According to the 2016 National Survey on Drug Use and Health (NSDUH), 11.5 million Americans aged 12 and older misused prescription pain relievers during the prior year, including 2.1 million initiating misuse for the first time.1,2 Opioid use disorder (OUD) is a serious and costly public health concern affecting employers and the workplace. Decreased on-the-job productivity (presenteeism) and employee absence due to health issues (sick days, inpatient or outpatient medical care, disability) related to OUD may result in significant indirect costs to employers beyond medical costs. Recent cost estimates of lost workplace productivity associated with OUD are lacking, however, earlier work suggests that the burden is substantial.
Birnbaum et al3 estimated that in 2007, costs due to opioid-related productivity loss (medically-related absenteeism, disability, presenteeism, incarceration, and premature death) exceeded $25 billion (USD2009), accounting for more than 45% of the total societal burden captured in that study. A study by Florence et al4 estimated that non-fatal lost productivity (reduced productive time/increased disability, and production lost for incarcerated individuals) accounted for 26% ($20.4 billion), and lost productivity due to fatality 27.3% ($21.4 billion) of the total societal costs associated with OUD in 2013.
Using more recent data than currently available in the published literature, the objective of this study was to assess workplace productivity loss, specifically work days lost due to medical visits and disability, associated with prescription OUD from the perspective of a self-insured employer.
A retrospective employer claims database analysis was conducted using OptumHealth Care Solutions, Inc.5 covering the time period from January 1, 2012 through March 30, 2016. The database contains de-identified administrative claims data for medical diagnoses and services, and prescription drugs for approximately 18 million privately-insured beneficiaries and their dependents from over 80 US-based self-insured companies. The data also provide information on disability claims for primary beneficiaries (ie, employees), and employees account for approximately 30% of the overall number of insured lives.
Employees were included in the study if they were a primary beneficiary from a company that reported disability data, between 12 and 64 years of age, and had at least one medical claim during the index period (January 1, 2013 to September 30, 2015) to ensure that they were users of healthcare services during this time period (Fig. 1). Employees were also required to be continuously enrolled in a non-HMO health plan throughout a 6-month baseline and a 12-month follow-up period around a specified index date (see below). Employees aged 65 and older were excluded from the study since the data likely did not reflect the full set of claims capturing their entire medical history.
Eligible employees were categorized into two study cohorts based on the presence or absence of OUD. The OUD cohort included employees with at least one medical claim for OUD (excluding heroin), based on the International Classification of Disease, Ninth Revision (Appendix A, https://links.lww.com/JOM/A694), during the index period. The date of first prescription OUD diagnosis was defined as the employee-specific index date.
The no OUD cohort was a 10% random sample of employees who had no medical claims associated with prescription or illicit OUD at any point throughout their medical history (ie, all available medical claims in the database). This longer look back period for controls was chosen in order to eliminate any potential lingering effects (ie, treatment costs) of OUD that may have occurred outside of the study period such as an undiagnosed relapse of OUD. The index date for employees in the no OUD cohort was randomly assigned throughout the index period. The 10% random sample was selected to reduce the overall cohort size and improve the interpretation of P-values when outcomes between the two cohorts were compared, since P-values often register significant differences as a result of large underlying sample sizes.
The mean number of work days (Monday through Friday) lost due to disability and medical visits and associated costs were compared between the OUD and control cohorts over a 12-month follow-up period centered around the index data (6-month pre-index and 12-month post-index). A follow-up period before and after index date was used because research by Kirson et al6 suggests that the burden of OUD often begins prior to a medical diagnosis with over 40% of the costs occurring in the 6 months before OUD diagnosis.
Work days lost due to disability was defined as the number of days for which an employee had a disability claim and was receiving benefits, plus the mandatory waiting (or elimination) period before disability benefits began. Most health plans require a waiting period during which the employee is absent from work before becoming eligible for disability benefits.
Absenteeism related to medical visits was captured by identifying work days during which an employee sought medical care. Specifically, medical claims for emergency department visits, inpatient stays, and substance use treatment facility stays were counted as a full day of absence, while claims for outpatient visits were counted as a half-day of absence from work. To avoid double-counting, disability days and absenteeism related to medical visits were classified to be mutually exclusive with disability days taking precedence should both events occur on the same day. Total work loss days included both categories of work absence.
Descriptive statistics (mean, standard deviation [SD], percent) were used to summarize and compare baseline characteristics of the OUD and no OUD cohorts. Wilcoxon rank-sum tests were used for comparing continuous variables and Chi-square tests or Fisher exact tests were used for comparing categorical variables.
To adjust for potential baseline differences between the two cohorts, matching was employed. A propensity score was generated for each employee using a multivariate logistic regression model to predict the probability of their being diagnosed with OUD taking into account sex, age, index year, census division, baseline healthcare utilization (emergency department and outpatient visits, days in substance use treatment centers and inpatient facilities, and total number of prescription drugs), baseline total costs (medical and pharmacy), and Charlson Comorbidity Index score (Appendix C, https://links.lww.com/JOM/A694).7 Employees in the two cohorts were then matched on a 1:1 basis using their propensity score (greedy8 matching within ±1/4 standard deviation) and the presence of baseline comorbid conditions for non-opioid substance use, psychiatric disorders, trauma, neuropathic pain, arthritis, fibromyalgia, back and neck pain, and general pain (Appendix B, https://links.lww.com/JOM/A694).
Differences in the total number of work loss days between matched cohorts of employees with and without OUD were calculated as the difference in the average number of work loss days per employee during the follow-up period. Daily costs for disability days were calculated based on the average daily disability benefits paid by employers, adjusted to 2018 USD based on the employment cost index.9 Daily costs for days of absence related to medical visits and days of absence waiting for disability benefits to begin were calculated based on average employee wage, adjusted to 2018 USD based on the medical care sub-index of the consumer price index.10P-values at the <0.05 level were considered statistically significant.
There were 8561 employees in the prescription OUD cohort and 93,191 in the no-OUD cohort that met study eligibility, resulting in 2311 matched pairs of employees with disability data (Fig. 2). The cohorts were well-matched on measured baseline characteristics (P > 0.05). Overall, the mean age in each cohort was approximately 37 years, slightly over half were men, and the average CCI (SD) was 0.3 (0.8) for employees in the OUD cohort and 0.2 (0.7) for employees in control cohort (Table 1). Non-opioid substance use occurred in 7.2% of employees, psychiatric disorders in 32.5%, and 18.6% had evidence of trauma. Only the mean (SD) number of outpatient visits during the baseline period varied statistically between cohorts with employees in the OUD cohort having fewer occurrences relative to the control cohort (6.7 [8.5] vs 8.2 [11.4], P < 0.001).
Work Loss Due to Disability
Approximately 8.8% of employees in the OUD cohort missed at least 1 day of work due to disability compared with 6.5% of employees in the control cohort (P = 0.003) (Table 2). Employees in the OUD cohort also had greater mean (SD) numbers of days missed while waiting for disability to begin (0.24 [1.4] vs 0.17 [1.0]; P = 0.035) and greater mean number of days missed while receiving disability benefits (9.5 [40.9] vs 5.6 [30.0]; P = 0.001) relative to controls (Table 2; Fig. 3). This difference resulted in higher mean (SD) annual work loss costs related to disability among employees with OUD ($2267 [$11,654] vs $1459 [$11,736]; P = 0.002) (Table 2; Fig. 4).
Work Loss Due to Medical Visits
The proportions of employees who had at least 1 day missed from work due to medical visits did not differ between groups, however, the mean (SD) number of days missed from work associated with seeking medical care was higher for employees in the OUD cohort compared with controls (17.8 [18.5] vs 10.0 [12.4]; P < 0.001) (Table 2; Fig. 3). This difference was primarily driven by the number of work days missed due to substance use treatment and outpatient visits. The corresponding mean (SD) annual costs due to absenteeism related to medical visits was higher for employees with OUD relative to controls ($5866 [$9925] vs $3939 [$7506], P < 0.001) (Table 2; Fig. 4).
Total Work Loss
Mean (SD) total work loss days (days of absence related to medical visits and disability/waiting for disability) were significantly higher for employees with OUD compared with controls (27.5 [31.1] vs 15.7 [20.0]; P < 0.001], translating into mean (SD) annual total work loss costs of $8193 ($14,694) and $5438 ($13,683) (P < 0.001), respectively (Table 2; Figs. 3 and 4). Absenteeism related to medical visits accounted for over 70% of the total work loss costs in both cohorts.
The results of this study suggest that prescription OUD imposes significant work loss costs on employers in the form of benefits paid to employees during work absences due to disability and costs arising from absence during the work week when employees are receiving medical care. We found that on average, an employee with an OUD had 11.8 excess work loss days annually, primarily driven by work missed due to substance use treatment and outpatient care. The corresponding annual excess costs were $2755, with absenteeism related to medical visits accounting for 70% and disability-related accounting for the remainder.
Our results are consistent with a 2003 to 2014 retrospective claims analysis reported by Johnston et al.11 The study compared work-loss, identified by worker's short-term disability claims, between those with and without an OUD diagnosis. Employees with OUD had a higher mean number of work-loss days compared with those without OUD (88.0 vs 60.2; P < 0.001) over a 12-month follow up period. The higher numerical values for mean days lost can be explained by differences in the analyses. Our calculations included employees with and without disability claims, whereas Johnston et al11 included only those with at least one disability claim.
Similar to our study, Rice et al12 reported excess annual work-loss costs for employees with OUD based on disability claims and medically-related absenteeism. The mean work-loss related costs for employees with OUD were $3773 (2012 USD) compared with $2528 for those without OUD (P < 0.001).
Recent legislative trends have focused on program development and treatment strategies aimed at curbing the opioid crisis. In 2016, the Comprehensive Addiction and Recovery Act was signed into law. The legislation addresses the full continuum of care from primary prevention of OUD to recovery support, including expanded access to addiction treatment services.13 Kuhn14 notes that employers can play a role in curbing the epidemic through awareness campaigns to include messaging around the implications of sharing medications, the recognition of OUD as a medical condition (to reduce stigma), and the availability of OUD treatment programs for employees. Such programs may improve employee health and have the added benefit of reducing employer costs.
The definition of work productivity loss examined in this study was limited to disability and medical visits data recorded in an administrative claims database. As such, the study was not able to assess economic impact of prescription OUD on work productivity loss related to unemployment, dropping out of the labor force, incarceration, presenteeism, and mortality. Therefore, the overall productivity loss associated with OUD estimated by this study is likely an underestimate. In addition, the generalizability of study results to employees covered by insurance arrangement other than commercial health plans may be limited.
This study was limited to beneficiaries who had at least one medical claim during the study period to ensure enrollment in the system. The impact of this restriction on study outcomes is unclear; costs may be either over- or underestimated depending on the health status of those who were excluded. For the purposes of matching beneficiaries with and without OUD, a period of 6-month pre-index was used to identify relevant baseline characteristics for both cohorts. However, for the non-OUD cohort only, a full look-back was performed (to the end of data availability) in order to exclude any individual with a prior OUD diagnosis. By doing so, our study was able to minimize any potential residual effects from OUD diagnosed outside of the study period, although some degree of selection bias may have occurred.
Finally, medical claims were utilized to identify prescription OUD in an employed population, however, not all employees with OUD will be diagnosed and have a corresponding record in the healthcare system. Employees with OUD who did not receive medical treatment for the condition were assigned to the OUD cohort. Underutilization of OUD diagnosis codes to avoid stigma associated with OUD also cannot be ruled out. A previous study estimated that the ratio of undiagnosed to diagnosed OUD may be as high as 3.3:1 suggesting that the costs of undiagnosed OUD could potentially exceed those of OUD for employers.12 Unfortunately, undiagnosed employees cannot be easily discerned from controls by administrative claims. The misclassification of employees with undiagnosed OUD may lead to an underestimation of the overall potential work loss and work loss costs associated with OUD. We expect, however, that any such exclusion would have only a moderate cost impact assuming those entering substance use treatment would have a corresponding diagnosis.
The opioid use disorder epidemic is an important topic from a societal perspective. Notwithstanding limitations, this analysis provides a framework for employers to assess the economic impact of prescription OUD in their organizations and may prompt employers to consider positive interventions such as rehabilitation assistance for those in need.
1. Center for Behavioral Health Statistics and Quality2016 National Survey on Drug Use and Health: Detailed Tables, Table 1.1A - Types of Illicit Drug Use in Lifetime, Past Year, and Past Month among Persons Aged 12 or Older. Rockville, MD: Substance Abuse and Mental Health Services Administration; 2017.
2. Center for Behavioral Health Statistics and Quality2016 National Survey on Drug Use and Health: Detailed Tables, Table 4.2B - Past Year Initiation of Pain Reliever Misuse among Persons Aged 12 or Older and Past Year Pain Reliever Misusers Aged 12 or Older, by Demographic Characteristics. Rockville, MD: Substance Abuse and Mental Health Services Administration; 2017.
3. Birnbaum HG, White AG, Schiller M, Cleveland JM, Roland CL. Societal costs of prescription opioid abuse, dependence, and misuse in the United States. Pain Med
4. Florence CS, Zhou C, Luo F, Xu L. The economic burden of prescription opioid overdose, abuse, and dependence in the United States, 2013. Med Care
5. OptumHealth Care Solutions Inc. Employer Claims Database, Q1-2013 - Q3-2015.
6. Kirson NY, Scarpati LM, Enloe CJ, Dincer AP, Birnbaum HG, Mayne TJ. The economic burden of opioid abuse: updated findings. J Manag Care Spec Pharm
7. Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care
8. Bergstralh EJ, Kosanke JL. Computerized matching of controls. Section of Biostatistics Technical Report 56. Minneapolis: Mayo Foundation; 1995.
9. U.S. Bureau of Labor Statistics, Employment Cost Index: Total compensation: All Civilian (ECIALLCIV), retrieved from FRED, Federal Reserve Bank of St. Louis. Available at: https://fred.stlouisfed.org/series/
(ECIALLCIV). Accessed April 30, 2019.
10. U.S. Bureau of Labor Statistics, Consumer Price Index for All Urban Consumers: Medical Care [CPIMEDNS], retrieved from FRED, Federal Reserve Bank of St. Louis. Available at: https://fred.stlouisfed.org/series/
. CPIMEDNS. Accessed April 30, 2019.
11. Johnston SS, Alexander AH, Masters ET, et al. Costs and work loss
burden of diagnosed opioid abuse among employees on workers compensation or short-term disability. J Occup Environ Med
12. Rice JB, Kirson NY, Shei A, et al. Estimating the costs of opioid abuse and dependence from an employer perspective: a retrospective analysis using administrative claims data. Appl Health Econ Health Policy
14. Kuhn S. Opioid addiction and implications for employers. Benefits Q