The health care worker cohort

The cohort is based on a survey of eldercare service employees conducted in 2004–5 in 35 Danish municipalities. All employees (12,744) received a written questionnaire that was administered by a local contact person in each municipality; 9,495 responded, a response rate of 75%. We excluded respondents who didn't report their gender, underweight workers, and administrative staff. We also excluded workers aged 30 years or younger because of the greater risk of type 1 diabetes (Figure 1). The remaining 7,305 health care workers were linked to disease and population registers via a unique personal identification number and followed for seven years. Detailed information on most of the measures has been reported elsewhere [18]–[20].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Participant Flow Chart.

https://doi.org/10.1371/journal.pone.0103425.g001

The National Diabetes Register

In 2006 Denmark established a National Diabetes Register (DNDR) that includes all incidences of diabetes since 1995 [26]. DNDR combines data from three different national registries: the National Patient Register (NPR) [27], the National Health Insurance Service Registry (NHISR) [28] and the Register of Medicinal Product Statistics (RMPS) [29]. Individuals must meet at least one of six criteria to be registered in DNDR: 1) receive a diagnosis of diabetes (data source, NPR); 2) receive diabetes-related chiropody care (NHISR); 3) have blood glucose measured five times in a calendar year (NHISR); 4) have at least two annual blood glucose measurements over a period of five consecutive years (NHISR); 5) purchase prescribed oral anti-diabetic medication (registered at second purchase, RMPS); 6) purchase prescribed insulin (RMPS). In addition to the unique personal identification number of each individual, DNDR contains data on sex, date of birth, date of death, and date of inclusion according to the entry criteria. However, the register does not distinguish between type 1 and type 2 diabetes. DNDR has been validated and found to be reliable as a source for linkable information on diabetes at a national level [15], [30].

Demographic variables

Age, information about ethnicity and household type was drawn from Statistics Denmark [31] through linking with the Danish Civil Registration System [32].

Measures of health

All measures of health originated from the baseline questionnaires, with the exception of diabetes data from DNDR. General health was self-assessed on a five-point scale ranging from ‘poor’ to ‘outstanding’ and dichotomised as poor if respondents selected one of the two lowest scores and good otherwise. Sleep problems were identified if at least one of five questions was scored higher than three on a six-point scale. Participants separately rated pain in the low back, neck/shoulders and knees as average pain during the last three months on a numerical rating scale from 0–9, where 0 represented ‘no pain’ and 9 represents ‘worst imaginable pain’. Based on the distribution of the rating, pain was categorised in four groups to make it easier to interpret than if using the continuous scale [33]. Depression was evaluated using the Danish version of the Major Depression Inventory (MDI), and the WHO-5 well-being scale was used to assess the risk for developing depression and stress [20], [34].

Measures of job-related factors

Working hours were self-reported in a single item. Ergonomic strain was assessed with the Hollmann index, a compound measure of body postures and weight lifted during the work day [20]. Strain was stratified in four categories based on the distribution of the continuous compound measure scale. Work-life balance was defined as poor if respondents reported that work was draining both time and energy from their private lives. Assessment of psychosocial working conditions used the same model as in a long-term sickness absenteeism survey [19], in which four scales—emotional demands, influence, role conflict and quality of leadership—were based on four items from the Copenhagen Psychosocial Questionnaire (COPSOQ) [35].

Measures of lifestyle

All lifestyle variables were self-assessed in the questionnaire. Leisure time physical activity was classified as good if respondents reported they were very active or fairly active. Physical capacity was rated across the domains of physical fitness, strength, endurance and body movement as the average score on a nine-point scale from worst (1) to best (9). On the basis of the distribution of the scores, the authors decided to stratify into four categories. Smokers were those reporting daily smoking. Finally Body Mass Index (BMI) was categorised as normal weight (18.5≤BMI<25), overweight (25≤BMI<30) and obese (BMI≥30).

The study was approved by the Danish Data Protection Agency. In Denmark, however, approval from Ethics Committees is not required for survey research. Furthermore, participation in the study was voluntary which implies that written informed consent is not necessary as this consent is implied in individual respondents' voluntary participation.

Statistical analyses

Initially, unadjusted bivariate (2×n) analyses compared each variable with the incidence of diabetes using Chi square analysis for trend. All variables were then included in a saturated Cox regression model with diabetes as the dependent variable using SAS Proc Genmod; this was used as the basis for a subsequent manual backward elimination based on log likelihood ratio tests to correct for multiple testing. A similar analytic approach followed, in which obesity at baseline was the dependent variable. However, since this was a cross-sectional design, the Logistic procedure in SAS 9.3 was used for a multiple logistic regression analysis. Manual backwards elimination using the log likelihood ratio test was used to fit the model. The Hosmer and Lemeshow goodness-of-fit test was used to evaluate the logistic analyses [36].

Diabetes and work

Pain was significantly associated with an increased diabetes risk in the bivariate analysis. In the same cohort, it has been shown that the more pain health care workers have, the higher the long-term sickness absenteeism [33], [37]. After controlling for lifestyle, however, we found little impact from occupational risk factors for developing diabetes. Therefore, if health care workers have so much musculoskeletal pain or are so drained of energy by their work that they cannot stay fit, it will have significant indirect effects on diabetes. The same factors are also expected to cause selection out of work.

Shift work was another factor that increased the risk of developing diabetes in our study. Potential explanations for this finding include the difficulty of maintaining a healthy lifestyle, e.g., a tendency to eat unhealthy foods and exercise less, and effects on the endocrine circadian rhythm [38]–[40]. Although a review of the epidemiological evidence found only moderate support for an association between shift work and metabolic syndrome and limited evidence for a relationship to diabetes [41], those with shift work who already had diabetes experienced more problems maintaining glycaemic control [42].

Interestingly, we found that quality of leadership influenced diabetes risk, but only high-quality leadership carried this effect, referencing those with medium-quality leadership; this did not hold true for low-quality leadership. A potential explanation is that high-quality leadership is important if people with diabetes have experienced support for increased sickness absence or health problems before they were diagnosed with diabetes, i.e. the opposite of the healthy worker effect [43]. More research is needed in this area because it is important to know how management could help employees with diabetes remain productive workers [9]. Alternatively, high-quality leadership may lead to over-commitment [44], which may again entail adverse health outcomes, such as diabetes.

There is scientific support that chronic stress—or, especially, lack of control—might cause diabetes among females. We thought that this cohort would provide insight into the strength of the stress-diabetes link [21]–[23], not least because this cohort has elsewhere identified that psychosocial demands and depression played a role in long-term sickness absenteeism [19]. However, our findings did not support this theory.

Obesity and work

Our results suggest that the occupational setting affects diabetes indirectly through the influence of obesity and physical inactivity. Referring to diabetes as a ‘lifestyle disease’ is misleading, implying that there is something wrong with the character of the employee, when the truth is that what we call ‘lifestyle’ is a combination of structural conditions for health and capacities for coping with both living and working conditions. Based on the present study, the occupational setting may be more important for the prevention of diabetes than previously thought. Our survey confirms that occupational risk factors were associated with obesity to a much higher extent than was seen for diabetes.

Another challenge is that the causes of obesity are tightly interwoven with parallel factors from both working and non-working life. Traditionally, obesity has not been regarded as an occupational health and safety (OHS) issue, but as a matter for workplace health promotion (WHP). However, more than thirty years of WHP have not been able to reverse the rising prevalence of obesity, and initiatives addressing nutrition and physical activities have only showed modest effect [45]–[47]. Lifestyle is not simply a confounder to occupational risk factors; it is both a causal factor and mediator because work has become so much more sedentary [48]. One would have expected health care workers to be in relatively good physical shape because they walk, lift, move a lot and generally have physically demanding work [18], [20]. However, new research shows that the majority of people who are considered to have strenuous physical work may still not reach activity intensity levels that could improve physical capacity [49].

T2DM has hitherto been considered as a condition only affecting the elderly and caused by factors related to individuals' private lives; it has, therefore, viewed as largely irrelevant to OHS. This view will need to be revised as both obesity and diabetes become more prevalent within the working population. Even a small increase in diabetes prevalence will have considerable impact on the labour market. The gradually increasing pension age means that diabetes will occur in many more people while they are still active in the labour market [50]. Instead of continuing with WHP alone, the challenge will be to investigate if it is possible to organise work itself to support and foster health promotion. If so, the occupational arena may be able to contribute much more to the improvement of public health than is the case today.

Strengths and limitations

A study limitation is that all baseline measures, including BMI, were based on self-report. However, self-report is expected to provide a reasonably reliable, albeit conservative, categorisation of obesity vs. non obesity [51]. A second limitation is that the DNDR does not distinguish between type 1 and type 2 diabetes; however, we estimate this had a minimal effect on our findings. Due to the low number of males in the survey, the results might not be generalizable to similar jobs with a different gender distribution. The dichotomous measure of ethnicity is probably too simple, since we know that Non-western populations have a higher diabetes risk than people of Western or Danish origin. However, the relatively low number of ethnic minority health workers hindered further stratification. One of the strengths of the study is the large sample size and high response rate. In addition, our ability to link high quality registers covering the entire Danish population with individually based information on work and lifestyle is unique. The homogenous population is also a strength of the study because it reduces the potential for confounding from the social gradient of health.