Tsai, Shan P. PhD; Wendt, Judy K. MPH; Ahmed, Farah S. MPH; Donnelly, Robin P. MB, ChB; Strawmyer, Thomas R. MD
Illness absence is one of the most important measures of workforce productivity. In an increasingly globalized market environment in which American industries must maximize productivity to be competitive, the direct and indirect costs from absenteeism remain a significant concern to corporate managers. In response to these increasing costs, a variety of health promotion and disability management programs have been initiated during the past 20 years with the goal of reducing health care costs and employee absenteeism and, ultimately, improving workforce productivity.1–10 Although several studies have confirmed the positive impact these programs can have on productivity, most are based on a cross-sectional study design and/or survey data. Very few studies have used a prospective study design and objective (ie, not self-reported) health and absence data.
Illness absence in a working population is a complex phenomenon and can be influenced by many factors such as age, gender, education, personal health risk factors, and work-related factors. Numerous studies have examined the relationship between these risk factors and employee absenteeism.1 Most studies have shown an increased body mass index (BMI) plays a significant role in employee absenteeism.11–13 However, one study did not find a relationship between BMI and employee absenteeism or presenteeism,14 and another study failed to find an association among male employees.1 Results have been similarly conflicting when examining cholesterol levels. One study demonstrated employees with cholesterol levels greater than 220 mg/dL spent 11% more time absent from work7 whereas another study found no association between increased cholesterol and worker productivity.12 Studies evaluating hypertension and absenteeism reported no difference in rates between employees with normal blood pressure readings and hypertensive employees.7,12,15 A few studies have also explored the relationship between the number of health risk factors and absenteeism, and some have demonstrated an increase in absenteeism with an increase in the number of risk factors.7,12
In this study, we use prospectively collected employee physical examination and illness absence data to quantify the impact of selected health risk factors on employee illness absence. Data were extracted from the Shell Health Surveillance System (HSS), which includes both occupational and nonoccupational illness absence data and employee physical examination data.16 The purpose of this study was to examine illness absence patterns for employees at a Texas oil refinery and chemical facility and to assess the impact of selected factors on absence frequency and severity. In addition, the benefit (in terms of increased productivity) of smoking cessation, and the impact of pre-hypertension and impaired fasting glucose level on absenteeism will also be quantified.
Materials and Methods
The study population consisted of all company employees working at the Shell Deer Park Manufacturing Complex between January 1, 1994, and December 31, 2003. The study population was dynamic, with employees entering and exiting during the observation period. The 2550 employees were identified through the Shell HSS, the data system used in the Company’s ongoing monitoring of employee health.
Morbidity data were extracted from the HSS, which includes all illness absence days reported in company personnel and payroll systems and are, therefore, virtually complete. The present study included only absences lasting 6 days or more to reduce the inclusion of nonhealth-related absences, such as those for child or elder care. Previous analyses of this population have found that, although the majority of absence events last 5 days or less, 85% to 90% of workdays lost are as the result of absences of 6 or more days’ duration. A physician-identified cause of absence was available for approximately 80% of the morbidity reports. The causes of morbidity were coded according to the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD9-CM).17 Only the primary cause was used in the analyses. Pregnancy and childbirth related absences were excluded.
Biometric and risk factor data were derived from the HSS, which contains all employee preplacement and periodic physical examinations since January 1978. The most current examination data was used for each subject, including 82% done since 1994. The average number of examinations performed did not differ between employees who were absent and those who were never absent, 13.8 and 13.3, respectively. Smoking history was used to identify current smokers, nonsmokers, and ex-smokers, and for this third group, to quantify the number of years since quitting. Overweight was defined as a body mass index (BMI = weight [in kilograms]/height [in meters2]) between 25 and 29 kg/m2, and obesity was defined as a BMI of 30 kg/m2 or greater. Borderline high cholesterol was defined as between 200 and 239 mg/dL and high cholesterol as 240 mg/dL or greater. Elevated triglycerides were those 150 mg/dL or higher. Prehypertension was defined as a diastolic blood pressure reading between 80 and 89 mm Hg or systolic pressure reading between 120 and 139 mm Hg, and hypertension was defined as a diastolic reading greater than or equal to 90 mm Hg or systolic reading greater than or equal to 140 mm Hg. Elevated glucose was categorized into two groups: between 110 and 124 mg/dL (impaired fasting glucose) and 125 mg/dL or greater.
Person-years at risk were calculated for each employee beginning January 1, 1994, or the date of hire (whichever was later) and ending at the closing date of the study (December 31, 2003), the date of retirement, the date of death, or the date of termination/transfer (whichever was earlier). The morbidity frequency rate was calculated as the number of absence episodes divided by person-years at risk multiplied by 100. Similarly, the average duration of absence per employee each year was calculated as the number of total workdays lost divided by person-years at risk. Comparisons of frequency rate were conducted using the z-test to determine the statistical significance of the difference between two binomial proportions at the 95% significance level (P < 0.05).18 A Bonferroni correction was used to adjust for cumulative type I error that may result from multiple comparisons, since the likelihood of detecting a significant difference increases with an increasing number of comparisons.19 The workdays lost were grouped into five categories, and a χ2 statistic was used to compare the average duration of absence between groups. This method assumes that if the distribution of days lost were equivalent, then the corresponding means would be the same. The study also ranked the relative importance of the individual risk factors on workdays lost. In generating this rank, it was assumed that each risk factor was acting independently, allowing the calculation of the proportion of workdays lost as the result of each risk factor. All statistical analyses were done using SAS System Software PC Version 8.2.20
The average age at the end of the study was 46 years, and the average duration of employment at Deer Park was 20 years. A total of 4492 episodes of absence were reported during the 10-year study period, with 170,256 workdays lost because of illness absence events longer than 5 days. This rate was equivalent to the absence from work of about 4% of the average workforce (approximately 72 employees) each year. Of the 2550 employees included in the study, nearly 50% had no absences lasting 6 or more days during the 10-year study period, whereas approximately 26% had three or more absences (Table 1). These employees (n = 663) accounted for 76% of the total workdays lost.
Morbidity frequency rates increased with age, ranging from 6 absences per 100 person-years for those employees less than 30 years old, to nearly 44 absences per 100 person-years for those employees aged 60 and older. As a whole, the rate of absence due to illness was similar between male and female employees (24.8 vs. 24.5 per 100 person-years) (Table 2). The average duration of absence also increased with age, with male employees less than 30 years old absent an average of 2.5 days per person per year, and male employees aged 60 and older absent 20 days per person. Similar patterns were seen among female employees, although there was a wider range of absence duration between the youngest and oldest age groups (Table 2). Overall, average duration of absence was almost 50% longer for females than for males.
Morbidity frequency rates for male employees declined from 25.5 per 100 person-years during 1994 to 1998 to 23.9 per 100 employees during 1999 to 2003. The rates for these two time periods remained about the same, 24.4 and 24.6 per 100 person-years, respectively, among female employees. The average duration of absence increased slightly for male employees, from 8.5 days per person-year during 1994 to 1998 to 9.3 days per person-year during 1999 to 2003. Conversely, the duration decreased from 14.5 days to 11.6 days per person-year for female employees during the same periods.
Among the study population, 23% of males were current smokers, 30% were ex-smokers, and 47% were nonsmokers. Prevalence of ever-smoking and current smoking was lower among female employees. More than 85% of male employees and 57% of female employees were either overweight or obese and 61% and 46% of male and female employees, respectively, were prehypertensive. Approximately half of male employees and 41% of female employees had cholesterol greater than 200 mg/dL. Similarly, greater than 45% of male and 20% of female employees had elevated triglycerides. The prevalence of raised fasting glucose was relatively low, with 20% of male and 7% of female employees having glucose levels greater than 110 mg/dL (Table 3).
As illustrated in Figs. 1 and 2, employees with selected health risk factors had higher absence frequency rates than employees without such risk factors. Among male employees, nonsmokers and ex-smokers had similar illness absence rates (P = 0.05), whereas current smokers had an almost 50% increased rate compared with nonsmokers (33.0 vs. 22.4 per 100 employees; P < 0.05). Males in the highest glucose bracket also experienced a 50% increase in rate compared with those in the lowest glucose bracket (36.9 vs. 23.2 per 100 employees; P < 0.05). Among female employees, ex-smokers had an almost 2-fold increase (35.0 vs. 17.7 per 100 employees; P < 0.05) and current smokers a nearly 2.5-fold increase (42.4 vs. 17.7 per 100 employees; P < 0.05) in their illness absence rate compared with nonsmokers. Obese female employees had a 3.5-fold increase in their illness absence rate compared with women with a normal BMI (47.7 vs. 13.3 per 100 employees; P < 0.05). Female hypertensive employees had twice the illness absence rate as nonhypertensive employees (35.1 vs. 15.6 per 100 employees; P < 0.05), and female employees in the highest glucose bracket had a 2-fold increased rate compared with women in the lowest bracket (60.9 vs. 25.0 per 100 employees; P < 0.05).
The average number of days lost per employee by risk factor is shown for males and females in Figs. 3 and 4. Current smokers, both male and female, had almost twice as many days lost compared with their nonsmoker counterparts (men, 13.3 vs. 7.0 days, P < 0.05; women, 23.3 vs. 12.3 days, P < 0.05). Obese employees had significantly greater workdays lost than those with normal weight (men, 10.5 vs. 6.8 days, P < 0.05; women 21.8 vs. 7.7 days, P < 0.05). Similar patterns also were noted for cholesterol, triglycerides, hypertension, and glucose among male employees. Because of a smaller number of female study subjects, the patterns are relatively unstable.
A further examination of obesity and smoking in a subset of male employees showed that employees in the highest risk category of each factor, compared with employees in the lowest risk category, had an increased risk of some common diseases. For example, obesity was associated with a 5.4-fold increase in the rate of diseases of the endocrine system (0.59 vs. 0.11 per 100 employees), a 2.6-fold increase in heart disease (2.12 vs. 0.82 per 100 employees), a 2-fold increase in injuries (3.26 vs. 1.59 per 100 employees), a 1.8-fold increase in respiratory diseases (3.81 vs. 2.14 per 100 employees) and a 1.4-fold increase in musculoskeletal disorders (4.96 vs. 3.67 per 100 employees). Smoking was associated with a 2.6-fold increase in the rate of infectious diseases (0.78 vs. 0.30 per 100 employees), a 1.97-fold increase in heart disease (2.44 vs. 1.24 per 100 employees), a 1.45-fold increase in all cancers (0.55 vs. 0.38 per 100 employees), a 1.43-fold increase in musculoskeletal diseases (5.17 vs. 3.62 per 100 employees) and a 1.42-fold increase in respiratory diseases (4.03 vs. 2.83 per 100 employees).
The relative importance of each individual risk factor on workdays lost was evaluated by examining the proportion of absence days among employees in each risk factor category. This analysis assumed that each risk factor was acting independently. Among male employees, 90% of all the workdays lost as the result of BMI were in employees with BMI greater than 25 kg/m2. Prehypertension and hypertension accounted for 81% of all workdays lost among employees with blood pressure information and elevated cholesterol and triglycerides accounted for 54% and 55%, respectively, of workdays lost as the result of these risk factors. Among female employees, BMI greater than 25 kg/m2 accounted for 80% of all workdays lost as the result of BMI. Similarly, prehypertension and hypertension accounted for 71% of all workdays lost among employees with this information and 56% of workdays lost by employees with smoking information were lost by current smokers.
The absence frequency rate steadily inclined with an increase in the number of risk factors present from no risk factors (11.8 per 100 employees), to one (16.3 per 100 employees), two (23 per 100 employees), three (27.4 per 100 employees), and four or more risk factors (32.3 per 100 employees). The number of workdays lost also increased with the number of risk factors present, with the least number of workdays lost by employees with zero risk factors (4.1 day), followed by one (6.4 days), two (8.8 days), three (9.3 days), and four or more risk factors (12.6 days) (Fig. 5).
As illustrated in Fig. 6, the absence frequency rate was highest among current smokers (34.0 per 100 employees), gradually declining for ex-smokers the longer they had quit (1 to 9 years: 28.2, 10 to 19 years: 24.9, 20+ years: 23.5 per 100 employees), and lowest among nonsmokers (21.8 per 100 employees). Similarly, the average number of workdays lost because of smoking was the greatest among current smokers (14.3 days), followed by ex-smokers (1 to 9 years: 11.0 days, 10 to 19 years: 8.8 days, 20+ yrs: 7.9 days), and lowest among nonsmokers (7.6 days). A similar pattern emerged for a subset of only male employees, for whom the absence frequency rates declined for ex-smokers the longer they had quit. For example, the absence rate among those who had quit 10 to 19 years and 20+ years was 23.4 and 22.8 per 100 employees, respectively, as compared with 22.4 per 100 employees among nonsmokers. A similar pattern was also noted for the average number of days lost since quitting smoking.
The Shell Health Surveillance System has been an important tool in assessing the effectiveness of health programs, formulating preventive strategies and evaluating employee health.21 To maintain and update this system requires substantial corporate commitment to pursue necessary data collection and perform analyses. The strength of this study is in the use of physical examination data to identify employee health risks and payroll records to quantify absence days. These objective and reliable measures minimize the potential for reporting and recall bias that may be introduced when using self-reported absenteeism and health risk data. In addition, the study includes only absences greater than five workdays; therefore absences unrelated to employee health, such as childcare and elder care related absences are unlikely to be included.
The relationship between years since an employee quit smoking and absenteeism suggests that smoking-cessation programs could have an immediate effect as well as increasing long-term improvement on productivity. Compared with smokers, lower absenteeism was observed 1 to 9 years after employees quit smoking. In addition, absence rates and duration among employees who quit more than 20 years approximate those of nonsmokers. If this result is generalizable, a further question to be addressed in future studies is to refine the years since quitting at which maximum benefit is observed, that is, how many years after quitting do ex-smokers become similar to nonsmokers in their absenteeism patterns? This type of information would be useful in the assessment of cost-benefit analysis of corporate smoking cessation programs.
Although the impact on absenteeism from known health risk factors such as smoking, obesity, high cholesterol, and hypertension have been documented,1–2,4,7,13 the potential impact of prehypertension on workers’ health and absenteeism has rarely been reported. In this study, we found higher absence rates and longer absence duration among male employees with prehypertension compared with those with normal blood pressure. Results are not consistent among women, probably because of small numbers. Future corporate health promotion programs should also address this prehypertensive group to minimize their long-term economic costs.
Abnormal glucose traditionally has been defined as a fasting blood glucose level greater than 125 mg/dL. Given the numerous adverse health outcomes of elevated blood glucose, we also examined a subgroup of employees who had impaired fasting glucose (110 to 124 mg/dL). Approximately 10% of males and 4% of females fell into this category. The absence rate for employees with impaired fasting glucose was higher than employees with a fasting glucose level lower than 110 mg/dL. A similar observation was noted for duration of absence for men but was not consistent among women. These findings suggest that early treatment for those with impaired fasting glucose would have a positive impact on employee absenteeism. It is noteworthy that the most current guidelines suggest that impaired fasting glucose be defined as 100 mg/dL and greater, lower than the prior standard.22 We have conducted an additional analysis to compare employees with fasting glucose levels between 100 and 109 mg/dL to those employees with levels lower than 100 mg/dL. There was no difference in duration of absence (7.9 vs. 7.9 days per employee) but the absence frequency rate was slightly higher (25.2 vs. 22.4 per 100 employees).
This study also shows that different health risks impact productivity differently. Compared with employees without the same risk factors, those factors that caused the greatest time lost were obesity, hypertension, and elevated triglycerides for men and obesity, hypertension and smoking for women. Although the average workdays lost were highest for those with elevated blood glucose (125+ mg/dL), because of its low frequency, the overall impact is not as great as obesity or hypertension.
The results of this study are consistent with the majority of the literature on health risk factors and employee absenteeism. However, a few studies have not detected a relationship between BMI and absenteeism or presenteeism. In one of these studies, the study compared employees with BMI greater than 24.9 kg/m2 to employees with BMI less than 18.5 kg/m2,14 whereas the current study defined overweight as a BMI between 25 and 29 kg/m2 and obesity as a BMI greater than or equal to 30 kg/m2. Perhaps the reason the Bole et al’s study14 was unable to detect a relationship is because their increased BMI category in fact combined overweight and obese employees and, if the effect is only detectable among obese employees, then combining the categories dilutes the effects. In a review of modifiable health risks and employee absenteeism, the studies reviewed consistently found a relationship between obesity in female employees and absenteeism. This correlation was also seen among male subjects in all but one study. However, the one study that did not find a correlation may have been compromised by the small number of male employees that were evaluated.1
The results of studies evaluating increased cholesterol and blood pressure levels and smoking and absenteeism also have been conflicting. One study did not find an association between increased cholesterol and blood pressure levels or smoking status and worker productivity.12 However, the sample used in this study was relatively young (average age, 33.2 years) and the job, customer service, required little in the way of physical demands. Meaning, young workers are unlikely to have decreased performance in handling customer service inquiries because of increased cholesterol or blood pressure.12 Furthermore, other studies that have been unable to detect a relationship may be explained by the fact that employees with illnesses resulting from these diseases would likely require invasive procedures and lengthy hospital stays. When this happens, they are given disability status and disability is not a reported measure of absenteeism in company records.1 The lack of a relationship between smoking and productivity was a result of the fact that the corporation participating in the study had a no smoking ban in the workplace so employees did not take extra time off to smoke.12
A number of limitations should be noted. Results for female employees are not stable because of the relatively small sample size. This is particularly true for the average number of days lost because there were four female employees who were at risk and had more than 1000 workdays lost. Also, this study uses workdays lost as the measure of productivity, which may have accounted for only one aspect of direct costs. The potential lost productivity costs from health-risk related absence are likely much higher.
As is the case in most studies of working populations, it could be argued that the “healthy worker” population somehow differs from employees who leave work. Thus, it is useful to compare the average number of absences between the group of workers employed during the study period compared with employees who exited the study period. In this study, 72% of the workers were employed during the whole study period. Among this group, the absence prevalence was 22.6 per 100 employees. Among the 28% of the employees who left the study period because of retirement, death or termination, the absence prevalence was 37.7 per 100 employees.
A similar question arising from this issue is whether absences are related to a shorter duration of employment. We have examined a subgroup of males ages 40 to 49 to assess the impact of duration of employment on illness absence. This middle age group was chosen because, in the younger age group, there are very few employees employed for a long duration and, in the oldest age group, there are very few employees employed for a short duration. It is interesting to note that the absence frequency rate was 19.4 per 100 employees for those who worked less than 10 years, compared with 23.4 and 23.6 for employment between 10 and 19 years and over 20 years, respectively.
In summary, a number of employee health risk factors have been shown in this study to be associated with increased absence frequency and duration of absence. Higher levels of risk factors are clearly associated with greater loss of productivity. Furthermore, a greater number of health risk factors correspond to a higher rate of absence and longer duration of absence. Reductions of employee health risk factors can be an effective means of improving employees’ health and increasing a company’s productivity. Results of this study could have significant implications for assessing the economic ramifications of an unhealthy workforce.
1. Aldana SG, Pronk NP. Health promotion programs, modifiable health risks, and employee absenteeism. J Occup Environ Med. 2001;43:36–46.
2. Serxner SA, Gold DB, Bultman KK. The impact of behavioral health risks on worker absenteeism. J Occup Environ Med. 2001;43:347–354.
3. Bertera RL. Behavioral risk factor and illness day changes with workplace health promotion: Two-year results. Am J Health Promot. 1993;7:365–373.
4. Jeffery RW, Forster JL, Dunn BV, French SA, McGovern PG, Lando HA. Effects of work-site health promotion on illness-related absenteeism. J Occup Med. 1993;35:1142–1146.
5. Jones RC, Bly JL, Richardson JE. A study of work site health promotion program and absenteeism. J Occup Med. 1990;32:95–99.
6. Bertera RL. The effects of workplace health promotion on absenteeism and employment costs in a large industrial population. Am J Public Health. 1990;80:1101–1105.
7. Bertera RL. The effects of behavioral risks on absenteeism and health care costs in the workplace. J Occup Med. 1991;33:1119–1124.
8. Walsh DC, Egdahl RH. Corporate perspectives on work site wellness programs: a report on the seventh Pew Fellows conference. J Occup Med. 1989;31:551–556.
9. Baun WB, Bernacki EJ, Tsai SP. A preliminary investigation: effect of a corporate fitness program on absenteeism and health care cost. J Occup Med. 1986;28:18–22.
10. Blair SN, Smith M, Collingwood TR, Reynolds R, Prentice MC, Sterling CL. Health promotion for educators: Impact on absenteeism. J Prev Med. 1986;15:166–175.
11. Riedel JE, Lynch W, Baase C, Hymel P, Peterson KW. The effect of disease prevention and health promotion on workplace productivity: a literature review. Am J Health Promot 2001;15:iii–v.
12. Burton WN, Conti DJ, Chen CY, Schultz AB, Edington DW. The role of health related risk factors and disease on worker productivity. J Occup Environ Med. 1999;41:863–877.
13. Tsai SP, Gilstrap EL, Colangelo TA, Menard AK, Ross CE. Illness absence at an oil refinery and petrochemical plant. J Occup Environ Med. 1997;39:455–462.
14. Boles M, Pelletier B, Lynch W. The relationship between health risks and work productivity. J Occup Environ Med. 2004;46:743–745.
15. Yen LT, Edington DW, Witting P. Prediction of prospective medical claims and absenteeism costs for 1284 hourly workers from a manufacturing company. J Occup Med. 1992;34:428–435.
16. Joyner RE, Pack PH. The Shell Oil Company’s computerized health surveillance system. J Occup Med. 1982;24:812–814.
17. International Classification of Disease, 9th Revision, Clinical Modification. Ann Arbor, Michigan: Commission of Professional Hospital Activities; 1978.
18. Remington RD, Schork MA. Statistics With Applications to the Biological and Health Sciences. 2nd ed. Englewood Cliffs, NJ: Prentice-Hall, Inc.; 1985.
19. Miller RA. Simultaneous Statistical Influence. 2nd ed. New York: Springer Verlag; 1981.
20. SAS Institute, Inc. SAS Procedures Guide Release 8.12. Gary, NC: SAS Institute, Inc.; 1999.
21. Tsai SP, Wendt JK, Cardarelli KM, Fraser AE. A mortality and morbidity study of refinery and petrochemical employees in Louisiana. Occup Environ Med. 2003;60:627–633.
22. American Diabetes Association. Diagnosis and Classification of Diabetes Mellitus. Diabetes Care. 2004;27(S1):S5–S10.