Pain.

Frost, C. F- (1989, August). WW.

Paper presented at the meeting of the Academy of Management,
Washington, DC.

Gaertner, G., Newman, P. ,Perry, S., Fisher, G. P., & Whitehead, K.

(1987 July, 7- 8) W

W. Paper presented at the Bureau
of Mines Technology Transfer Seminar, _ Pettsburgh, PA.

Galvin, D. E. (1986) Health promotion, disability management, and

rehabilitation in the workplace. Wm, 41(9-
10), 218- 223.

Gapen, P (1990, October). Whittling down workers' comp costs.
3nsin:ss_and_1:1:a1111, 35- 48

Gice, J H., & Tompkins, K. (1987). Evaluation ofa formal return-
to- work program in a hospital setting. Legal Insight, 2(3), 9-
12.

Glass, G. V. & Hopkins, K. D. (1984). W
W Englewood Cliffs NJ

Prentice- Hall, Inc.

Gross, C. M. (1988). Ergonomic workplace assessments are the first

step in injury treatments WW, 51(5)
16- 19

Habeck, R. V., Leahy, M. 1., Hunt, H. A, Chan, F. & Welch, E. M.
(1991). Employer factors related to workers' compensation

claims and disability management. Wm
3111:1111, 311(3) 210- 226

142

Hill, 8., & Gipson, K. (1990, May). How Weyerhaeuser improved its
workers' compensation program. W, 38- 44.

Hunt, H A., Habeck, R V., VanTol, B. C., & Scully, S. M. (1993).

121W: (Technical
Report #93- 004). Kalamazoo, MI: W. E. Upjohn Institute for

Employment Research.

Jacobs, H. C., & Nieburg, J. T. (1989). Accident investigations.
WM, BA, 13- 16

Kasdan, A. S., & McElwain, A. P. (1989). Return- to- work programs

following occupational hand injuries. Q9_c_u_p_a119na_l_M_e_di_c_in_e_;_
$1919 9£1h9 An 81:11:11.3, 4(3), 539- 545.

Kavianian, H. R., Rao, J. K., & Sanchez, V. F. (1989). Management
thinking and decision making styles: Their effects on
occupational safety and environmental health. mm

mm, 34(9). 24-27.

Kline, P. (1994). W New York:
Routledge.

Koch, C. (1988, May). A task of orchestration. W,
23-25.

LeClair, S. W., & Mitchell, K. (1989, May). MM

010: '1 1 ’00"'. .11 o 1,}.-

WW Columbus, OH: International

Center for Industry, Labor, and Rehabilitation.

Lewin, D., & Schecter, S. (1991). Four factors lower disability rates.
EemmneLMLnal, 19(5), 99-103-

Lewin, D. & Schecter, S. M. (1990).Ih9_¢e19_r_m1m1_s_a_n_d_
391W (Unpublished monograph)

New York: Human Resource Health Institute, Inc.

Lewis-Beck, M. S. (1980). WM
Beverly Hills, CA: Sage Publications, Inc.

Lutness, J. (1987). Measuring up: Assessing safety with climate

surveys WW, 16(2) 20- 26

143

Marcus, A. A. (1988). Implementing externally inducted innovations:
A comparison or rule-bound and autonomous approaches.

MAW, 31(2), 235-256.

McAfee, R. B., & Winn, A. R. (1989). The use of
incentives/feedback to enhance work place safety: A critique

of the literature. MmaLQLSamLRmmh, 23(1), 7-19-

McDonald, M. (1990, May). You can control your disability costs.
W311, 21- 36

Michigan Department of Labor. (1990). W, 1990.
Lansing, MI: Author.

Norusis, M. J. (1994). SPSS Rmfgssmnal 819111111123, 6,1. Chicago, IL:
SPSS Inc.

OSHA (1990).Emono.mi:s_1mtam_mmm:m:n1_suid:lin:s_fo_1

W (OSHA Publication No. 3121). Washington,
D. C. U. S. Government Printing Office.

Pedhazur, E. J., & Schmelkin, L. P. (1991). MW
Wm Hillsdale, NJ Lawrence

Erlbaum Associates, Publishers.

Phillips, J. F. (1989). What gets measured gets done. Mama],
9113:111, £0), 39-41-

Prevention Leadership Forum. (1987). mm
19119115,. Washington, DC: Washington Business Group on
Health.

Rappaport, W. (1988). Comprehensive 'OH & EIS' systems help firms

combine related data.Q99_u,pa119_na1_H_e_aj1h_&_S_a£911,5_Z(6), 52-
61.

Rosen, R. H. (1936). H:a1m1_c_om1ani:s;___A_h1man_1:so11::s
W. New York: American Management Association.

Schmitt, N. (1996). Uses and abuses of coefficient alpha.
Wm, 3(4) 350- 353

Schwartz, G. E, & Beggelman, M. (1986, May). Assessing the need
for rehabilitation providers. W, 15-16.

144

Schwartz, G E., Watson, S. D, Galvin, D. E., & Lipoff, E. (1989).
DWI: Washington DC:

Washington Business Group on Health/Institute for
Rehabilitation and Disability Management.

Shrey, D. E. (1988). controlling worker injury and disability costs:
Disability management strategies for the employer. 11:19

Want-11,3326

Sims, R. H. (1988). WW
I] [I E' l E l . . I I
W5. (Doctoral dissertation, The Rand Graduate School,

1988). University Microfilms International, Order no. 8909266.

Smith, G. T., & McCarthy, D. M. (1995). Methodological
considerations in the refinement of clinical assessment

instruments. WW, 1(3), 300-308.

Smith M J (1983 Juric) WWW
W95. A grant application submitted to the

Department of Health and Human Services, Minnesota.

Stern, L. (1990, May). Why EAPs are worth the investment. 8115111955
a11_d_H:al111, 14-19.

Susser, P. A. (1989). Growing concern for workers' repetitive motion
injuries has made their prevention OSHA's issue of the nineties.

EmM11m:n1_R:111Ms_I:da1, 1.6, 243-248.

Tate, D. G (1992). Factors influencing injured employees return to

work WWW,23(2),17-
20.

Tate, D. G., Habeck, R. V., & Galvin, D. E. (1986). Disability
management. Origins, concepts, and principles for practice.

1.91mal_of_Ap111:.d_3:hab113a119.n_C.o_uns:ling 11(3) 5-12

Victor, R B. (1985, May). Work place safety: Present costs, future
trends 3usin:s.s_and_1:1:al111, 7-10

Victor, R. B., Gardner, J. A., Sweeney, D., & Telles, C. A (1992)
W Cambridge, MA: Workers

Compensation Research Institute.

145

Vojtecky, M. A. (1988) Education for job safety and health 111111111
Ed1.c.a1_i_o_n_Q1_a11:_11, L0), 289- 298

Walker, J. M. (1992). Injured worker helplessness: Critical
relationships and systems level approach for intervention.

WWW, 2(4) 201- 209

Wiesel, S. W., Boden, S. D., & Feffer, H. L. (1994). A quality-
based protocol for management of musculosketetal injuries: A

ten year prospective outcome study. WW
W,30.1,164-176

Webb, G. R., Redman, S., Wilkinson, C., & Sanson- Fisher, R. W.
(1989). Filtering effects in reporting work injuries. A9_c_i_d_9111

WWW, 21(2) 115-123

Young, K. (1989, February). Why you can’t afford to overlook safety.
B.usi_n:ss_and_H:a1111,4l-43

146

HICHIGRN sreTE UNIV. LIBRnRIEs
11111111111111111111111111111111111111111111111
31293017141403