Mental health, absenteeism and earnings at a large manufacturing worksite

BACKGROUND: A few recent studies have examined the relationship between mental illness and labor market variables. The findings are inconsistent, however, and leave unanswered many questions concerning both the nature and magnitude of the relationship. AIMS OF THE STUDY: A recently available worksite-based data set is analyzed to explore the relationship between symptoms of emotional and psychological problems and employee absenteeism and earnings among employees at a large US worksite. METHODS: The analysis was based on data collected through a random and anonymous survey of workers at a large US manufacturing worksite. Two measures of absenteeism are combined - days absent during the past 30 days due to sickness or injury and days absent during the past 30 days because the employee did not want to be at work - to create both a dichotomous (i.e., ever absent) and a continuous (i.e., number of days absent) absenteeism variable. Annual earnings were measured as personal earnings from the primary job. Various statistical models were tested to determine the independent and joint (with alcohol and illicit drug use) relationship between symptoms of emotional problems and labor market variables. RESULTS: The analysis consistently finds that workers who report symptoms of emotional/psychological problems have higher absenteeism and lower earnings than otherwise similar coworkers. This finding is robust to model specification and to the inclusion of comorbid conditions such as alcohol and illicit drug use. DISCUSSION: This study contributes new information to the literature in this area by estimating the effects of emotional/psychological symptoms on two important labor market variables: absenteeism and earnings. Several specifications of the absenteeism and earnings equations were estimated to test the independent effect of emotional symptoms and the joint effects of emotional symptoms and other comorbid conditions. The results suggest that employers should consider the productivity losses associated with workers' mental health when designing worksite-based programs such as employee assistance programs (EAPs). LIMITATIONS: Unlike national surveys of households or individuals, the sample does not include unemployed individuals or those outside the labor force. Therefore, the decision to participate in the labor market can not be modeled. In addition, the study relies on voluntary self-reported survey data that may suffer from underreporting of substance use and emotional symptoms. Although respondents were repeatedly assured about confidentiality, if underreporting does exist, it may be more acute than in household surveys because respondents may be more worried about job loss if they self-report drug or alcohol use at the worksite. CONCLUSIONS: All four measures of emotional symptoms had a positive and statistically significant relationship with absenteeism and a negative and statistically significant relationship with personal earnings. These findings were robust across all specifications, even when the effects of other potentially confounding factors (i.e., alcohol and drug use variables) are included. In addition, the number of days intoxicated and cigarette use in the past year appear to be significantly related to earnings even after controlling for emotional symptoms. Finally, the explanatory power of the models is relatively high for cross-sectional data, especially for the earnings regressions. IMPLICATIONS FOR HEALTH CARE PROVISION AND USE: The findings from this worksite suggest that employers might do well to reassess the priorities of their EAPs and consider directing more of their resources to diagnosing and assisting employees with emotional and psychological distress. IMPLICATIONS FOR HEALTH POLICY FORMULATION: It is strongly suggestive that mental health status is related to absenteeism and earnings for employees at this worksite. However, most employer-based programs and policies are designed to dissuade the use of alcohol and illicit drugs by workers (e.g., employee drug and alcohol testing) rather than addressing other employee behaviors and problems. IMPLICATIONS FOR FURTHER RESEARCH: Numerous opportunities are present to collect similar data from other worksites and settings to determine whether these models and results are robust.