Kowlessar, Niranjana M. PhD; Goetzel, Ron Z. PhD; Carls, Ginger Smith PhD; Tabrizi, Maryam J. MS, CHES; Guindon, Arlene MPH
* Demonstrate familiarity with previous research data on the financial burden of modifiable health risks for employers.
* Summarize the new findings on how specific health risks contribute to medical and productivity costs.
* Discuss how the new findings could help in targeting worksite health promotion programs, including the health risks likely to lead to cost savings.
Several prior studies have shown that modifiable health risk factors impose a substantial financial burden on employers.1–4 Both the number and type of health risks have been found to correlate directly with future medical and pharmacy claims costs.2,4,5 Anderson et al1 found that approximately 25% of total health care costs can be attributed to 10 modifiable risk factors among six large employers. Another study that evaluated the impact of clusters of risk factors established higher medical care costs for employees with high biometric laboratory values and poor emotional health.6
Studies that have evaluated the impact of health risks on workplace productivity including absenteeism and presenteeism have estimated annual costs due to lost productive time at $1392 to $2592 per employee at risk.7 The presence of multiple risk factors in employees has also been shown to affect the magnitude of productivity losses.7 Boles et al8 found that employees with multiple risk factors reported greater productivity losses, with a mean percentage of annual absenteeism losses ranging from 0.0% to 6.3% reported for employees with 0 to 8 health risks, respectively, and a mean percentage of presenteeism losses ranging from 1.3% to 25.9%, respectively. Short self-report tools such as the Work Limitations Questionnaire (WLQ) have proven effective in measuring these losses.9 Other studies have also found significant associations between health risks and diminished worker productivity.4,10
There is growing evidence that worksite health promotion programs can improve employee health, reduce health-related productivity losses, and generate a positive return on investment.11–13 As employers develop, evaluate, and justify their budgets for population health management initiatives, they routinely seek to quantify the financial value or return that such initiatives will produce for the organization. Insight into the costs of modifiable health risk factors has been extremely powerful in building a business case for health improvement investments within an organization.
This study examines the Mayo Clinic health assessment14,15 (referred to as “health assessment” throughout this article) instrument in examining the relationship between health risks and costs for a large employer. Comparing results of this analysis to those of prior studies can help researchers and employers better estimate the economic correlates of certain common health risks. Furthermore, the analysis presented here offers alternative risk reduction scenarios that are likely to produce medical cost savings and productivity improvements, which in turn would enable employers to fine-tune their worksite health promotion programs to achieve the largest financial impact.
This is a cross-sectional analysis using descriptive and multivariate methods to measure the relationships between 11 modifiable health risks and three financial outcomes for a large US employer. The three outcomes are medical care/pharmaceutical costs, presenteeism (on-the-job health-related work productivity loss), and incidental absences. We report the prevalence of each of the 11 measured health risks and estimate the impact of individual risks on costs using multivariate models that hold other risks and demographic variables constant.
Subjects in this study were employees of a large multinational company based in the Midwestern United States. In an effort to create a world-class wellness program, reinforce a genuine commitment to its workforce, contribute to the health of its business, and further distinguish itself as an employer of choice, this employer created a multipronged wellness program aimed at increasing employee engagement, improving their health, moderating health care cost trends, and increasing worker productivity. In addition to administering a health assessment, the Mayo Clinic provided health promotion programs to workers that included access to a Web portal, telephonic lifestyle coaching, and health newsletter.
Medical and pharmaceutical claims data for the study population were extracted from the Thomson Reuters MarketScan® Commercial Claims and Encounters Database. Health assessment data were obtained from the Mayo Clinic. These data were combined into a comprehensive analytic file spanning 3 years (January 1, 2005, to December 31, 2008).
The target population for this study was active employees aged 18 to 64 years. Inactive employees and employee spouses, partners, and dependents were not included in the analysis. We also excluded employees who did not complete a health assessment during the study period, and pregnant women, who may have especially high costs due to labor and delivery. These women were identified by searching medical claims for pregnancy-related costs as well from self-reported pregnancy status on the health assessment. Employees were required to have at least 12 months of eligibility for health care coverage during the calendar year that the health assessment was completed. Employees in capitated health maintenance organization plans were excluded from the medical care cost analysis but were retained for the absenteeism and presenteeism studies. The final study sample included 77,410 employees. Excluding employees enrolled in capitated plans for the medical care cost analysis resulted in a sample of 63,013 employees. Employees who completed the presenteeism question (only asked in the 2008 health assessment) totaled 38,298.
The first health assessment completed in each calendar year by the employee was included in the study; for example, employees who completed a health assessment in 2005 and 2006 would have had two assessments examined in the study, whereas employees who completed a health assessment in 2005, 2006, and 2007 would have had three. We excluded a small number of health assessments (<1%) that were completed in the same calendar year as a previous assessment.
Outcomes examined included measures of medical care costs, presenteeism, and absenteeism. Specifically, we examined the total allowed amounts recorded for medical care and pharmaceuticals claims, measured during the calendar year during which the health assessment was completed. Total allowed costs included employee co-payments and deductibles as well as payments made by the employer. Average annual total allowed amounts were $3901; the median annual total allowed amounts were $1492; and the standard deviation was $9840. We did not exclude outliers in our analysis.
Presenteeism was measured using the WLQ, which was included as part of the health assessment for the employer in 2008. The WLQ was scored to report the percent of time unproductive at work. We converted percent unproductive time to missed workdays by multiplying the percentage amount by the typical number of paid workdays each year (250). Missed workdays were monetized by determining the average annual value of compensation and benefits paid to the company's employees, which was $80,000 a year or $320 a day.
Data on absences were obtained from self-reported health-related missed workdays collected by the health assessments each year. Since some response categories were in ranges of days (eg, 10 to 15 days), we imputed the midpoint of each selected range. We multiplied the imputed number of absent days by the average employee daily compensation ($320), to calculate the annual absenteeism cost for each employee.
We examined the following 11 health risks based on the Mayo Clinic risk definitions: (1) overweight/obesity, (2) high blood pressure, (3) high blood glucose, (4) high cholesterol, (5) inadequate exercise, (6) poor nutrition, (7) poor emotional health, (8) high triglycerides, (9) poor safety practices, (10) tobacco use, and (11) high alcohol consumption. Each of the 11 health risks was dichotomized into high risk (“1”) versus lower risk (“0”). For risks with missing information on risk status (likely because of the employee not completing the question(s) required to define the risk), we created variables that indicated “missing” or “unknown” risk status. Risk definitions are provided in the appendix.
The employer in this study launched its first online health assessment in 2004 and offered voluntary follow-up coaching for employees at risk for nutrition, exercise, weight, or stress. Over the next several years, the company expanded its health management efforts by broadening the online offering of health assessments to spouses, providing supplemental information campaigns that targeted different learning styles of participants, assessing readiness to change, accelerated targeted telephonic outreach to those at risk, and incorporated incentives to encourage participation and engagement. The overall participation rates in the health assessment increased 32% over previous levels, resulting in approximately 50% of the eligible population completing a health assessment.
The Mayo Clinic health assessment is designed to provide individual feedback on one's health status to prepare and motivate individuals to move toward action. The health assessment has, as its scientific foundation, nationally recognized guidelines from Healthy People 2010, US Surgeon Generals' reports, and updated health risk guidelines from the National Institutes of Health and the American Diabetes Association. These guidelines are translated into questions and algorithms by a team of clinical and behavioral experts, and the tool is under guidance and ongoing review by Mayo Clinic medical staff.
The Mayo Clinic health assessment is driven by sophisticated branching logic that considers gender, age, and risk. A follow-up action plan is tailored by risk level and readiness to change for both messaging and links to relevant resources and targeted health management programs.
We extracted demographic variables including age, gender, and region of residence (Northeast, North Central/Midwest, South, West, and unknown region) from the medical care eligibility files. We then included the following variables in the multivariate analyses to control for potential confounding: age, gender, geographic location, type of health plan, job type (union or nonunion, salaried, or hourly), and year of data (2005 to 2008).
Health assessment data were pooled for each year and linked to medical care costs in the same calendar year; thus, some employees could have had multiple health assessments across different study years. We performed descriptive studies to examine the distribution of baseline characteristics, health risks, and costs for medical care services, presenteeism, and absenteeism. We then reported the prevalence of health risks for the study sample and calculated the average annual costs associated with each health risk. We then performed additional descriptive studies, weighting the data to account for multiple health assessments completed by some employees. Thus, the unit of analysis for the weighted analysis was the employee. Weighted results were not materially different from unweighted results so we report here the unweighted results for brevity.
Next, we performed multivariate analyses to measure the relationships between the health risks and each outcome, controlling for all 11 health risks, baseline demographics, type of health plan, and job type. Medical care costs and presenteeism were modeled using generalized linear models implemented with a gamma distribution and a log link function. In contrast to traditional ordinary least squares regression, a generalized linear model provides more robust coefficient estimates.16 Moreover, the log link function of the mean allows for the coefficients to be directly transformed back into the dollar scale and avoids the issue of potentially biased estimates that may result from using smearing estimation methods.17 Absenteeism was also modeled using a generalized linear model, but with a negative binomial distribution and log link function.
Finally, we modeled several risk reduction scenarios to estimate the hypothetical savings that might be achieved by reducing the prevalence of health risks among employees at the company.
Demographic characteristics are reported in Table 1 for employees eligible for the medical cost analysis and separately for the absenteeism and presenteeism analyses. Characteristics of each of the samples were very similar with 41% to 44% of employees between the ages of 45 and 54 years and most employees (64% to 70%) in nonunion plans. In addition, the majority of workers were salaried (84% to 86%), men (68% to 72%), and residing in the Western region (66% to 79%).
Table 2 presents the prevalence of each risk in the three samples used in this study: the absenteeism sample consisting of 158,541 health assessments, the medical cost sample consisting of 125,484 health assessments, and the presenteeism sample consisting of 38,298 health assessments. The most common risks were in the areas of poor nutrition (83.9% to 84.6%), safety (68.2% to 71.0%), weight (67.5% to 70.3%), and emotional health (53.6% to 60.1%). Risk information was missing on some risks. Cholesterol and triglycerides risk had the highest amount of missing information; 20.3% to 23.8% of health assessments were missing cholesterol risk, and 28.0% to 31.4% of employees were missing triglycerides risk information. Some risks (emotional health, exercise, safety, nutrition) had no or negligible amounts of missing data.
Table 2 also shows the medical care costs, absenteeism, and presenteeism days for employees who were at risk before controlling for other risks and demographics. For example, employees at risk for weight had $992 higher medical care expenditures, 0.4 additional days of absenteeism, and 0.7 additional days for presenteeism, compared with employees not at risk.
Coefficients and standard errors from multivariate analyses of the association between individual health risks and costs are shown in Table 3. Many of the predictors were statistically significant. Increased age was associated with higher medical care costs but with lower absenteeism and presenteeism. Female gender was associated with higher medical care costs, and increased absenteeism and presenteeism. Employees in union-negotiated medical plans tended to have lower medical care costs, but higher absenteeism and presenteeism. Hourly workers were associated with higher medical care costs and presenteeism, but lower absenteeism.
TABLE 3-a. Coefficie...Image Tools
Table 4 uses the model coefficients in Table 3 along with study data to present the regression adjusted costs or days associated with high and lower risk, and the impact of each risk. For instance, adjusted costs for employees at high risk for blood pressure were $4754 annually per at-risk employee, compared with $3677 for employees at lower risk, controlling for confounders. This difference, $1077, was determined to be the impact of blood pressure risk on costs, and it was statistically significant at 95% confidence levels. Thus, the risk for high blood pressure was associated with 29.3% higher medical care costs ($1077/$3677 × 100).
Similar results are shown for absenteeism and presenteeism. Blood pressure risk was associated with 2.28 absent days, compared with 2.10 days for lower risk. Thus, blood pressure risk was associated with 0.18 higher absent days. Valued at $320/day, blood pressure risk cost $57 extra dollars annually per employee, controlling for confounders. This was also statistically significant at 95% confidence levels.
High risks for blood pressure ($1077), blood glucose ($2310), and exercise ($848) were associated with the highest medical care costs (all statistically significant). A few risks (cholesterol, nutrition, safety, and alcohol) were associated with lower medical care costs. Cholesterol, nutrition, and safety were statistically significant. For absenteeism, with the exception of cholesterol, safety, and alcohol, the remaining health risks were associated with higher absenteeism costs. Being at risk for blood glucose ($146), weight ($113), emotional health ($176), and exercise ($131) was associated with the highest absenteeism costs (all statistically significant). Tobacco risk was associated with $143 higher absenteeism costs, but the result was not statistically significant. One risk (cholesterol) was associated with lower absenteeism costs but this result was also not statistically significant. All risks, with the exception of cholesterol risk, were significantly associated with higher presenteeism costs, with emotional health risk being associated with the highest presenteeism costs of $1056.
Results in the above-mentioned tables showed the costs associated with having each risk, for employees at risk. Table 5 illustrates the potential cost savings from five different risk reduction scenarios. The first column shows the risk prevalence at baseline, which is reproduced from Table 4. Each of the other columns shows risk prevalence for various risk reduction scenarios. Results in the bottom panel of Table 5 show the adjusted annual per capita costs at baseline and for each scenario. The difference between baseline and scenario costs represents the potential per capita savings for that scenario. Baseline adjusted costs were $3312 for medical care, $656 for absenteeism, and $949 for presenteeism. For all cost outcomes, the largest savings were obtained in scenario 5, which reduced the prevalence of emotional health, nutrition, weight, exercise, and tobacco risk, by 2%. Medical care cost savings for scenario 5 were $22.10 per capita or $1,710,482 for all 77,410 people in the study. Presenteeism savings for scenario 5 were $37.04 per capita or $2,867,296 for the entire sample. Absenteeism savings for scenario 5 were $10.69 per capita or $827,414 for the entire sample.
A large US employer was interested in understanding its employees' health risk profile and the relationship between that risk profile and key financial outcomes related to health care costs and worker productivity. By linking the health risk profile of each employee to relevant financial metrics, the employer was able to determine the relative economic and productivity cost burdens placed on the company by workers with health risks, and consider targeted health improvement interventions aimed at reducing the costliest risks.
Our results are consistent with other studies that have found depression or poor emotional health to be a significant predictor of medical costs. For example, the Health Enhancement Research Organization study found depression and stress to have the largest impact on medical care costs.3 A study by Goetzel et al6 of Novartis employees found that high laboratory biometric values and poor emotional health predicted high medical care costs. That study, which also employed the Mayo Clinic health assessment, produced results similar to those of this study since both analyses found that poor emotional health and biometric risk factors, such as high blood pressure high blood glucose, and obesity, were significant predictors of higher costs.
In contrast to the studies using the Health Enhancement Research Organization data, which found weight and blood glucose to be stronger predictors of medical costs compared to high blood pressure and high cholesterol, this study found overweight/obesity, high blood pressure, high blood glucose, high triglycerides, and inadequate physical activity to be strong predictors of high medical costs. These biometric health risks, including overweight/obesity, high blood pressure, and high blood glucose, were also significantly related to higher medical costs in the study of Novartis employees.6
Our findings regarding the impact of health risks on health-related productivity costs are also consistent with previous work. Risk factors related to emotional health, such as depression and stress, have been shown to predict absenteeism and presenteeism. In this study, poor emotional health was a significant predictor of both absenteeism and presenteeism. In addition, similar to the results from the study of health risks of Novartis employees, we found that biometric risks such as high blood glucose and excess weight were significant predictors of higher absenteeism costs.
For this employer, our analysis projects large medical care cost savings from reductions in health risks related to high blood glucose, high blood pressure, and low physical activity whereas the strongest potential for presenteeism savings lies in reducing emotional health risks. Potential absenteeism savings, however, are relatively small compared with medical care and presenteeism savings.
Stepping back and examining the areas where cost savings are most likely from risk reduction, it appears that targeting high blood glucose, high blood pressure, poor emotional health, and inadequate exercise would be most beneficial for this employer. Although the sample used for each outcome was different, the characteristics of each sample were very similar, which suggests that results from one sample (eg, sample with medical care data) may be generalized to the other samples (eg, sample with presenteeism data).
There are several limitations to this study worth noting. First, although we analyzed a large and geographically diverse sample, the results may not be generalizable to companies with employee characteristics dissimilar to this employee population.
Second, the question regarding absenteeism is self-reported on the health assessment and thus may be subject to bias. Although many studies have used self-reported absenteeism measures, and comparisons between employer administrative sickness/absenteeism data to self-report have found high levels of agreement,18,19 there are other studies that have shown significant underestimates of hours and days missed from work and overestimates of hours worked, compared with employer payroll records.20 The latter findings would likely result in underestimates of cost. Short et al21 found that while self-reported health care utilization and absenteeism can be relied upon as proxies for financial outcome measures when the recall period is within one month, the ability to extrapolate results from one month to a year, to infer annual medical and absenteeism losses, is inexact and subject to recall bias.
Third, because we did not have access to blood pressure measurements and laboratory test values for blood glucose and cholesterol, self-reports were used as evidence of risk. The reliance upon self-report for these conditions may have resulted in misclassification of employees, since some employees who reported having the condition may be managing the illness through medication. Thus, our estimates of increased costs associated with these conditions may be understated, since the estimates include employees whose conditions may be under control.
Fourth, since we performed a pooled cross-sectional analysis, we did not account for the change in employee risk profiles and corresponding costs over time, nor did we determine how changes in risk influence changes in costs. This makes it difficult to infer a causal relationship between risks and costs.
Fifth, several risk factors had missing data, especially biometric risks. To control for that limitation, we included indicators in the regression models to flag risk factor predictors that were missing data. However, missing data for biometric risk factors have been noted to be a problem by other studies of this nature. We were unable to determine whether these missing data would skew our results in a negative or a positive direction.
The Health Enhancement Research Organization study has served as the foundation for exploring the relationships between health risks and costs. However, subsequent studies have shown that these relationships may vary depending upon the populations studied and the instruments used to assess health risks. Definitions of health risks vary by health assessment instrument and that can lead to significant variations in estimates of health risks as well as variations in the associations between risks and costs. Therefore, it is important to replicate initial results with diverse populations using different survey tools. This study provides one such replication.
A strength of this analysis was that the employer studied contributed a large employee population to enable a distinction among risk factors and obtain increased precision in our study estimates. Furthermore, this study uses the validated WLQ to assess health-related work performance, in addition to medical care and absenteeism costs.
Our analysis of data from a large US employer illustrates how common, modifiable health risks impact not only direct medical costs, but also costs associated with health-related productivity here defined as workers' absenteeism and presenteeism. This analysis also highlights potential cost saving opportunities available to employers who implement evidence-based and effective health promotion and risk reduction programs. Employers who put in place targeted and comprehensive health promotion programs that reduce health risks among workers may see savings across multiple cost categories.
The authors thank Diane Everett and Kristin Wood at Mayo Foundation for their help in the preparation and review of the manuscript. We also thank Dr Phil Hagen for his senior leadership on the development of Mayo Clinic's Health Assessment. The health assessment serves as a key component of this study.
Mayo Foundation provided funding for the preparation of this article. The opinions expressed in this article are the authors' and do not necessarily represent the opinions of Thomson Reuters, Emory University, or Mayo Foundation.
APPENDIX Definition of Health Risks
The Mayo Clinic Health Assessment was the instrument used to define high risk in this study. Definitions of high risk follow.
* Alcohol risk is defined as imbibing more than fourteen drinks or seven alcoholic drinks per week for men and women, respectively, a positive CAGE score (The CAGE score is based on answers to four questions: C—Have you ever felt you should cut down on your drinking? A—Have people annoyed you by criticizing your drinking? G—Have you ever felt bad or guilty about your drinking? E—Eye opener: Have you ever had a drink first thing in the morning to steady your nerves or to get rid of a hangover?), or both; or if pregnant, any alcohol consumption.
* Emotional health risk is determined when individuals respond that they have moderate to very high stress levels, are sad, or both, or depressed so that these moods interfere with their job or personal life.
* Nutrition risk is present when there is medium to very high daily dietary fat intake, consumption of fewer than five daily servings of fruits and vegetables or both.
* Physical inactivity (lack of exercise) risk occurs with less than 30 minutes of moderate activity 5 or more days a week, or less than 60 minutes of vigorous activity weekly.
* Safety risk is defined as failure to meet basic safety recommendations including seat belt use, driving within the posted speed limit, helmet use when riding bicycles or motorcycles, driving or riding in a motor vehicle when the driver has been drinking, lack of sunscreen use, or not having a fire extinguisher or smoke alarm at home.
* Blood pressure risk is present with diagnosed hypertension or reported blood pressure values greater than or equal to 120/80 mm Hg.
* Blood glucose risk is present with diagnosed diabetes, fasting blood glucose values greater than or equal to 100 mg/dL, or nonfasting blood glucose values greater than or equal to 140 mg/dL.
* Cholesterol risk occurs with Low-density lipoprotein cholesterol values greater than 100 mg/dL or low-density lipoprotein cholesterol values greater than 130 mg/dL in the presence or absence of comorbidities, respectively, or if low-density lipoprotein value and comorbidity status is unknown, with total cholesterol values greater than or equal to 200 mg/dL.
* Triglyceride risk is defined as fasting triglyceride values greater than 150 mg/dL.
* Weight risk is present when body mass index is less than 18.5 kg/m2 or greater than or equal to 25 kg/m2.
* Tobacco risk occurs with use of any tobacco product including cigarettes, pipes, cigars, snuff, or chew.
1. Anderson D, Whitmer R, Goetzel R, Ozminkowski R, Wasserman J, Serxner S. The relationship between modifiable health risks and group-level health care expenditures. Am J Health Promot. 2000;15:45–52.
2. Burton W, Chen C, Conti D, Schultz A, Edington D. Measuring the relationship between employees' health risk factors and corporate pharmaceutical expenditures. J Occup Environ Med. 2003;45:793–802.
3. Goetzel R, Anderson D, Whitmer R, Ozminkowski R, Dunn R, Wasserman J. The relationship between modifiable health risks and health care expenditures: an analysis of the multi-employer HERO health risk and cost database. J Occup Environ Med. 1998;40:843–854.
4. Yen L, Schultz A, Schnueringer E, Edington D. Financial costs due to excess health risks among active employees of a utility company. J Occup Environ Med. 2006;48:896–905.
5. Yen L, Edington D, Witting P. Associations between health risk appraisal scores and employee medical claims costs in a manufacturing company. Am J Health Promot. 1991;6:46–54.
6. Goetzel R, Carls G, Wang S, et al. The relationship between modifiable health risk factors and medical expenditures, absenteeism, short-term disability, and presenteeism among employees at Novartis. J Occup Environ Med. 2009;51:487–499.
7. Burton W, Chen C, Conti D, Schultz A, Pransky G, Edington D. The association of health risks with on-the-job productivity. J Occup Environ Med. 2005;47:769–777.
8. Boles M, Pelletier B, Lynch W. The relationship between health risks and work productivity. J Occup Environ Med. 2004;46:737–745.
9. Ozminkowski R, Goetzel R, Chang S, Long S. The application of two health and productivity instruments at a large employer. J Occup Environ Med. 2004;46:635–648.
10. Tsai S, Wendt J, Ahmed F, Donnelly R, Strawmyer T. Illness absence patterns among employees in a petrochemical facility: impact of selected health risk factors. J Occup Environ Med. 2005;47:838–846.
11. Chapman LS. Meta-evaluation of worksite health promotion economic return studies: 2005 update. Am J Health Promot. 2005;19:1–11.
12. Goetzel RZ, Juday TR, Ozminkowski RJ. What's the ROI? A Systematic Review of Return on Investment (ROI) Studies of Corporate Health and Productivity Management Initiatives. AWHP's Worksite Health Summer; 1999.
13. Ozminkowski RJ, Dunn RL, Goetzel RZ, Cantor RI, Murnane J, Harrison M. A return on investment evaluation of the Citibank, N.A., health management program. Am J Health Promot. 1999;14:31–43.
14. Hagen P, Bond A, Rehman H, Molella R, Murad M. Have you had a tetanus booster in the last 10 years? Sensitivity and specificity of the question. Patient Educ Counsel. 2008;70:403–406.
15. Lim L, Williams D, Hagen P. Validation of a five-point self-rated stress score. Am J Health Promot. 2005;19:438–441.
16. Manning W, Mullahy J. Estimating log models: to transform or not to transform? J Health Econ. 2001;20:461–494.
17. Manning W. The logged dependent variable, heteroscedasticity, and the retransformation problem. J Health Econ. 1998;17:283–295.
18. Rees D, Cooper C. Research note: reliability of self-report sickness absence data in the health service. Health Serv Manag Res. 1993;6:140–141.
19. Revicki D, Irwin D, Reblando J, Simon G. The accuracy of self-reported disability days. Med Care. 1994;32:401–404.
20. Kessler R, Barber C, Beck A, et al. The world health organization health and work performance questionnaire (HPQ). J Occup Environ Med. 2003;45:156–174.
21. Short M, Goetzel R, Pei X, et al. How accurate are self-reports? Analysis of self-reported health care utilization and absence when compared with administrative data. J Occup Environ Med. 2009;51:786–796.
This article has been cited 1 time(s).
Health AffairsWellness Incentives In The Workplace: Cost Savings Through Cost Shifting To Unhealthy WorkersHealth Affairs
©2011The American College of Occupational and Environmental Medicine