SEARCH

SEARCH BY CITATION

Keywords:

  • health surveys;
  • workforce participation;
  • employment;
  • absenteeism;
  • cost of illness

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

Objective: To describe the relationship between obesity class and workforce participation and the influence of demographic, socioeconomic, and comorbid disease states on this relationship using population-based Canadian data.

Research Methods and Procedures: Responses from 73, 531 adults surveyed in the Canadian Community Health Survey 2000 to 2001 who provided complete information regarding variables of interest were analyzed. Workforce participation was defined as individuals reporting that they held and were present at a job or business in the week before survey administration. The association between obesity and workforce participation was explored using logistic regression after adjusting for demographic, socioeconomic, and obesity-related comorbidities.

Results: In univariate analysis, obese individuals had lower odds of participating in the workforce. In the fully adjusted model, increasing obesity was associated with decreasing odds of workforce participation, with Class I, II, and III obesity having odds ratios (95% confidence interval) of 0.94 (0.89 to 0.99), 0.85 (0.77 to 0.94), and 0.66 (0.57 to 0.78), respectively. Obese individuals were also less likely to be employed and more likely to be absent from work.

Discussion: Obesity is associated with lower workforce participation. This association appears to be independent of associated comorbidity and sociodemographic factors. These results indicate that the economic impact of obesity alone on workforce productivity is larger than previous reports suggest.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

The prevalence of obesity and severe obesity continues to increase throughout developed and developing nations (1, 2, 3). Obesity causes significant morbidity and mortality (4, 5) and has been described as a major public health crisis (1, 3, 6, 7, 8). In addition to the associated deleterious impact on health, the economic impact of obesity is substantial. It has been estimated that 6% to 10% of national health care expenditures are attributable to obesity (1, 9, 10, 11, 12, 13), with societal costs estimated to be greater than those attributable to cigarette smoking and alcoholism (14). The economic impact will likely continue to increase as the prevalence of this disorder rises.

Although recent work has concentrated on the direct medical costs of obesity, less is known regarding the impact of obesity on labor market outcomes, which may have important implications for indirect costs. Obesity has been associated with increased absenteeism in selected populations (15, 16, 17, 18). In addition, compared with their non-obese counterparts, obese persons may incur a wage penalty of 0.7% to 6.3% (19). In part, this may indicate lower productivity rates for obese workers.

Little is known about the relationship between obesity and overall workforce participation on a population-wide basis. If increasing levels of obesity predict decreased workforce participation, the indirect economic losses as a result of obesity may be substantial and the overall economic costs of obesity markedly underestimated. Because workforce non-participation and absenteeism have been associated with impaired health status and performance in other functional spheres (20, 21), obesity may also impart sweeping implications for home and personal productivity.

We hypothesized that increasing degrees of obesity would be associated with decreasing workforce participation (defined as encompassing both non-employment and absenteeism) and that this relationship would be independent of obesity-related comorbidities. We utilized the population-based Canadian Community Health Survey (CCHS)1 to study the relationship between obesity and workforce participation over a 1-week time frame. We also examined whether obesity was associated with a lower risk of employment and a higher risk of absenteeism on a national level.

Research Methods and Procedures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

Subjects for this analysis were obtained from the CCHS 2000 to 2001, a population-based household survey of non-institutionalized persons from the Canadian population conducted by Statistics Canada (22). The primary objective of this survey is to provide timely cross-sectional estimates of health determinants, health status, and health system use at a subprovincial level. Respondents to this survey provided detailed information regarding socioeconomic status, demographics, medical status, labor force participation, and self-reported height and weight. This survey uses a multistage stratified cluster design, provides cross-sectional data representative of 98% of the Canadian population over the age of 12, and attained an 80% overall response rate (22).

We selected a subsample appropriate for testing the hypothesis that obesity is associated with workforce participation, by analyzing respondents between the ages of 20 and 59 (23), who provided self-reported height and weight. All survey respondents were asked questions regarding their activity in the labor force. We used responses from the questions “last week, did you work at a job or business (including part-time jobs, seasonal work, contract work, self-employment, baby-sitting and any other paid work, regardless of the number of hours worked)?” and “last week, did you have a job or business from which you were absent?” Workforce participation was defined as individuals indicating that they both worked at a job or business and were present at that job for the week before survey administration. Employment status was defined as those who indicated they were employed in the previous week, regardless of whether they were actually present at their job in the week before survey administration. Absenteeism was defined as those who indicated that they were employed but were absent from their work in the previous week. Subjects with missing data for the variables of interest were excluded from the analysis (Figure 1).

image

Figure 1. Overview of definition of sample for analysis from the CCHS.

Download figure to PowerPoint

A labor force participation model was used, which assumes that a set of demographic, economic, and health status variables influence an individual's incentive or ability to work (24, 25, 26, 27). These variables were identified a priori, with specific emphasis on identifying those comorbid conditions that may be associated with obesity and impact employment status. The relationship between individual variables and workforce participation was determined by a series of bivariate analyses. Variables that were statistically significant (p < 0.05) on bivariate analysis were then included in subsequent regression models.

A set of binary logistic regression equations were constructed using workforce participation status as the dependent variable in each model. The demographic and socioeconomic variables considered included age (20 to 29, 30 to 39, 40 to 49, and 50 to 59 years old), gender, ethnic origin, region of residence, marital status, level of education, and obesity-related comorbidities (type 2 diabetes, hypertension, heart disease, cancer, and stroke) (28, 29, 30, 31). Depression was also considered and was defined as being present if the probability of major depression was ≥80% based on the Composite International Diagnostic Interview (32).

Obesity was calculated from self-reported height and weight and categorized according to the World Health Organization criteria (1, 33) into Classes I to III (BMI, 30 to 34.9, 35 to 39.9, and ≥40 kg/m2, respectively). An alternate classification considered those subjects with Class II obesity or greater. To account for the non-equal probability of selection in the CCHS due to the complex sampling design, sample weights were applied to obtain population-based estimates. All statistical tests were performed using SPSS 13.0 (SPSS Inc., Chicago, IL).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

From the full study sample (n = 130, 880), 81, 938 individuals were between the ages of 20 and 59. Of these, 73, 531 provided complete information for both workforce participation status and other variables of interest. Complete data for absenteeism were available in 58, 289 persons (Figure 1). Characteristics of the individuals in the workforce participation data set are shown in Table 1, stratified by obesity class. The fraction of individuals who participated in the workforce decreased with increasing obesity class.

Table 1. . Distribution of demographic, socioeconomic, and health-related characteristics of participants by BMI category
 n = 73, 531
CharacteristicsNormal (18.5 to 24.9)Class I (30 to 34.9)Class II (35 to 39.9)Class III (≥40)
Employment status/absenteeism    
 Employed79.8179.2574.7766.53
 Absent from work (58, 289)6.557.487.978.46
Age (years)    
 20 to 2928.4115.2417.2214.46
 30 to 3927.3825.8725.7424.90
 40 to 4927.5731.1830.2035.47
 50 to 5916.6427.7026.8525.17
Gender    
 Men44.1757.7849.8937.55
 Women55.8342.2250.1162.45
Ethnicity    
 White82.4790.5492.4293.88
 Visible minority17.539.467.586.12
Marital status    
 Married or common law62.0871.9866.8357.92
 Widowed, separated, or divorced9.409.2211.7714.03
 Single28.5218.8021.4028.06
Education    
 Less than high school13.1320.4722.3825.17
 High school or greater86.8779.5377.6274.83
Diabetes    
 Present1.255.739.7613.19
 Absent98.7594.2790.2486.81
High blood pressure    
 Present4.2816.9822.9628.09
 Absent95.7283.0277.0471.91
Effects from stroke    
 Present0.260.650.450.56
 Absent99.7499.3599.5599.44
Heart disease    
 Present1.583.395.625.14
 Absent98.4296.6194.3894.86
Cancer    
 Present0.850.901.601.95
 Absent99.1599.1098.4098.05
Depression    
 Present5.031.130.450.18
 Absent90.1190.0985.3881.50

All variables except for region of residence achieved statistical significance in bivariate analysis. In unadjusted logistic regression analyses, with normal BMI class of 18.5 to 24.9 kg/m2 used as the reference category, increasing severity of obesity was associated with decreasing odds of workforce participation. Odds ratios were 0.94, 0.75, and 0.53 for Classes I to III obesity, respectively (Table 2). This statistically significant, graded relationship persisted for all classes of obesity when demographic and socioeconomic parameters were included as covariates in the model (Table 2).

Table 2. . Unadjusted and partially adjusted logistic regression for workforce participation
 (n = 73, 531)
VariablesOR95% CI
  • OR, odds ratio; CI, confidence interval.

  • *

    Adjusted for age, gender, ethnic origin, marital status, and education.

  • Normal weight (18.5 to 24.9), Class I (30 to 34.5), Class II (35 to 39.9), and Class III (≥40).

Unadjusted  
 Obesity (reference: normal weight)  
  Class I obesity0.940.89 to 0.99
  Class II obesity0.750.69 to 0.82
  Class III obesity0.530.46 to 0.62
Adjusted*  
 Obesity (reference: normal weight)  
  Class I obesity0.890.84 to 0.94
  Class II obesity0.760.69 to 0.84
  Class III obesity0.580.50 to 0.68

When demographic, socioeconomic, and health characteristic covariates were included in the full logistic regression model, persons in all classes of obesity continued to have significantly diminished odds of workforce participation (Table 3).

Table 3. . Logistic regression for workforce participation
 (n = 73, 531)
VariablesOR95% CI
  • OR, odds ratio; CI, confidence interval.

  • *

    Normal weight (18.5 to 24.9), Class I (30 to 34.5), Class II (35 to 39.9), and Class III (≥40).

Obesity* (reference: normal weight)  
 Class I obesity0.940.89 to 0.99
 Class II obesity0.850.77 to 0.94
 Class III obesity0.660.57 to 0.78
Age in years (reference: 20 to 29)  
 30 to 391.241.18 to 1.31
 40 to 491.361.28 to 1.43
 50 to 590.750.71 to 0.80
Gender (reference: men)  
 Women0.460.44 to 0.48
Ethnic origin (reference: white)  
 Visible minority0.690.66 to 0.73
Marital status (reference: married or common law)  
 Widowed, separated, or divorced0.940.89 to 1.00
 Single0.830.79 to 0.87
Education (reference: less than high school)  
 High school or greater2.222.12 to 2.32
Diabetes (reference: absent)0.700.63 to 0.77
High blood pressure (reference: absent)0.880.83 to 0.94
Effects from stroke (reference: absent)0.300.23 to 0.40
Heart disease (reference: absent)0.460.41 to 0.51
Cancer (reference: absent)0.560.48 to 0.65
Depression (reference: absent)0.590.56 to 0.62

A fully adjusted logistic regression analysis of the relationship between employment status and obesity severity resulted in similar point estimates, although statistical significance was present for only Class II and III obesity (Table 4). A separate logistic regression model examining the relationship between obesity and absenteeism found significantly higher odds of absenteeism in those with Class I obesity (Table 4). An alternate classification of BMI resulted in an odds ratio of 1.17 (95% confidence interval 0.99 to 1.38) for absenteeism for those with a BMI of ≥35 kg/m2.

Table 4. . Fully adjusted* logistic regression for non-employment status and absenteeism
 Employment (n = 73, 531)Absenteeism (n = 58, 289)
VariablesOR95% CIOR95% CI
  • OR, odds ratio; CI, confidence interval.

  • *

    Adjusted for age, gender, ethnic origin, marital status, education, diabetes, high blood pressure, stroke, heart disease, cancer, and depression.

  • Normal weight (18.5 to 24.9), Class I (30 to 34.5), Class II (35 to 39.9), and Class III (≥40).

Obesity (reference: normal weight)    
 Class I obesity0.970.91 to 1.031.151.03 to 1.27
 Class II obesity0.860.77 to 0.951.150.95 to 1.38
 Class III obesity0.640.54 to 0.761.240.89 to 1.73

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

This population-based analysis demonstrates that increasing severity of obesity is associated with decreasing odds of actively participating in the workforce. This association is independent of other demographic, socioeconomic, and health conditions, including chronic medical conditions that may be influenced by the presence of obesity, such as type 2 diabetes and cardiovascular disease. To our knowledge, the association of obesity and workforce participation and employment status has not been described on a population basis.

Our analysis was confined to workforce participation, an economic activity of the adult population. Decreasing involvement in the workforce has been associated with health status and capacity in other functional spheres (20, 21); thus, it is likely that severe obesity influences more than simply paid labor activities. The overall effects of obesity on function and personal productivity, of which we have described only the employment effects, could be very substantial.

Our point estimates are concordant with other findings of workplace absenteeism. Other investigators have described the association between BMI and absenteeism or sick days in employed populations and found that increasing BMI is associated with increased absenteeism (15, 16, 17). A 15-year follow-up of 19, 518 Finnish persons reported 0.63 and 0.52 more years of work disability in men and women, respectively, as defined by receiving a work disability pension (18). However, many of these studies did not account for the impact of other medical conditions, suffered from a high non-response rate, or studied only select populations. Our analysis supports these findings of increased absenteeism from a population perspective, although it suffers from a very short time frame for assessment of absenteeism. For this reason, the analysis may be underpowered to demonstrate an association, despite the large sample size.

Although obesity contributes to the development of multiple serious chronic medical disorders, many investigators have not controlled for these associated medical conditions when assessing the attributable direct health care costs of obesity (34). Others have used mathematical modeling based on attributable risk to calculate direct medical costs (35). However, in a recent analysis of direct medical costs in obesity that controlled for chronic disease, it was found that obesity remained an independent predictor of medical expenditures (36). Here, we performed two analyses, one in which the associated medical conditions were included in the model and one in which they were not. The latter model likely overestimates the influence of obesity on workforce participation. In contradistinction to other analyses of direct costs, where much of the excess resource use is attributable to chronic disease states that develop as a result of obesity, particularly type 2 diabetes and cardiovascular disease (35, 37, 38), the relationship persisted even when these conditions were accounted for. In fact, the independent association of Class III obesity on workforce participation is similar in magnitude to that of type 2 diabetes.

Use of cross-sectional databases allows for observations and associations but does not establish causation or temporal sequence or allow exploration of mechanisms of effect. We speculate that many factors may be responsible for the association of obesity and workforce participation, aside from other chronic illness and depression that may be associated with obesity. As explored by Baum et al. (19) in their analysis of wage rate disparity in obesity, there may be internal and external factors contributing. The persistence of the effect after controlling for age and education lessens the likelihood that employer discrimination on these grounds plays a significant role. Obesity per se may also impact worker productivity independently of the development of associated disease states such as diabetes and cardiovascular disease. This may be mediated, perhaps, through capacity for physical exertion or musculoskeletal disease that may play a role in the case of those jobs requiring exertion or manual labor. Although acute disorders were not reported in this survey, the inclusion of chronic musculoskeletal disorders including osteoarthritis and back pain as a covariate in a secondary analysis did not substantively alter the results. Personal preferences may also play a role, and it has been hypothesized that those who are obese may prefer leisure activities over work activities (19), which may influence both the development of obesity and worker training and effort to seek employment. Finally, external factors such as discrimination in hiring may also contribute to our findings.

Other investigators have reported that obesity is associated with lower health-related quality of life in both mental and physical domains (39, 40), even when underlying comorbidities are controlled for (41). This decrement in mental and physical health has been found to influence the frequency of days with activity limitations (42). Our findings of employment status related to obesity alone support the hypothesis that obesity itself may have a negative impact on functional status, independently of other chronic medical conditions.

Confidence in the validity of a model can be increased if its results are intuitive and supported by other studies. Non-employment has been found to be more common in those of increasing age, female gender, non-white ethnic origin, single, and lower educational attainment (43, 44). Similarly, type 2 diabetes has been associated with non-employment in several studies (44, 45, 46, 47). In a recent review, return to work after stroke occurred in 11% to 85% of patients (48) and is predicted with greater accuracy when such variables as severity and location of injury are considered. Although a careful attempt was made to control for medical conditions that may be associated with obesity and also impact on an individual's ability to work, we could not control for the severity of these conditions given the limitations of health survey questions. Nevertheless, the consistency of the results of this model with other published data increases its credibility.

Although we have found a considerable association of workforce participation and obesity, its effects are most marked with severe obesity, particularly Class III. Using a conservative assumption of an 18-hour work week, a minimum wage rate of $6.00 per hour (49), an estimated prevalence of Class III obesity of 1% in adults ages 20 to 59 (calculated from the CCHS 2000 to 2001), and current demographic data (50), the total lost productivity over a 1-year time frame due to workforce non-participation in 2004 for Class III obesity alone may be as high as $187 million (Canadian). Although the fraction of the population with severe obesity is at present relatively small, it is important to point out that the proportion of those with severe obesity is increasing faster than obesity itself (51, 52). Over a 14-year time frame, Class III obesity quadrupled in the U.S. population, whereas all classes of obesity doubled (51). If this trend continues, these findings may have relevance for increasingly large fractions of the population, with the associated productivity cost implications.

The survey data used for this analysis have limitations. It has been observed that self-reported weight and height are lower and higher, respectively, than objective measures of the same (53), leading to systematically lower BMI. This would tend to bias toward the null; thus, the association described here may an underestimate. Also, the data set analyzed excludes institutionalized adults, which may bias the results. Errors may also exist for participant self-reported chronic medical conditions, although other investigators have found that self-reported disease and data from the medical record correlate well, with κ values of 0.73 to 0.80 for such illnesses as cardiovascular disease, hypertension, myocardial infarction, and diabetes (54). Although non-response was relatively low, this may also impart systematic bias to this analysis. Furthermore, the employment effects may be greater than we have found because we were unable to account for the effect of decreased lifespan (55, 56) on workforce participation. Finally, the time frame for sampling for workforce activity was very short, which may bias the results to the null. However, an analysis of obesity class and workforce activity defined by responses to a survey question assessing workforce participation in the previous 12 months resulted in similar point estimates and statistical significance in a fully adjusted model (data not shown).

To conclude, we report, using population-based survey data, that increasing severity of obesity is associated with decreased participation in the workforce. This association persists after accounting for demographic, socioeconomic, and other important related chronic medical conditions such as type 2 diabetes and cardiovascular disease. This finding has implications for the potential impact of interventions aimed at reducing obesity with specific reference to indirect productivity costs.

Acknowledgement

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

This work was supported by the Institute of Health Economics and the University of Alberta. This project utilized a data set made publicly available by Statistics Canada; however, the opinions expressed herein do not necessarily represent the views of Statistics Canada. The authors accept sole responsibility for study design, analyses, and conclusions.

Footnotes
  • 1

    Nonstandard abbreviation: CCHS, Canadian Community Health Survey.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References