- Email: [email protected]
Worksite Health and Safety Climate: Scale Development and Effects of a Health Promotion Intervention
27, 111–119 (1998) PM970253
PREVENTIVE MEDICINE ARTICLE NO.
Worksite Health and Safety Climate: Scale Development and Effects of a Health Promotion Intervention1 Karen Basen-Engquist, Ph.D., M.P.H.,*,2 Karen Suchanek Hudmon, Dr.P.H., R.Ph.,*,† Mary Tripp, M.P.H.,* and Robert Chamberlain, Ph.D.† *Department of Behavioral Science and †Department of Epidemiology, The University of Texas M. D. Anderson Cancer Center, Houston, Texas 77030
Background. Environmental influences on health and health behavior have an important place in research on worksite health promotion. We tested the validity and internal consistency of a new measure of organizational health and safety climate that was used in a large randomized trial of a worksite cancer prevention program (the Working Well Trial). The resulting scales then were applied to assess intervention effects. Method. This study uses data from a subset of 40 worksites in the Working Well Trial. Employees at 20 natural gas pipeline worksites and 20 rural electrical cooperatives completed a cross-sectional questionnaire at baseline and 3-year follow-up. Results. A factor analysis of this self-report instrument produced a two-factor solution. The resulting health and safety climate scales had good internal consistency (Cronbach’s a = 0.74 and 0.82, respectively) and concurrent validity. The health climate scale was correlated more highly with organizational measures that were indicative of a supportive health climate than those indicating supportive safety climate, while the reverse was true of the safety climate scale. Changes in health climate were associated with the number of smoking and smokeless tobacco programs offered at the worksites at the time of the 3-year follow-up (r = 0.46 and 0.42, respectively). The scales were not correlated with most employee health behaviors. The health climate scores increased at intervention worksites, compared with scores at control worksites (F[1,36] = 7.57, P = 0.009). Conclusions. The health and safety climate scales developed for this study provide useful instruments for measuring organizational change related to worksite health promotion activities. The Working Well Intervention resulted in a significant improvement in worksite health climate. © 1998 Academic Press
1 This work was supported by a Cooperative Agreement from the National Cancer Institute, Grant U01 CA51671. 2 To whom reprint requests should be addressed.
Key Words: organizational climate; cancer prevention; worksite health promotion; scale development; validity. INTRODUCTION
Health promotion and organization theories suggest that effective worksite health promotion programs must address environmental factors that influence health behavior. Environmental factors derived from sociobehavioral theoretical frameworks, such as Social Cognitive Theory [1] and Diffusion of Innovations Theory, [2] have been discussed extensively in the health promotion literature. We describe a new measure of an environmental factor often studied in the organizational literature—organizational climate. Traditionally, organizational climate has been studied for its effects on organizational outcomes such as attendance, absenteeism, productivity, performance, and job satisfaction. For this study, organizational climate related to health and safety was investigated as it relates to health promotion at the worksite. Within the framework of a randomized trial of a worksite cancer prevention program, we examined the psychometric properties of an instrument designed to measure the worksite climate regarding health and safety and assessed the effect of the intervention on this climate. This study is unique because it provides data on both the validity and the reliability of the climate measure, in addition to assessing the sensitivity of the measure to climate change resulting from an organizational intervention. The literature suggests a wide range of environmental factors that can affect an organization’s health promotion programming and the participation of its employees. These include organizational management style; [3] industry type; [4,5] organization rank; [5] worksite size; [4,6] organizational structure; [7] degree of support from top management, [6] including incentives or recognition for healthy behaviors or participation in a health promotion program; [8] modeling of healthy behaviors by the CEO; [8] policies reinforcing healthy behaviors; [8,9] employee autonomy and con-
111 0091-7435/98 $25.00 Copyright © 1998 by Academic Press All rights of reproduction in any form reserved.
112
BASEN-ENGQUIST ET AL.
trol over work activities; [10] and subcultures within the organization [10]. One environmental factor in particular, organizational climate, is the focus of this study. Organizational climate has been identified as an important environmental influence on employee participation in worksite health promotion programs and on the effectiveness of such programs [7,8,11–13]. Although organizational climate has been studied in educational, business, and health care institutions since the 1950s, there is little consensus among researchers regarding the definition of organizational climate [14,15]. Nevertheless, the conceptual components central to the various views can be summarized. Organizational climate is a set of internal characteristics that distinguishes one organization from another, is experienced by members of the organization, influences their behavior, and is based on their collective perception of the organizational environment [16–19]. Climate is influenced by the physical environment, organizational characteristics such as centralization and formalization, formal and informal relationships, personalities or organization members, and leadership [15,17]. The interactive perspective on climate formation suggests that an organization’s members interact with the situational environment to create climate [16,18]. By responding to a situation or some organizational characteristic according to shared perceptions, individuals in an organization create climate [18]. Along with individual motives and abilities, climate in turn influences behavior [16–19]. This concept is similar to that of reciprocal determinism in Social Cognitive Theory (i.e., the idea that environment, behavior, and personal factors mutually influence each other) [1]. For this reason, it is important to address climate when attempting to change the behavior of an organization’s members [15,18]. Studies of health promotion have dealt with the relationship of organizational climate to health behaviors and participation in health promotion programs in the workplace [11–13]. Some researchers have investigated general organizational climate as it is represented in the organizational literature, [12] while others have specifically investigated the health-related dimensions of climate [11,13]. Dimensions of climate such as perceived supervisor support, [12] perceived control over work matters, [12] and worktime flexibility to allow time for personal health needs [11] are associated with employee participation in health promotion programs. Health climate also has been associated with exercise habits, healthy nutrition habits, and smoking status [13]. We identified only four instruments that specifically assess the health and safety climate and its relationship to the implementation of worksite health promotion programs: (1) the Safety Climate Questionnaire, a measure of safety concerns in the workplace; [20,21] (2)
the Organizational Climate Survey, a measure of general organizational climate, with added items related to health promotion; [11,22] (3) the AT&T Health Audit, a measure of general climate, health climate, and social support systems, but created to measure organizational culture; [23] and (4) the Worksite Health Climate Scales (WHCS), a measure of organizational support, interpersonal support, and health norms [13]. Empirical investigation of the psychometric properties of these instruments is limited. Construct validity and reliability testing has been replicated in only a few instances, and rarely are both validity and reliability assessed for a given instrument. A principal-components factor analysis of the Safety Climate Questionnaire, the only safety climate scale located in the literature, originally resulted in an eight-factor solution [20]. In a second study this structure could not be replicated, but the data did support the use of a smaller three-factor model [21]. Psychometric analysis of the existing health climate scales has focused on reliability testing. Factor analysis, which is typically one of the first steps in scale validation, has not been reported for any of these scales. For the health climate subscales of the Organizational Climate Survey, [11] a estimates of internal consistency reliabilities ranged from 0.70 to 0.88, and no further reliability testing has been reported. When the AT&T Health Audit’s test–retest reliability was assessed, the proportion of respondents giving identical answers to single items at the two time periods ranged from 58 to 99%; the proportion of identical responses tended to be somewhat lower, 61 to 78%, for the health norms and values items. Criterion and concurrent validity were supported for the AT&T Health Audit [23]. For the WHCS, a estimates of internal consistency reliability of all 12 subscales of the WHCS ranged from 0.73 to 0.94 in a population of small worksites [13]. Reliability testing of the WHCS at larger worksites yielded similar results, with scale reliability ranging from 0.69 to 0.96 [24]. Low correlations among subscales showed that they are conceptually distinct, although they tap related constructs [13,24]. Differences in health climate among worksites were investigated in a few studies, [13] as were the relationships of health climate, health status, and the performance of health behaviors [13,23]. We could identify no assessments of the effect of a health promotion intervention on worksite health or safety climate, nor do we know of any examinations of the relationships between health and safety climate and variables such as the number of health promotion programs available at the worksite, organizational characteristics, and safety behaviors. For these reasons, the purposes of this study were to (1) estimate the reliability and validity of the health and safety climate scale developed for a randomized trial of a worksite cancer prevention program and (2)
HEALTH AND SAFETY CLIMATE
assess the effect of the worksite cancer prevention intervention on organizational climate of health and safety. Because intervention strategies addressed social and environmental factors associated with the adoption and practice of healthy behaviors, thereby creating a supportive environment for such behaviors, we hypothesized that health climate would improve as a result of the intervention. Because the intervention did not emphasize worksite safety, we expected no change in safety climate. METHOD
Background Data were collected as part of the Working Well Trial, the largest worksite cancer control intervention trial in the United States [25]. The primary hypothesis of the Working Well Trial was that a sustained 2-year cancer control worksite health promotion intervention addressing dietary change and smoking cessation would be more effective than a minimal intervention (consisting of widely available print materials) in achieving both individual and environmental changes. The study was conducted at 114 worksites and coordinated by four study centers. The results presented here derive from the study center at The University of Texas M. D. Anderson Cancer Center (UTMDACC), where investigators were responsible for a subset of 40 worksites. The project was reviewed and approved by the UTMDACC institutional review board. Design The Working Well Trial used a randomized, matched-pair research design in which the worksite was the unit of randomization and analysis. Worksites were randomized after baseline assessment; they were stratified, matched into pairs, and randomly assigned within pairs to intervention or control group. Stratification variables included presence of a cafeteria, worksite size, type of smoking policy, company type, gender distribution, distribution of blue- and white-collar jobs, and response rate to the baseline survey. The intervention group received the program described below. The UTMDACC control group received widely available print materials such as posters and brochures. Employees at each worksite completed a cross-sectional questionnaire at baseline and at a 3-year follow-up. Description of the Sample The worksites studied by the UTMDACC study center included 20 natural gas pipeline worksites and 20 rural electrical cooperatives; each of these sites employed between 72 and 621 employees. The worksites were located primarily in the southeast quadrant of the United States, from Texas eastward to Florida and northward to Nebraska and Colorado. Of the 6,867 em-
113
ployees who responded to the baseline survey at the UTMDACC worksites, the majority were male (80.3%) and white (90.5%). Most of the sample was currently married (81.6%). Ninety-six percent had at least a high school education; 57.3% had education beyond high school. Intervention and control worksites had no detectable demographic differences at baseline or followup [26]. Working Well Intervention The Working Well Intervention, based on individual, organizational, and community organization theories, focused on (1) promoting and building employee awareness of preventive behaviors, (2) action and skills training, and (3) behavior maintenance and relapse prevention. The risk behaviors targeted at the UTMDACC intervention worksites were smoking, smokeless tobacco use, and diets low in fruits, vegetables, and fiber and high in fat. The activities conducted to address these risk behaviors included promotional and awareness materials and activities, such as posters, brochures, and videos; self-assessments that enabled employees to assess their risk behavior or health status and compare their results with a standard; self-help materials that provided step-by-step instructions for helping people build the skills they needed to adopt healthy behaviors; and direct education for smoking cessation or dietary change offered through multisession support groups or classes. The intervention was implemented using a participatory strategy that involved selecting an employee from each intervention worksite to serve as the coordinator for the Working Well program and selecting a committee of employees to serve as the employee advisory board (EAB). Worksite coordinators and EABs were responsible for planning and implementing activities and tailoring them to their worksite. They were given a program manual outlining study protocols and general information about implementing health promotion programs in the workplace. Intervention staff from UTMDACC provided support and technical assistance to the worksite coordinators and EABs. Data Collection The baseline data were collected between September and December 1990, and the follow-up data were collected between September and December 1993. Questionnaires were administered by both worksite personnel and UTMDACC staff members to groups of employees at the worksite. Employees who were absent or could not complete the survey on the scheduled day were given a questionnaire to be completed and mailed to the study center. All employees were recruited to complete the questionnaire. Response rates were 89% at baseline and 86% at follow-up. Organizational interviews with key worksite infor-
114
BASEN-ENGQUIST ET AL.
mants (CEO or highest ranking employee at worksite, personnel director, health and safety director, etc.) were conducted by telephone. UTMDACC staff members conducted baseline interviews from September to December 1990, and the final interviews were conducted from September to December 1993.
peated-measures analysis of variance with one withinworksites factor (time of measurement) and two between-worksites factors (treatment condition and type of industry). The unit of analysis was the worksite for tests of concurrent validity and intervention effects, because health and safety climate is conceptualized as a worksite characteristic.
Measurement and Analysis Procedures RESULTS
A panel of researchers with experience in the study of organizational climate developed a series of 18 items to assess the organizational health and safety climate of the UTMDACC Working Well sites. The items were administered to 6,867 participants. Psychometric analyses to determine whether the items were sufficiently related to justify formation of one or more health and safety climate scales included assessment of item characteristics, interitem correlations and internal consistency, and an exploratory factor analysis to determine the underlying structure of the scales [27,28]. A matrix of interitem correlations was used as input for unweighted least-squares factor extraction with oblique rotation. An oblique rather than orthogonal rotation was selected because we hypothesized that two factors, health climate and safety climate, would emerge, and that they would be related statistically. Because the questionnaire items had not been pretested or used in previous studies, items that did not exhibit desirable psychometric properties were eliminated. A factor loading of 0.40 was the cutoff for retaining an item, although the selection of a factor analysis solution also was based on theory, the communality estimates, the residual matrix, and the percentage of variance explained by the extracted factors. The psychometric results reported here are derived from baseline measures. We also conducted a factor analysis on the scale items from the follow-up data, to determine if the factor analysis solution derived from the baseline data could be replicated. The item analysis, internal consistency, and factor analysis were all calculated using the individual employee as the unit of analysis. To assess the concurrent validity of the climate scales, Pearson correlation coefficients were calculated between worksite means of the scale scores and theoretically related organizational characteristics that were assessed in other sections of the employee questionnaire. We also assessed the relationship between changes in health and safety climate and the health promotion programs offered in the workplace, as reported at follow-up in interviews with key informants in the worksite. To evaluate the relationship between climate and employee behavior, Pearson correlation coefficients were calculated between worksite means on the scale scores and worksite means of worker health and safety behavior. To assess the effect of the Working Well program on health and safety climate, baseline and follow-up scale scores were analyzed using a re-
Item and Factor Analyses The majority of items had a standard deviation of at least 1 point on the 5-point Likert response scales, as shown in Table 1. Following reversal of the coding for negatively worded items (items 7, 11, and 12), item means ranged from 2.39 to 4.60, with higher response values indicating more positive health and safety climate. A reliability analysis including 5,947 subjects who responded to all 18 items showed the mean interitem correlation coefficient to be 0.23, with coefficients ranging from −0.06 to 0.59. Item and factor analysis procedures resulted in the elimination of 7 items (items 1, 7, 10, 11, 12, 13, and 14). Items 10, 11, and 12 appeared to be conceptually unrelated to the other items as well as being statistically unrelated. In the factor analysis, items 1, 7, and 13 exhibited low factor loadings (less than 0.35). Finally, item 14 was omitted because of split-loadings of borderline magnitude (0.35 and 0.40). The final solution had two factors: a 6-item ‘‘safety climate’’ factor (factor 1) and a 5-item ‘‘health climate’’ factor (factor 2). Factor 1 accounted for 32.7% of the total variance, and factor 2 accounted for an additional 9% of the variance. In the residual matrix, 5% of the residuals had absolute values higher than 0.05. Table 2 shows the results of the two-factor solution. Both the factor pattern, which includes the item loadings, and the factor structure are presented. The structure matrix takes into account the interrelatedness of the two factors, which resulted from the oblique rotation. The two factors had a substantial interfactor correlation coefficient of 0.48. We analyzed the final 11 items using the follow-up data and found similar results. The first factor, on which the 6 safety climate items had high loadings, accounted for 36.8% of the variance. The 5 health climate items had high loadings on the second factor, which accounted for 9.4% of the variance. The correlation between the two factors was 0.50. Formation of Scale Scores and Internal Consistency of Scales Based on the results of the factor analyses, scale scores were created to represent each factor. Because the unweighted scores were highly correlated with the weighted factor scores for both scales (r > 0.98 for both
HEALTH AND SAFETY CLIMATE
TABLE 1 Characteristics of 18 Questionnaire Items Items 1. Around here they look at your safety record when they consider you for promotion. 2. I can usually work safely and still get my work done. 3. New workers get training in how to be safe on the job. 4. I think safety is very important to my employer. 5. My supervisor encourages me to make changes to improve my health. 6. Equipment is always kept in safe operating condition. 7. Supervisors here worry more about getting the job done than about the employee’s safety. 8. Most people here are very safety conscious. 9. Supervisors always enforce safety rules. 10. I think the health insurance benefits offered here are very good. 11. At least half of the workers here are overweight. 12. At least half of the workers here use tobacco. 13. I think management would support more programs to improve employee health 14. I think my health is important to my employer. 15. Around here they look at how well you take care of your health when they consider you for promotion. 16. Most employees here are very health conscious. 17. Supervisors always enforce health-related rules (smoking policies, requirements about medical examinations, etc.). 18. At my workplace, sometimes we talk with each other about improving our health and preventing disease.
No. of responses
Mean
Standard deviation
6,450
3.12
1.40
6,658
4.60
0.76
6,599
4.14
1.17
6,658
4.51
0.91
6,585
2.60
1.41
6,604
3.91
1.15
6,580
3.55
1.40
6,659
4.11
0.97
6,591
3.85
1.12
6,661
3.81
1.30
6,605
2.78
1.18
6,531
2.98
1.27
6,600
3.78
1.19
6,641
3.85
1.19
6,482
2.39
1.21
6,584
2.99
1.02
6,547
2.93
1.34
6,626
2.87
1.30
factors), the unweighted scale scores were used in subsequent analyses. Cronbach’s a estimates of internal consistency for the safety climate and health climate scales were 0.82 and 0.74, respectively. Concurrent Validity To obtain evidence in support of the scales’ concurrent validity, we examined relationships between the scale scores and other measured variables that we hypothesized to be related to organizational health or
115
safety climate. Table 3 shows correlations between the scale scores and the health- and safety-related criterion variables. With the exception of ‘‘At my workplace, it is easy to eat a healthy diet,’’ health-related criterion variables were more closely related to the health climate than the safety climate scale scores. The size of the association with the health-related criteria averaged 0.55 for the health climate scale and 0.29 for the safety climate scale. Likewise, the safety-related criterion variables were more strongly associated with the safety climate (average correlation 4 0.56) than with the health climate scale scores (average correlation 4 0.33). One would expect the health-related criteria to be related to some extent to safety climate and the safety-related criteria to be related to health climate, because the two factors are correlated. In addition, we assessed correlations between the health and safety climate scales and the workers’ selfreported health behaviors (Table 4). Eight of 11 correlations were nonsignificant, indicating that worksite means of the health and safety climate scales were unrelated to the worksite-level means of individual health and safety behavior. Two of the 3 significant correlations were hypothesized as being null associations. Average daily servings of fruits and vegetables and grams of fiber consumed by employees were negatively related to safety climate; safety climate was not expected to be related to these variables. However, the percentage of employees who wear safety sunglasses when working outdoors was positively related to safety climate, as would be expected. The relationship between the changes in the health and safety climate scale scores and the number of health promotion programs offered during the intervention period also was assessed (Table 5). The mean worksite change in health climate was 0.12 (standard deviation 0.18); the mean change in safety climate was −0.02 (standard deviation 0.17). The change in health climate from baseline to follow-up was related to the number of smoking programs and smokeless tobacco use programs offered during the past 2 years. However, change in health climate was not associated with screening programs, health and safety programs, exercise and weight control programs, or nutrition programs. Safety climate was not significantly related to any of the types of programs, although the correlation coefficient for the relationship to smoking programs was 0.30 (P 4 0.055). The final question concerned the effects of the Working Well Intervention on worksite health and safety climate. The repeated-measures analysis of variance model included the main effects for time (baseline or follow-up), group (intervention or control), type of industry (gas pipeline or electrical cooperative), and all two- and three-way interactions. Primarily, we were interested in the treatment-by-time interaction, as this term indicated whether change in the climate scales
116
BASEN-ENGQUIST ET AL.
TABLE 2 Results of the Two-Factor Solution, Individual Level of Analysis (N 4 5,947 Employees) Factor pattern
Items Safety climate I think safety is very important to my employer. New workers get training in how to be safe on the job. Most people here are very safety conscious. Equipment is always kept in safe operating condition. Supervisors always enforce safety rules. I can usually work safely and still get my work done. Health climate At my workplace, sometimes we talk with each other about improving our health and preventing disease. Most employees here are very health conscious. Around here they look at how well you take care of your health when they consider you for promotion. My supervisor encourages me to make changes to improve my health. Supervisors always enforce health-related rules (smoking policies, requirements about medical examinations, etc.).
Factor 1, safety climate
Factor structure
Factor 2, health climate
Factor 1, safety climate
Factor 2, health climate
Communality estimate
0.79 0.67 0.63 0.62 0.61 0.54
−0.07 0.04 0.09 0.14 0.15 −0.10
0.76 0.69 0.68 0.69 0.68 0.49
0.31 0.36 0.40 0.44 0.44 0.16
0.58 0.47 0.47 0.49 0.48 0.25
−0.07 0.03
0.63 0.62
0.23 0.33
0.59 0.64
0.36 0.41
−0.02
0.58
0.26
0.57
0.33
0.07
0.57
0.35
0.61
0.37
0.11
0.56
0.37
0.61
0.38
TABLE 3 Concurrent Validity with Organizational Health- and Safety-Related Variables Measured at Baseline, Worksite Level of Analysis (N 4 40 Worksites)
Health-related criterion measures There is a lot of information on healthy eating where I work. (1 4 strongly agree . . . 5 4 strongly disagree) At my workplace, it is easy to eat a healthy diet. (1 4 strongly agree . . . 5 4 strongly disagree) How much encouragement for eating low-fat foods do you get from your co-workers? (1 4 very much . . . 5 4 none) How often do people at work smoke in places where smoking is not allowed? (1 4 almost always . . . 5 4 never) How concerned does management seem about the effects of tobacco smoke on workers’ health? (1 4 extremely concerned . . . 5 4 not concerned) How concerned does management seem about whether you eat healthful food? (1 4 extremely concerned . . . 5 4 not concerned) Safety-related criterion measures How concerned do you think your employer is about your exposure to possibly harmful chemicals or other substances you work with? (1 4 extremely concerned . . . 5 4 not concerned) In my workplace, safer chemicals have been substituted for more harmful chemicals. (0 4 no; 1 4 yes) In my workplace, exhaust ducts or enclosures have been installed to collect harmful substances before they enter the workplace air. (0 4 no; 1 4 yes) In my workplace, ventilation systems are maintained regularly. (0 4 no; 1 4 yes) In my workplace, adequate protective equipment such as respirators or gloves is readily available. (0 4 no; 1 4 yes) In my workplace, harmful chemicals are labeled. (0 4 no; 1 4 yes) In my workplace, the equipment I work with is adequately maintained. (0 4 no; 1 4 yes) In my workplace, warning signs are posted in areas where hazardous chemicals may be present. (0 4 no; 1 4 yes) In my workplace, most exposures to possibly harmful substances are lowered by workplace controls, such as ventilation or enclosures. (0 4 no; 1 4 yes) In the past year, how much training have you received from your employer on the safe use of harmful substances? (1 4 none . . . 5 4 4 hr or more)c How concerned does management seem about reducing exposures to work hazards? (1 4 extremely concerned . . . 5 4 not concerned) n 4 5 items; higher scores indicative of climates more conducive to worker health. n 4 6 items; higher scores indicative of climates more conducive to worker safety. c Includes only respondents who indicated that they work with potentially harmful substances. * Correlation is statistically significant at P < 0.05. a b
Health climatea
Safety climateb
−0.75* −0.30 −0.48*
−0.24 −0.38* 0.15
0.24
0.18
−0.67*
−0.36*
−0.89*
−0.42*
−0.34*
−0.64*
0.14 0.22
0.39* 0.21
0.45* 0.11
0.57* 0.46*
0.39* 0.46* 0.22
0.68* 0.80* 0.45*
0.45*
0.58*
0.27
0.45*
−0.57*
−0.88*
117
HEALTH AND SAFETY CLIMATE
TABLE 4 Concurrent Validity with Self-Reported Health Behaviors Measured at Baseline, Worksite Level of Analysis (N 4 40 Worksites) Health behavior Total servings of fruits and vegetables per day Percentage of calories from fat Grams of fiber per 1,000 calories Grams of dietary fiber per day Percentage of worksite employees who Are current smokers Are current smokeless tobacco users Wear protective clothing when working outdoorsc,d Wear sun-protecting lotion when working outdoorsc,d Wear safety sunglasses when working outdoorsc,d Percentage of worksite employees who, after working with possibly harmful substances, wash hands before eating meals or snacksc,e Percentage of worksite employees who, after working with possibly harmful substances, wash hands before smoking or using smokeless tobaccoc,e,f
Health climatea
Safety climateb
−0.02 0.25 0.02 −0.15
−0.43* 0.30 −0.33* −0.26
−0.14 −0.10
0.17 −0.01
0.13
0.25
0.14
0.03
0.06
0.35*
TABLE 5 Correlations between Changes in Health Climate and Safety Climate and Number of Health Promotion Programs Offered during Previous 2 Years, as Measured at Follow-Up, Worksite Level of Analysis (N 4 40 Worksites) Type of program
Health climatea
Safety climateb
Screening Health and safety programs Smoking programs Smokeless tobacco programs Exercise and weight control programs Nutrition programs
−0.25 0.27 0.46* 0.42* 0.06 0.08
−0.17 0.18 0.30 0.27 0.08 0.03
n 4 5 items; higher scores indicative of climates more conducive to worker health; mean change in health climate 4 0.12, standard deviation 4 0.18. b n 4 6 items; higher scores indicative of climates more conducive to worker safety; mean change in safety climate 4 −0.02, standard deviation 4 0.17. * Correlation is statistically significant at P < 0.05. a
−0.17
0.05
DISCUSSION
−0.12
−0.09
a n 4 5 items; higher scores indicative of climates more conducive to worker health. b n 4 6 items; higher scores indicative of climates more conducive to worker safety. c Excludes white-collar employees. d Excludes employees who indicated that they do not work outdoors. e Excludes employees who do not work with potentially harmful substances. f Excludes employees who indicated that they do not use tobacco. * Correlation is statistically significant at P < 0.05.
between baseline and follow-up differed between intervention and control worksites. The treatment-by-time interaction was significant for health climate (F[1,36] 4 7.57, P 4 0.009). Examination of the means in Table 6 indicates that the health climate at intervention worksites increased from baseline to final measurement more than it did at control worksites. The main effect for time was also significant (F[1,36] 4 21.47, P 4 0.000), indicating overall health climate increased during the study. Safety climate at the intervention worksites remained unchanged between baseline and follow-up, but the control worksites showed a trend toward a decrease in safety climate. However, this interaction term was not significant (F[1,36] 4 3.57, P 4 0.067). For safety climate, the industry-by-time interaction was significant (F[1,36] 4 6.02, P 4 0.02), indicating a decrease in safety climate over time at the electrical cooperatives, but not at the gas pipeline worksites. No other interactions or main effects were significant.
The health and safety climate scales developed for this study are useful instruments for measuring organizational change related to worksite health promotion activities. The scales exhibited good internal consistency and were related to other conceptually relevant organizational constructs. Our study demonstrated that health and safety climate is related to other variables, indicative of an environment in which healthand safety-related behaviors are supported. These associations confirmed the convergent and divergent validity of the health and safety climate scales, in that TABLE 6 Intervention Effects on Health and Safety Climate, Worksite Level of Analysis (N 4 40 Worksites)
Health climatea Intervention Natural gas pipeline Electrical cooperative Control Natural gas pipeline Electrical cooperative Safety climateb Intervention Natural gas pipeline Electrical cooperative Control Natural gas pipeline Electrical cooperative
Baseline mean (SD)
Final mean (SD)
2.78 (0.24) 2.71 (0.15) 2.84 (0.29) 2.75 (0.20) 2.75 (0.18) 2.74 (0.23)
2.97 (0.26) 2.91 (0.24) 3.03 (0.28) 2.80 (0.24) 2.78 (0.23) 2.81 (0.27)
4.22 (0.23) 4.27 (0.15) 4.18 (0.29) 4.23 (0.27) 4.22 (0.24) 4.23 (0.31)
4.25 (0.24) 4.33 (0.16) 4.16 (0.28) 4.16 (0.29) 4.23 (0.28) 4.08 (0.29)
a Treatment group × time interaction is significant, F(1,36) 4 7.57, P 4 0.009; main effect for time is significant, F(1,36) 4 21.47, P 4 0.000. b Treatment group × time interaction is not significant, F(1,36) 4 3.57, P 4 0.067; industry × time interaction effect is significant, F(1,36) 4 6.02, P 4 0.02.
118
BASEN-ENGQUIST ET AL.
the health-related variables were more highly correlated with the health climate scale, and the safetyrelated variables were more highly correlated with the safety climate scale. The health climate scale was related to the presence of tobacco use cessation programs in the workplace at the end of the study, indicating that changes in organizational climate helped support the implementation of health promotion programs, as predicted by theory [7,8]. Concurrent validity was not established in the relationship of health and safety climate to employee health and safety behavior. Although the literature on organizational climate leads us to believe that employee behavior should be affected by climate, [16–19] our findings did not support this relationship. It is possible that such a relationship is found in other studies that focus on behaviors confined primarily to the workplace (e.g., absenteeism, productivity). Health and safety behaviors may be influenced to some extent by workplace climate, but they likely are affected more directly by nonworksite factors. For example, although a worksite smoking policy or smoking cessation program may alter smoking behavior somewhat, this behavior is probably influenced more heavily by other factors, such as level of addiction and smoking by significant others. The Working Well Intervention did not have a significant impact on smoking, [29] suggesting that smoking in particular is influenced heavily by nonworksite factors. The health and safety climate scales would benefit from additional validity and reliability testing. The design of the current study did not allow us to assess the scales’ test–retest reliability. Furthermore, our factor analysis was exploratory, and because several items were dropped to provide a good solution, the stability of the results is uncertain. However, the similar factor analytic solution produced by the follow-up data strengthens our confidence in the results. The scales would benefit from further testing on other populations and confirmatory factor analysis. The health climate scale was sufficiently sensitive to detect changes in the worksite environment. The health climate scores at worksites at which a cancer prevention intervention took place increased from baseline to follow-up, whereas scores at control worksites remained the same. The design of the intervention, which emphasized building organizational capacity for conducting health promotion programs, probably influenced this change. Because the intervention focused on diet and tobacco use, as opposed to safetyrelated behavior, we did not expect to see increases in safety climate. Even if the intervention had included more safety content, however, we might not have detected much change; the safety climate scores were quite high at baseline, and thus the scale may not have been sufficiently sensitive to measure subtle differences in the top range. This may be so because the
types of industries in which we worked are heavily regulated, and adherence to safety practices is a pervasive industry expectation. Greater variance on the safety climate scale scores might be found among other industries. A strength of this study is that there were a sufficient number of worksites to conduct worksite-level analyses. This is critical when answering research questions about organizational characteristics, [13] but is not done in some studies because too few worksites are studied. A second strength is that two sources of data were used for criterion variables in assessing concurrent validity; items from the same employee questionnaire that assessed climate and data from interview about health promotion activities with key informants within the worksites. We were somewhat limited in the type of data available from the employee survey that could be used to assess the concurrent validity. In selecting the criterion variables from the employee questionnaire, we attempted to identify variables that did not rely on individual perception, in order to distinguish them sufficiently from the climate construct. However, because there were few such variables on the employee survey we had to rely on more subjective criterion variables. Although the more subjective variables tended to have higher correlations with the scales than the more objective criteria, overall the findings support the concurrent validity of the scales. CONCLUSIONS
The health and safety climate scales had good internal consistency reliability, and their validity is supported by correlations between scale scores and conceptually related organizational characteristics. The lack of association between health and safety climate scores and health and safety behaviors was disappointing, but may be due to the multiple influences on health and safety behaviors or insufficient power caused by using the worksite as the unit of analysis. Two additional findings increased our confidence in the scales’ validity: the relationship between changes in health climate and the number of tobacco use programs offered and the increase in health climate scores among worksites receiving the cancer prevention intervention compared with control worksites. Health promotion theory stresses the importance of changes in the environment to support changes in behavior. Organizational theory also suggests the importance of environmental influences on the implementation of health promotion in the workplace. Despite this emphasis on organizational influences, measurement of environmental and organizational characteristics has lagged behind measurement of intrapersonal variables associated with change in health behavior. Our data suggest that this measurement tool can be useful
HEALTH AND SAFETY CLIMATE
for worksite health promotion programs, both for assessing receptivity to health promotion programs before the intervention and for assessing change over the course of a program.
14. 15.
REFERENCES 1. Bandura A. Social foundations of thought and action: a social cognitive theory. Englewood Cliffs (NJ): Prentice-Hall, 1986. 2. Rogers EM. Diffusion of innovations. New York: Free Press, 1983. 3. Witte K. Managerial style and health promotion programs. Soc Sci Med 1993;36:227–35. 4. Glasgow RE, Sorensen G, Corbett K. Worksite smoking control activities: prevalence and related worksite characteristics from the COMMIT study, 1990. Prev Med 1992;21:688–700. 5. Hollander RB, Lengermann JJ. Corporate characteristics and worksite health promotion programs: survey findings from Fortune 500 companies. Soc Sci Med 1988;26:491–501. 6. Glasgow RE, Hollis JF, Ary DV, Lando HA. Employee and organizational factors associated with participation in an incentivebased worksite smoking cessation program. J Behav Med 1990; 13:403–18. 7. Gottlieb NH, McLeroy KR. Social health. In: O’Donnell MP, Harris JS, editors. Health promotion in the workplace. Albany: Delmar Pub., 1994. 8. Allen J, Bellingham R. Building supportive cultural environments. In: O’Donnell MP, Harris JS, editors. Health promotion in the workplace. Albany: Delmar Pub., 1994. 9. Emont SL, Cummings KM. Organizational factors affecting participation in a smoking cessation program and abstinence among 68 auto dealerships. Am J Health Promot 1990;5:107–14. 10. Terborg JR. Organizational context. In: Johnson K, Larosa JH, Scheirer CJ, Wolle JM, editors. Methodological issues in worksite research: proceedings of a workshop, April 1988. Washington: U.S. Department of Health and Human Services, U.S. Public Health Service, 1988. 11. Rost K, Connell C, Schechtman K, Barzilai B, Fisher EB Jr. Predictors of employee involvement in a worksite health promotion program: Health Educ Q 1990;17:395–407. 12. Sloan RP, Gruman JC. Participation in workplace health promotion programs: the contribution of health and organizational factors. Health Educ Q 1988;15:269–88. 13. Ribisl KM, Reischl TM. Measuring the climate for health at or-
16. 17. 18. 19. 20. 21.
22.
119
ganizations: development of the worksite health climate scales. J Occup Med 1993;35:812–24. James LR, Jones AP. Organizational climate: a review of theory and research. Psychol Bull 1974;81:1096–112. Hoy WK, Tarter CJ, Kottkamp RB. Open schools/healthy schools: measuring organizational climate. Newbury Park: Sage, 1991. Forehand GA, Gilmer B. Environmental variation in studies of organizational behavior. Psychol Bull 1964;62:361–82. Field RHG, Abelson MA. Climate: a reconceptualization and proposed model. Hum Relat 1982;35:181–201. Moran ET, Volkwein JF. The cultural approach to the formation of organizational climate. Hum Relat 1992;45:19–47. Schneider B. Organizational climate: an essay. Pers Psychol 1975;28:447–479. Zohar D. Safety climate in industrial organizations: theoretical and applied implications. J Appl Psychol 1980;65:96–102. Brown RL, Holmes H. The use of a factor-analytic procedure for assessing the validity of an employee safety climate model. Accid Anal Prev 1986;18:455–70. Franklin J. Down the organization: influence processes across levels of hierarchy. Admin Sci Q 1975;20:153–64.
23. Scofield ME, Martin W. Development of the AT&T health audit for measuring organizational health. State Art Rev Occup Med 1990;5:755–88. 24. Brooks L. Employee perceptions of the worksite health climate at a large company [thesis]. Houston: Univ. of Texas School of Public Health, 1995. 25. Abrams DB, Boutwell WB, Grizzle J, Heimendinger J, Sorensen G, Varnes J. Cancer control at the workplace: the Working Well Trial. Prev Med 1994;23:1–13. 26. Heimendinger J, Feng Z, Emmons K, Stoddard A, Kinne S, Biener L, Sorenson G, Abrams D, Varnes J, Boutwell B. The Working Well Trial: baseline dietary and smoking behaviors of employees and related worksite characteristics. Prev Med 1995;24: 180–93. 27. Nunnally JC, Bernstein IH. Psychometric theory. New York: McGraw–Hill, 1994. 28. Gorsuch RL. Factor analysis. Hillsdale (NJ): Erlbaum, 1983. 29. Sorensen G, Thompson B, Glanz K, Feng Z, Kinne S, DiClemente C, Emmons K, Heimendinger J, Probart C, Lichtenstein E. Work site-based cancer prevention: primary results from the Working Well Trial. Am J Public Health 1996;86:939–47.
PREVENTIVE MEDICINE ARTICLE NO.
Worksite Health and Safety Climate: Scale Development and Effects of a Health Promotion Intervention1 Karen Basen-Engquist, Ph.D., M.P.H.,*,2 Karen Suchanek Hudmon, Dr.P.H., R.Ph.,*,† Mary Tripp, M.P.H.,* and Robert Chamberlain, Ph.D.† *Department of Behavioral Science and †Department of Epidemiology, The University of Texas M. D. Anderson Cancer Center, Houston, Texas 77030
Background. Environmental influences on health and health behavior have an important place in research on worksite health promotion. We tested the validity and internal consistency of a new measure of organizational health and safety climate that was used in a large randomized trial of a worksite cancer prevention program (the Working Well Trial). The resulting scales then were applied to assess intervention effects. Method. This study uses data from a subset of 40 worksites in the Working Well Trial. Employees at 20 natural gas pipeline worksites and 20 rural electrical cooperatives completed a cross-sectional questionnaire at baseline and 3-year follow-up. Results. A factor analysis of this self-report instrument produced a two-factor solution. The resulting health and safety climate scales had good internal consistency (Cronbach’s a = 0.74 and 0.82, respectively) and concurrent validity. The health climate scale was correlated more highly with organizational measures that were indicative of a supportive health climate than those indicating supportive safety climate, while the reverse was true of the safety climate scale. Changes in health climate were associated with the number of smoking and smokeless tobacco programs offered at the worksites at the time of the 3-year follow-up (r = 0.46 and 0.42, respectively). The scales were not correlated with most employee health behaviors. The health climate scores increased at intervention worksites, compared with scores at control worksites (F[1,36] = 7.57, P = 0.009). Conclusions. The health and safety climate scales developed for this study provide useful instruments for measuring organizational change related to worksite health promotion activities. The Working Well Intervention resulted in a significant improvement in worksite health climate. © 1998 Academic Press
1 This work was supported by a Cooperative Agreement from the National Cancer Institute, Grant U01 CA51671. 2 To whom reprint requests should be addressed.
Key Words: organizational climate; cancer prevention; worksite health promotion; scale development; validity. INTRODUCTION
Health promotion and organization theories suggest that effective worksite health promotion programs must address environmental factors that influence health behavior. Environmental factors derived from sociobehavioral theoretical frameworks, such as Social Cognitive Theory [1] and Diffusion of Innovations Theory, [2] have been discussed extensively in the health promotion literature. We describe a new measure of an environmental factor often studied in the organizational literature—organizational climate. Traditionally, organizational climate has been studied for its effects on organizational outcomes such as attendance, absenteeism, productivity, performance, and job satisfaction. For this study, organizational climate related to health and safety was investigated as it relates to health promotion at the worksite. Within the framework of a randomized trial of a worksite cancer prevention program, we examined the psychometric properties of an instrument designed to measure the worksite climate regarding health and safety and assessed the effect of the intervention on this climate. This study is unique because it provides data on both the validity and the reliability of the climate measure, in addition to assessing the sensitivity of the measure to climate change resulting from an organizational intervention. The literature suggests a wide range of environmental factors that can affect an organization’s health promotion programming and the participation of its employees. These include organizational management style; [3] industry type; [4,5] organization rank; [5] worksite size; [4,6] organizational structure; [7] degree of support from top management, [6] including incentives or recognition for healthy behaviors or participation in a health promotion program; [8] modeling of healthy behaviors by the CEO; [8] policies reinforcing healthy behaviors; [8,9] employee autonomy and con-
111 0091-7435/98 $25.00 Copyright © 1998 by Academic Press All rights of reproduction in any form reserved.
112
BASEN-ENGQUIST ET AL.
trol over work activities; [10] and subcultures within the organization [10]. One environmental factor in particular, organizational climate, is the focus of this study. Organizational climate has been identified as an important environmental influence on employee participation in worksite health promotion programs and on the effectiveness of such programs [7,8,11–13]. Although organizational climate has been studied in educational, business, and health care institutions since the 1950s, there is little consensus among researchers regarding the definition of organizational climate [14,15]. Nevertheless, the conceptual components central to the various views can be summarized. Organizational climate is a set of internal characteristics that distinguishes one organization from another, is experienced by members of the organization, influences their behavior, and is based on their collective perception of the organizational environment [16–19]. Climate is influenced by the physical environment, organizational characteristics such as centralization and formalization, formal and informal relationships, personalities or organization members, and leadership [15,17]. The interactive perspective on climate formation suggests that an organization’s members interact with the situational environment to create climate [16,18]. By responding to a situation or some organizational characteristic according to shared perceptions, individuals in an organization create climate [18]. Along with individual motives and abilities, climate in turn influences behavior [16–19]. This concept is similar to that of reciprocal determinism in Social Cognitive Theory (i.e., the idea that environment, behavior, and personal factors mutually influence each other) [1]. For this reason, it is important to address climate when attempting to change the behavior of an organization’s members [15,18]. Studies of health promotion have dealt with the relationship of organizational climate to health behaviors and participation in health promotion programs in the workplace [11–13]. Some researchers have investigated general organizational climate as it is represented in the organizational literature, [12] while others have specifically investigated the health-related dimensions of climate [11,13]. Dimensions of climate such as perceived supervisor support, [12] perceived control over work matters, [12] and worktime flexibility to allow time for personal health needs [11] are associated with employee participation in health promotion programs. Health climate also has been associated with exercise habits, healthy nutrition habits, and smoking status [13]. We identified only four instruments that specifically assess the health and safety climate and its relationship to the implementation of worksite health promotion programs: (1) the Safety Climate Questionnaire, a measure of safety concerns in the workplace; [20,21] (2)
the Organizational Climate Survey, a measure of general organizational climate, with added items related to health promotion; [11,22] (3) the AT&T Health Audit, a measure of general climate, health climate, and social support systems, but created to measure organizational culture; [23] and (4) the Worksite Health Climate Scales (WHCS), a measure of organizational support, interpersonal support, and health norms [13]. Empirical investigation of the psychometric properties of these instruments is limited. Construct validity and reliability testing has been replicated in only a few instances, and rarely are both validity and reliability assessed for a given instrument. A principal-components factor analysis of the Safety Climate Questionnaire, the only safety climate scale located in the literature, originally resulted in an eight-factor solution [20]. In a second study this structure could not be replicated, but the data did support the use of a smaller three-factor model [21]. Psychometric analysis of the existing health climate scales has focused on reliability testing. Factor analysis, which is typically one of the first steps in scale validation, has not been reported for any of these scales. For the health climate subscales of the Organizational Climate Survey, [11] a estimates of internal consistency reliabilities ranged from 0.70 to 0.88, and no further reliability testing has been reported. When the AT&T Health Audit’s test–retest reliability was assessed, the proportion of respondents giving identical answers to single items at the two time periods ranged from 58 to 99%; the proportion of identical responses tended to be somewhat lower, 61 to 78%, for the health norms and values items. Criterion and concurrent validity were supported for the AT&T Health Audit [23]. For the WHCS, a estimates of internal consistency reliability of all 12 subscales of the WHCS ranged from 0.73 to 0.94 in a population of small worksites [13]. Reliability testing of the WHCS at larger worksites yielded similar results, with scale reliability ranging from 0.69 to 0.96 [24]. Low correlations among subscales showed that they are conceptually distinct, although they tap related constructs [13,24]. Differences in health climate among worksites were investigated in a few studies, [13] as were the relationships of health climate, health status, and the performance of health behaviors [13,23]. We could identify no assessments of the effect of a health promotion intervention on worksite health or safety climate, nor do we know of any examinations of the relationships between health and safety climate and variables such as the number of health promotion programs available at the worksite, organizational characteristics, and safety behaviors. For these reasons, the purposes of this study were to (1) estimate the reliability and validity of the health and safety climate scale developed for a randomized trial of a worksite cancer prevention program and (2)
HEALTH AND SAFETY CLIMATE
assess the effect of the worksite cancer prevention intervention on organizational climate of health and safety. Because intervention strategies addressed social and environmental factors associated with the adoption and practice of healthy behaviors, thereby creating a supportive environment for such behaviors, we hypothesized that health climate would improve as a result of the intervention. Because the intervention did not emphasize worksite safety, we expected no change in safety climate. METHOD
Background Data were collected as part of the Working Well Trial, the largest worksite cancer control intervention trial in the United States [25]. The primary hypothesis of the Working Well Trial was that a sustained 2-year cancer control worksite health promotion intervention addressing dietary change and smoking cessation would be more effective than a minimal intervention (consisting of widely available print materials) in achieving both individual and environmental changes. The study was conducted at 114 worksites and coordinated by four study centers. The results presented here derive from the study center at The University of Texas M. D. Anderson Cancer Center (UTMDACC), where investigators were responsible for a subset of 40 worksites. The project was reviewed and approved by the UTMDACC institutional review board. Design The Working Well Trial used a randomized, matched-pair research design in which the worksite was the unit of randomization and analysis. Worksites were randomized after baseline assessment; they were stratified, matched into pairs, and randomly assigned within pairs to intervention or control group. Stratification variables included presence of a cafeteria, worksite size, type of smoking policy, company type, gender distribution, distribution of blue- and white-collar jobs, and response rate to the baseline survey. The intervention group received the program described below. The UTMDACC control group received widely available print materials such as posters and brochures. Employees at each worksite completed a cross-sectional questionnaire at baseline and at a 3-year follow-up. Description of the Sample The worksites studied by the UTMDACC study center included 20 natural gas pipeline worksites and 20 rural electrical cooperatives; each of these sites employed between 72 and 621 employees. The worksites were located primarily in the southeast quadrant of the United States, from Texas eastward to Florida and northward to Nebraska and Colorado. Of the 6,867 em-
113
ployees who responded to the baseline survey at the UTMDACC worksites, the majority were male (80.3%) and white (90.5%). Most of the sample was currently married (81.6%). Ninety-six percent had at least a high school education; 57.3% had education beyond high school. Intervention and control worksites had no detectable demographic differences at baseline or followup [26]. Working Well Intervention The Working Well Intervention, based on individual, organizational, and community organization theories, focused on (1) promoting and building employee awareness of preventive behaviors, (2) action and skills training, and (3) behavior maintenance and relapse prevention. The risk behaviors targeted at the UTMDACC intervention worksites were smoking, smokeless tobacco use, and diets low in fruits, vegetables, and fiber and high in fat. The activities conducted to address these risk behaviors included promotional and awareness materials and activities, such as posters, brochures, and videos; self-assessments that enabled employees to assess their risk behavior or health status and compare their results with a standard; self-help materials that provided step-by-step instructions for helping people build the skills they needed to adopt healthy behaviors; and direct education for smoking cessation or dietary change offered through multisession support groups or classes. The intervention was implemented using a participatory strategy that involved selecting an employee from each intervention worksite to serve as the coordinator for the Working Well program and selecting a committee of employees to serve as the employee advisory board (EAB). Worksite coordinators and EABs were responsible for planning and implementing activities and tailoring them to their worksite. They were given a program manual outlining study protocols and general information about implementing health promotion programs in the workplace. Intervention staff from UTMDACC provided support and technical assistance to the worksite coordinators and EABs. Data Collection The baseline data were collected between September and December 1990, and the follow-up data were collected between September and December 1993. Questionnaires were administered by both worksite personnel and UTMDACC staff members to groups of employees at the worksite. Employees who were absent or could not complete the survey on the scheduled day were given a questionnaire to be completed and mailed to the study center. All employees were recruited to complete the questionnaire. Response rates were 89% at baseline and 86% at follow-up. Organizational interviews with key worksite infor-
114
BASEN-ENGQUIST ET AL.
mants (CEO or highest ranking employee at worksite, personnel director, health and safety director, etc.) were conducted by telephone. UTMDACC staff members conducted baseline interviews from September to December 1990, and the final interviews were conducted from September to December 1993.
peated-measures analysis of variance with one withinworksites factor (time of measurement) and two between-worksites factors (treatment condition and type of industry). The unit of analysis was the worksite for tests of concurrent validity and intervention effects, because health and safety climate is conceptualized as a worksite characteristic.
Measurement and Analysis Procedures RESULTS
A panel of researchers with experience in the study of organizational climate developed a series of 18 items to assess the organizational health and safety climate of the UTMDACC Working Well sites. The items were administered to 6,867 participants. Psychometric analyses to determine whether the items were sufficiently related to justify formation of one or more health and safety climate scales included assessment of item characteristics, interitem correlations and internal consistency, and an exploratory factor analysis to determine the underlying structure of the scales [27,28]. A matrix of interitem correlations was used as input for unweighted least-squares factor extraction with oblique rotation. An oblique rather than orthogonal rotation was selected because we hypothesized that two factors, health climate and safety climate, would emerge, and that they would be related statistically. Because the questionnaire items had not been pretested or used in previous studies, items that did not exhibit desirable psychometric properties were eliminated. A factor loading of 0.40 was the cutoff for retaining an item, although the selection of a factor analysis solution also was based on theory, the communality estimates, the residual matrix, and the percentage of variance explained by the extracted factors. The psychometric results reported here are derived from baseline measures. We also conducted a factor analysis on the scale items from the follow-up data, to determine if the factor analysis solution derived from the baseline data could be replicated. The item analysis, internal consistency, and factor analysis were all calculated using the individual employee as the unit of analysis. To assess the concurrent validity of the climate scales, Pearson correlation coefficients were calculated between worksite means of the scale scores and theoretically related organizational characteristics that were assessed in other sections of the employee questionnaire. We also assessed the relationship between changes in health and safety climate and the health promotion programs offered in the workplace, as reported at follow-up in interviews with key informants in the worksite. To evaluate the relationship between climate and employee behavior, Pearson correlation coefficients were calculated between worksite means on the scale scores and worksite means of worker health and safety behavior. To assess the effect of the Working Well program on health and safety climate, baseline and follow-up scale scores were analyzed using a re-
Item and Factor Analyses The majority of items had a standard deviation of at least 1 point on the 5-point Likert response scales, as shown in Table 1. Following reversal of the coding for negatively worded items (items 7, 11, and 12), item means ranged from 2.39 to 4.60, with higher response values indicating more positive health and safety climate. A reliability analysis including 5,947 subjects who responded to all 18 items showed the mean interitem correlation coefficient to be 0.23, with coefficients ranging from −0.06 to 0.59. Item and factor analysis procedures resulted in the elimination of 7 items (items 1, 7, 10, 11, 12, 13, and 14). Items 10, 11, and 12 appeared to be conceptually unrelated to the other items as well as being statistically unrelated. In the factor analysis, items 1, 7, and 13 exhibited low factor loadings (less than 0.35). Finally, item 14 was omitted because of split-loadings of borderline magnitude (0.35 and 0.40). The final solution had two factors: a 6-item ‘‘safety climate’’ factor (factor 1) and a 5-item ‘‘health climate’’ factor (factor 2). Factor 1 accounted for 32.7% of the total variance, and factor 2 accounted for an additional 9% of the variance. In the residual matrix, 5% of the residuals had absolute values higher than 0.05. Table 2 shows the results of the two-factor solution. Both the factor pattern, which includes the item loadings, and the factor structure are presented. The structure matrix takes into account the interrelatedness of the two factors, which resulted from the oblique rotation. The two factors had a substantial interfactor correlation coefficient of 0.48. We analyzed the final 11 items using the follow-up data and found similar results. The first factor, on which the 6 safety climate items had high loadings, accounted for 36.8% of the variance. The 5 health climate items had high loadings on the second factor, which accounted for 9.4% of the variance. The correlation between the two factors was 0.50. Formation of Scale Scores and Internal Consistency of Scales Based on the results of the factor analyses, scale scores were created to represent each factor. Because the unweighted scores were highly correlated with the weighted factor scores for both scales (r > 0.98 for both
HEALTH AND SAFETY CLIMATE
TABLE 1 Characteristics of 18 Questionnaire Items Items 1. Around here they look at your safety record when they consider you for promotion. 2. I can usually work safely and still get my work done. 3. New workers get training in how to be safe on the job. 4. I think safety is very important to my employer. 5. My supervisor encourages me to make changes to improve my health. 6. Equipment is always kept in safe operating condition. 7. Supervisors here worry more about getting the job done than about the employee’s safety. 8. Most people here are very safety conscious. 9. Supervisors always enforce safety rules. 10. I think the health insurance benefits offered here are very good. 11. At least half of the workers here are overweight. 12. At least half of the workers here use tobacco. 13. I think management would support more programs to improve employee health 14. I think my health is important to my employer. 15. Around here they look at how well you take care of your health when they consider you for promotion. 16. Most employees here are very health conscious. 17. Supervisors always enforce health-related rules (smoking policies, requirements about medical examinations, etc.). 18. At my workplace, sometimes we talk with each other about improving our health and preventing disease.
No. of responses
Mean
Standard deviation
6,450
3.12
1.40
6,658
4.60
0.76
6,599
4.14
1.17
6,658
4.51
0.91
6,585
2.60
1.41
6,604
3.91
1.15
6,580
3.55
1.40
6,659
4.11
0.97
6,591
3.85
1.12
6,661
3.81
1.30
6,605
2.78
1.18
6,531
2.98
1.27
6,600
3.78
1.19
6,641
3.85
1.19
6,482
2.39
1.21
6,584
2.99
1.02
6,547
2.93
1.34
6,626
2.87
1.30
factors), the unweighted scale scores were used in subsequent analyses. Cronbach’s a estimates of internal consistency for the safety climate and health climate scales were 0.82 and 0.74, respectively. Concurrent Validity To obtain evidence in support of the scales’ concurrent validity, we examined relationships between the scale scores and other measured variables that we hypothesized to be related to organizational health or
115
safety climate. Table 3 shows correlations between the scale scores and the health- and safety-related criterion variables. With the exception of ‘‘At my workplace, it is easy to eat a healthy diet,’’ health-related criterion variables were more closely related to the health climate than the safety climate scale scores. The size of the association with the health-related criteria averaged 0.55 for the health climate scale and 0.29 for the safety climate scale. Likewise, the safety-related criterion variables were more strongly associated with the safety climate (average correlation 4 0.56) than with the health climate scale scores (average correlation 4 0.33). One would expect the health-related criteria to be related to some extent to safety climate and the safety-related criteria to be related to health climate, because the two factors are correlated. In addition, we assessed correlations between the health and safety climate scales and the workers’ selfreported health behaviors (Table 4). Eight of 11 correlations were nonsignificant, indicating that worksite means of the health and safety climate scales were unrelated to the worksite-level means of individual health and safety behavior. Two of the 3 significant correlations were hypothesized as being null associations. Average daily servings of fruits and vegetables and grams of fiber consumed by employees were negatively related to safety climate; safety climate was not expected to be related to these variables. However, the percentage of employees who wear safety sunglasses when working outdoors was positively related to safety climate, as would be expected. The relationship between the changes in the health and safety climate scale scores and the number of health promotion programs offered during the intervention period also was assessed (Table 5). The mean worksite change in health climate was 0.12 (standard deviation 0.18); the mean change in safety climate was −0.02 (standard deviation 0.17). The change in health climate from baseline to follow-up was related to the number of smoking programs and smokeless tobacco use programs offered during the past 2 years. However, change in health climate was not associated with screening programs, health and safety programs, exercise and weight control programs, or nutrition programs. Safety climate was not significantly related to any of the types of programs, although the correlation coefficient for the relationship to smoking programs was 0.30 (P 4 0.055). The final question concerned the effects of the Working Well Intervention on worksite health and safety climate. The repeated-measures analysis of variance model included the main effects for time (baseline or follow-up), group (intervention or control), type of industry (gas pipeline or electrical cooperative), and all two- and three-way interactions. Primarily, we were interested in the treatment-by-time interaction, as this term indicated whether change in the climate scales
116
BASEN-ENGQUIST ET AL.
TABLE 2 Results of the Two-Factor Solution, Individual Level of Analysis (N 4 5,947 Employees) Factor pattern
Items Safety climate I think safety is very important to my employer. New workers get training in how to be safe on the job. Most people here are very safety conscious. Equipment is always kept in safe operating condition. Supervisors always enforce safety rules. I can usually work safely and still get my work done. Health climate At my workplace, sometimes we talk with each other about improving our health and preventing disease. Most employees here are very health conscious. Around here they look at how well you take care of your health when they consider you for promotion. My supervisor encourages me to make changes to improve my health. Supervisors always enforce health-related rules (smoking policies, requirements about medical examinations, etc.).
Factor 1, safety climate
Factor structure
Factor 2, health climate
Factor 1, safety climate
Factor 2, health climate
Communality estimate
0.79 0.67 0.63 0.62 0.61 0.54
−0.07 0.04 0.09 0.14 0.15 −0.10
0.76 0.69 0.68 0.69 0.68 0.49
0.31 0.36 0.40 0.44 0.44 0.16
0.58 0.47 0.47 0.49 0.48 0.25
−0.07 0.03
0.63 0.62
0.23 0.33
0.59 0.64
0.36 0.41
−0.02
0.58
0.26
0.57
0.33
0.07
0.57
0.35
0.61
0.37
0.11
0.56
0.37
0.61
0.38
TABLE 3 Concurrent Validity with Organizational Health- and Safety-Related Variables Measured at Baseline, Worksite Level of Analysis (N 4 40 Worksites)
Health-related criterion measures There is a lot of information on healthy eating where I work. (1 4 strongly agree . . . 5 4 strongly disagree) At my workplace, it is easy to eat a healthy diet. (1 4 strongly agree . . . 5 4 strongly disagree) How much encouragement for eating low-fat foods do you get from your co-workers? (1 4 very much . . . 5 4 none) How often do people at work smoke in places where smoking is not allowed? (1 4 almost always . . . 5 4 never) How concerned does management seem about the effects of tobacco smoke on workers’ health? (1 4 extremely concerned . . . 5 4 not concerned) How concerned does management seem about whether you eat healthful food? (1 4 extremely concerned . . . 5 4 not concerned) Safety-related criterion measures How concerned do you think your employer is about your exposure to possibly harmful chemicals or other substances you work with? (1 4 extremely concerned . . . 5 4 not concerned) In my workplace, safer chemicals have been substituted for more harmful chemicals. (0 4 no; 1 4 yes) In my workplace, exhaust ducts or enclosures have been installed to collect harmful substances before they enter the workplace air. (0 4 no; 1 4 yes) In my workplace, ventilation systems are maintained regularly. (0 4 no; 1 4 yes) In my workplace, adequate protective equipment such as respirators or gloves is readily available. (0 4 no; 1 4 yes) In my workplace, harmful chemicals are labeled. (0 4 no; 1 4 yes) In my workplace, the equipment I work with is adequately maintained. (0 4 no; 1 4 yes) In my workplace, warning signs are posted in areas where hazardous chemicals may be present. (0 4 no; 1 4 yes) In my workplace, most exposures to possibly harmful substances are lowered by workplace controls, such as ventilation or enclosures. (0 4 no; 1 4 yes) In the past year, how much training have you received from your employer on the safe use of harmful substances? (1 4 none . . . 5 4 4 hr or more)c How concerned does management seem about reducing exposures to work hazards? (1 4 extremely concerned . . . 5 4 not concerned) n 4 5 items; higher scores indicative of climates more conducive to worker health. n 4 6 items; higher scores indicative of climates more conducive to worker safety. c Includes only respondents who indicated that they work with potentially harmful substances. * Correlation is statistically significant at P < 0.05. a b
Health climatea
Safety climateb
−0.75* −0.30 −0.48*
−0.24 −0.38* 0.15
0.24
0.18
−0.67*
−0.36*
−0.89*
−0.42*
−0.34*
−0.64*
0.14 0.22
0.39* 0.21
0.45* 0.11
0.57* 0.46*
0.39* 0.46* 0.22
0.68* 0.80* 0.45*
0.45*
0.58*
0.27
0.45*
−0.57*
−0.88*
117
HEALTH AND SAFETY CLIMATE
TABLE 4 Concurrent Validity with Self-Reported Health Behaviors Measured at Baseline, Worksite Level of Analysis (N 4 40 Worksites) Health behavior Total servings of fruits and vegetables per day Percentage of calories from fat Grams of fiber per 1,000 calories Grams of dietary fiber per day Percentage of worksite employees who Are current smokers Are current smokeless tobacco users Wear protective clothing when working outdoorsc,d Wear sun-protecting lotion when working outdoorsc,d Wear safety sunglasses when working outdoorsc,d Percentage of worksite employees who, after working with possibly harmful substances, wash hands before eating meals or snacksc,e Percentage of worksite employees who, after working with possibly harmful substances, wash hands before smoking or using smokeless tobaccoc,e,f
Health climatea
Safety climateb
−0.02 0.25 0.02 −0.15
−0.43* 0.30 −0.33* −0.26
−0.14 −0.10
0.17 −0.01
0.13
0.25
0.14
0.03
0.06
0.35*
TABLE 5 Correlations between Changes in Health Climate and Safety Climate and Number of Health Promotion Programs Offered during Previous 2 Years, as Measured at Follow-Up, Worksite Level of Analysis (N 4 40 Worksites) Type of program
Health climatea
Safety climateb
Screening Health and safety programs Smoking programs Smokeless tobacco programs Exercise and weight control programs Nutrition programs
−0.25 0.27 0.46* 0.42* 0.06 0.08
−0.17 0.18 0.30 0.27 0.08 0.03
n 4 5 items; higher scores indicative of climates more conducive to worker health; mean change in health climate 4 0.12, standard deviation 4 0.18. b n 4 6 items; higher scores indicative of climates more conducive to worker safety; mean change in safety climate 4 −0.02, standard deviation 4 0.17. * Correlation is statistically significant at P < 0.05. a
−0.17
0.05
DISCUSSION
−0.12
−0.09
a n 4 5 items; higher scores indicative of climates more conducive to worker health. b n 4 6 items; higher scores indicative of climates more conducive to worker safety. c Excludes white-collar employees. d Excludes employees who indicated that they do not work outdoors. e Excludes employees who do not work with potentially harmful substances. f Excludes employees who indicated that they do not use tobacco. * Correlation is statistically significant at P < 0.05.
between baseline and follow-up differed between intervention and control worksites. The treatment-by-time interaction was significant for health climate (F[1,36] 4 7.57, P 4 0.009). Examination of the means in Table 6 indicates that the health climate at intervention worksites increased from baseline to final measurement more than it did at control worksites. The main effect for time was also significant (F[1,36] 4 21.47, P 4 0.000), indicating overall health climate increased during the study. Safety climate at the intervention worksites remained unchanged between baseline and follow-up, but the control worksites showed a trend toward a decrease in safety climate. However, this interaction term was not significant (F[1,36] 4 3.57, P 4 0.067). For safety climate, the industry-by-time interaction was significant (F[1,36] 4 6.02, P 4 0.02), indicating a decrease in safety climate over time at the electrical cooperatives, but not at the gas pipeline worksites. No other interactions or main effects were significant.
The health and safety climate scales developed for this study are useful instruments for measuring organizational change related to worksite health promotion activities. The scales exhibited good internal consistency and were related to other conceptually relevant organizational constructs. Our study demonstrated that health and safety climate is related to other variables, indicative of an environment in which healthand safety-related behaviors are supported. These associations confirmed the convergent and divergent validity of the health and safety climate scales, in that TABLE 6 Intervention Effects on Health and Safety Climate, Worksite Level of Analysis (N 4 40 Worksites)
Health climatea Intervention Natural gas pipeline Electrical cooperative Control Natural gas pipeline Electrical cooperative Safety climateb Intervention Natural gas pipeline Electrical cooperative Control Natural gas pipeline Electrical cooperative
Baseline mean (SD)
Final mean (SD)
2.78 (0.24) 2.71 (0.15) 2.84 (0.29) 2.75 (0.20) 2.75 (0.18) 2.74 (0.23)
2.97 (0.26) 2.91 (0.24) 3.03 (0.28) 2.80 (0.24) 2.78 (0.23) 2.81 (0.27)
4.22 (0.23) 4.27 (0.15) 4.18 (0.29) 4.23 (0.27) 4.22 (0.24) 4.23 (0.31)
4.25 (0.24) 4.33 (0.16) 4.16 (0.28) 4.16 (0.29) 4.23 (0.28) 4.08 (0.29)
a Treatment group × time interaction is significant, F(1,36) 4 7.57, P 4 0.009; main effect for time is significant, F(1,36) 4 21.47, P 4 0.000. b Treatment group × time interaction is not significant, F(1,36) 4 3.57, P 4 0.067; industry × time interaction effect is significant, F(1,36) 4 6.02, P 4 0.02.
118
BASEN-ENGQUIST ET AL.
the health-related variables were more highly correlated with the health climate scale, and the safetyrelated variables were more highly correlated with the safety climate scale. The health climate scale was related to the presence of tobacco use cessation programs in the workplace at the end of the study, indicating that changes in organizational climate helped support the implementation of health promotion programs, as predicted by theory [7,8]. Concurrent validity was not established in the relationship of health and safety climate to employee health and safety behavior. Although the literature on organizational climate leads us to believe that employee behavior should be affected by climate, [16–19] our findings did not support this relationship. It is possible that such a relationship is found in other studies that focus on behaviors confined primarily to the workplace (e.g., absenteeism, productivity). Health and safety behaviors may be influenced to some extent by workplace climate, but they likely are affected more directly by nonworksite factors. For example, although a worksite smoking policy or smoking cessation program may alter smoking behavior somewhat, this behavior is probably influenced more heavily by other factors, such as level of addiction and smoking by significant others. The Working Well Intervention did not have a significant impact on smoking, [29] suggesting that smoking in particular is influenced heavily by nonworksite factors. The health and safety climate scales would benefit from additional validity and reliability testing. The design of the current study did not allow us to assess the scales’ test–retest reliability. Furthermore, our factor analysis was exploratory, and because several items were dropped to provide a good solution, the stability of the results is uncertain. However, the similar factor analytic solution produced by the follow-up data strengthens our confidence in the results. The scales would benefit from further testing on other populations and confirmatory factor analysis. The health climate scale was sufficiently sensitive to detect changes in the worksite environment. The health climate scores at worksites at which a cancer prevention intervention took place increased from baseline to follow-up, whereas scores at control worksites remained the same. The design of the intervention, which emphasized building organizational capacity for conducting health promotion programs, probably influenced this change. Because the intervention focused on diet and tobacco use, as opposed to safetyrelated behavior, we did not expect to see increases in safety climate. Even if the intervention had included more safety content, however, we might not have detected much change; the safety climate scores were quite high at baseline, and thus the scale may not have been sufficiently sensitive to measure subtle differences in the top range. This may be so because the
types of industries in which we worked are heavily regulated, and adherence to safety practices is a pervasive industry expectation. Greater variance on the safety climate scale scores might be found among other industries. A strength of this study is that there were a sufficient number of worksites to conduct worksite-level analyses. This is critical when answering research questions about organizational characteristics, [13] but is not done in some studies because too few worksites are studied. A second strength is that two sources of data were used for criterion variables in assessing concurrent validity; items from the same employee questionnaire that assessed climate and data from interview about health promotion activities with key informants within the worksites. We were somewhat limited in the type of data available from the employee survey that could be used to assess the concurrent validity. In selecting the criterion variables from the employee questionnaire, we attempted to identify variables that did not rely on individual perception, in order to distinguish them sufficiently from the climate construct. However, because there were few such variables on the employee survey we had to rely on more subjective criterion variables. Although the more subjective variables tended to have higher correlations with the scales than the more objective criteria, overall the findings support the concurrent validity of the scales. CONCLUSIONS
The health and safety climate scales had good internal consistency reliability, and their validity is supported by correlations between scale scores and conceptually related organizational characteristics. The lack of association between health and safety climate scores and health and safety behaviors was disappointing, but may be due to the multiple influences on health and safety behaviors or insufficient power caused by using the worksite as the unit of analysis. Two additional findings increased our confidence in the scales’ validity: the relationship between changes in health climate and the number of tobacco use programs offered and the increase in health climate scores among worksites receiving the cancer prevention intervention compared with control worksites. Health promotion theory stresses the importance of changes in the environment to support changes in behavior. Organizational theory also suggests the importance of environmental influences on the implementation of health promotion in the workplace. Despite this emphasis on organizational influences, measurement of environmental and organizational characteristics has lagged behind measurement of intrapersonal variables associated with change in health behavior. Our data suggest that this measurement tool can be useful
HEALTH AND SAFETY CLIMATE
for worksite health promotion programs, both for assessing receptivity to health promotion programs before the intervention and for assessing change over the course of a program.
14. 15.
REFERENCES 1. Bandura A. Social foundations of thought and action: a social cognitive theory. Englewood Cliffs (NJ): Prentice-Hall, 1986. 2. Rogers EM. Diffusion of innovations. New York: Free Press, 1983. 3. Witte K. Managerial style and health promotion programs. Soc Sci Med 1993;36:227–35. 4. Glasgow RE, Sorensen G, Corbett K. Worksite smoking control activities: prevalence and related worksite characteristics from the COMMIT study, 1990. Prev Med 1992;21:688–700. 5. Hollander RB, Lengermann JJ. Corporate characteristics and worksite health promotion programs: survey findings from Fortune 500 companies. Soc Sci Med 1988;26:491–501. 6. Glasgow RE, Hollis JF, Ary DV, Lando HA. Employee and organizational factors associated with participation in an incentivebased worksite smoking cessation program. J Behav Med 1990; 13:403–18. 7. Gottlieb NH, McLeroy KR. Social health. In: O’Donnell MP, Harris JS, editors. Health promotion in the workplace. Albany: Delmar Pub., 1994. 8. Allen J, Bellingham R. Building supportive cultural environments. In: O’Donnell MP, Harris JS, editors. Health promotion in the workplace. Albany: Delmar Pub., 1994. 9. Emont SL, Cummings KM. Organizational factors affecting participation in a smoking cessation program and abstinence among 68 auto dealerships. Am J Health Promot 1990;5:107–14. 10. Terborg JR. Organizational context. In: Johnson K, Larosa JH, Scheirer CJ, Wolle JM, editors. Methodological issues in worksite research: proceedings of a workshop, April 1988. Washington: U.S. Department of Health and Human Services, U.S. Public Health Service, 1988. 11. Rost K, Connell C, Schechtman K, Barzilai B, Fisher EB Jr. Predictors of employee involvement in a worksite health promotion program: Health Educ Q 1990;17:395–407. 12. Sloan RP, Gruman JC. Participation in workplace health promotion programs: the contribution of health and organizational factors. Health Educ Q 1988;15:269–88. 13. Ribisl KM, Reischl TM. Measuring the climate for health at or-
16. 17. 18. 19. 20. 21.
22.
119
ganizations: development of the worksite health climate scales. J Occup Med 1993;35:812–24. James LR, Jones AP. Organizational climate: a review of theory and research. Psychol Bull 1974;81:1096–112. Hoy WK, Tarter CJ, Kottkamp RB. Open schools/healthy schools: measuring organizational climate. Newbury Park: Sage, 1991. Forehand GA, Gilmer B. Environmental variation in studies of organizational behavior. Psychol Bull 1964;62:361–82. Field RHG, Abelson MA. Climate: a reconceptualization and proposed model. Hum Relat 1982;35:181–201. Moran ET, Volkwein JF. The cultural approach to the formation of organizational climate. Hum Relat 1992;45:19–47. Schneider B. Organizational climate: an essay. Pers Psychol 1975;28:447–479. Zohar D. Safety climate in industrial organizations: theoretical and applied implications. J Appl Psychol 1980;65:96–102. Brown RL, Holmes H. The use of a factor-analytic procedure for assessing the validity of an employee safety climate model. Accid Anal Prev 1986;18:455–70. Franklin J. Down the organization: influence processes across levels of hierarchy. Admin Sci Q 1975;20:153–64.
23. Scofield ME, Martin W. Development of the AT&T health audit for measuring organizational health. State Art Rev Occup Med 1990;5:755–88. 24. Brooks L. Employee perceptions of the worksite health climate at a large company [thesis]. Houston: Univ. of Texas School of Public Health, 1995. 25. Abrams DB, Boutwell WB, Grizzle J, Heimendinger J, Sorensen G, Varnes J. Cancer control at the workplace: the Working Well Trial. Prev Med 1994;23:1–13. 26. Heimendinger J, Feng Z, Emmons K, Stoddard A, Kinne S, Biener L, Sorenson G, Abrams D, Varnes J, Boutwell B. The Working Well Trial: baseline dietary and smoking behaviors of employees and related worksite characteristics. Prev Med 1995;24: 180–93. 27. Nunnally JC, Bernstein IH. Psychometric theory. New York: McGraw–Hill, 1994. 28. Gorsuch RL. Factor analysis. Hillsdale (NJ): Erlbaum, 1983. 29. Sorensen G, Thompson B, Glanz K, Feng Z, Kinne S, DiClemente C, Emmons K, Heimendinger J, Probart C, Lichtenstein E. Work site-based cancer prevention: primary results from the Working Well Trial. Am J Public Health 1996;86:939–47.