The impact of hepatitis C virus infection on work absence, productivity, and healthcare benefit costs

Authors


  • Potential conflict of interest: Drs. Su and Corey-Lisle own stock in Bristol-Myers-Squibb. Drs. Kleinman and Brook received grants from Bristol-Myers-Squibb.

Abstract

Chronic hepatitis C virus (HCV) infection is generally considered an asymptomatic disease. However, studies have shown that HCV has a substantial negative impact on patients' quality of life and functioning. This study was designed to compare absenteeism, productivity, and health cost between employees with and without HCV infection in the United States. Employee records from multiple large employers in the United States were obtained from the Human Capital Management Services Research Reference Database and were assessed for demographics, salary, healthcare use, work loss, and workers' compensation. HCV-infected subjects were identified by International Classification of Diseases 9th revision Clinical Modification codes. Controls were randomly selected from employees not diagnosed with HCV. T-tests and chi-square tests were used to determine if there were differences in demographic characteristics. Regression modeling compared days absent (among benefit-eligible employees) and productivity (among employees with data on task-oriented activities), while controlling for the impact of confounding factors. A total of 339,456 subjects were evaluated. Employees with HCV (n = 1664) had significantly more lost work days per employee than the control cohort (n = 337,792), including sick leave, short-term disability, and long-term disability. HCV-infected workers had 4.15 more days of absence per employee than the control cohort. Productivity was measured by units of work processed per hour; employees with HCV processed 7.5% fewer units per hour than employees without HCV (P > 0.05). All healthcare benefit costs among HCV employees were significantly higher than the same costs among employees without HCV. Overall, the total incremental difference was $8352 per year. Conclusion: This real world study provides evidence that there is a substantial indirect burden of illness and describes a relationship between HCV infection, productivity, increased absenteeism, and higher healthcare benefit costs. Hepatology 2010

Hepatitis C virus (HCV) infection is a blood-borne infection characterized by inflammation and destruction of liver cells. It is a major health problem worldwide, affecting 180 million people globally.1 No vaccine is available currently, and HCV infection is projected to become a substantial health and economic burden over the next 10-20 years.2

Chronic HCV infection is defined as ongoing liver cell injury with inflammation lasting longer than 6 months. HCV is generally considered an asymptomatic disease in its early stage and may present with nonspecific symptoms such as fatigue, flu-like symptoms, muscle pain, joint pain, abdominal pain, intermittent low-grade fevers, itching, sleep disturbances, appetite changes, and mood swings.3 Over a 20-year period, 20% of chronically infected patients will develop cirrhosis despite a lack of clinical signs or symptoms.4, 5 HCV-related cirrhosis leads to continued decline in liver function, and ≈6% of patients per year will develop hepatic decompensation, 4% per year will develop HCV, and 3%-4% per year can be expected to require liver transplantation or die.5

Despite a lack of clinical signs or symptoms in early chronic HCV infection, HCV-infected patients have been reported to experience lower health-related quality of life compared with the general population.6, 7 For patients treated with peginterferon alpha-2a monotherapy or a combination of interferon alfa-2b plus ribavirin, reduced health-related quality of life during treatment has been linked with increased loss of work productivity and activity impairment.8, 9 However, there is limited research examining the impact of HCV alone on patients' work loss and productivity.

This study was designed to compare absenteeism, productivity, and health benefit and comorbidity costs between employees with and without HCV infection. These factors characterize the economic and humanistic disease burden for this population.

Abbreviations:

HCMS RRDb, Human Capital Management Services Research Reference Database; HCV, hepatitis C virus; ICD-9, International Classification of Diseases 9th revision Clinical Modification.

Materials and Methods

Data Source and Cohort Selection.

This retrospective cohort study assessed administrative claims data from the Human Capital Management Services Research Reference Database (HCMS RRDb) from January 1, 2001, to March 31, 2007. The HCMS RRDb collects anonymous data from different United States employers (including manufacturing, insurance, retail, transportation, telecommunications, healthcare, and pharmaceuticals). The data are geographically dispersed throughout the United States and across a variety of health plans.

The HCMS RRDb includes demographic, salary, healthcare use, prescription drug coverage, work loss, and workers' compensation data for employees from multiple large employers in the United States. All personal information is in compliance with the Health Insurance Portability and Accountability Act. The database is consistent with the United States civilian labor force in terms of age and sex proportions.

HCV subjects were identified by International Classification of Diseases 9th revision Clinical Modification (ICD-9) codes (at least one of the following primary, secondary, or tertiary ICD-9 codes: 070.41, 070.44, 070.51, 070.54, or 070.7x). This cohort included HCV-diagnosed subjects with or without treatment with interferon, peginterferon, or ribavirin. Non-HCV controls were employees without HCV (those that did not have above-HCV ICD-9 codes). The index date used the “average” start treatment date from the subjects who had HCV treatment. This index date is also used both for HCV subjects without treatment and for non-HCV controls.

Throughout the study, subcohorts were used based on the availability of specific data elements. For example, the place of service components only included persons with place of service data. Similarly, absence benefits were also limited to persons enrolled in the specific benefit.

Outcome Measures.

The study outcome measures included health-related employee days absent; benefit healthcare and prescription drug costs for both the full sample and a subsample of employees with costs data that specified where the care was performed; prevalence of comorbidities; and at-work productivity.

Absence included lost time due to sick leave, short-term disability, long-term disability, and workers' compensation claims. Prescriptions filled at pharmacies were included in the prescription drug cost category. Infused and injected therapies were included in the healthcare costs category. Comorbid conditions were defined according to the U.S. Agency for Healthcare Research and Quality's 17 major diagnostic categories and specific categories (261 total).10 At-work productivity was measured in proprietary units of work output developed by the HCMS group and used in prior research.11-14

Statistical Analysis.

For demographic data in all cohorts, mean values and 95% confidence intervals were calculated. Demographic data between the two cohorts were compared using t tests for continuous variables and chi-square tests for discrete variables. Values were considered statistically significant at P ≤ 0.05. A two stage regression technique was used for all cost absence dependent variables.15-17 First, logistic regression was used to predict the likelihood of having any healthcare costs or absences during the study period. Second, generalized linear regressions of healthcare costs or absences were used to estimate average annual healthcare costs or absences for those employees with positive healthcare costs or absences. Next, those results were then combined to yield estimates of annual healthcare costs or absences for all employees in the population. The productivity models used only the second stage of the regression methodology, because the productivity analysis excludes employees with zero productivity during the time period. All regression models controlled for differences between cohorts in age, sex, marital status, race, exempt status, full-time/part-time status, salary, location, Charlson comorbidity index, and number of months of plan enrollment. The models only controlled for linear forms of the continuous variables (age, salary, Charlson comorbidity index, and number of months of plan enrollment). Outcome estimates were produced by evaluating the models at the study population means of the controlled variables.

For the comorbidity comparisons, 20 control employees were randomly matched on demographic characteristics to each employee with HCV. The prevalence of comorbidities was calculated as the percentage of persons within the cohort who had a claim with an ICD-9 code within the particular diagnostic category.

To compare the prevalence of comorbidities between cases and controls, 95% confidence intervals were calculated for the comorbidity odds ratio using the Woolf method.18 The prevalence of each comorbidity was considered to be different between cohorts (P < 0.05) if the confidence interval for the odds ratio did not include 1.0. The mental and drug-related specific categories reported were chosen based on guidance from the study team and reviewers.

Results

Descriptive statistics for employees in the overall HCV (n = 1664) and control (n = 337,792) cohorts are shown in Table 1. Descriptive statistics for the other samples used in this study are available in Supporting Tables 1–3. The results show that employees with HCV were significantly older (by more than 5 years) than employees in the control cohort (P < 0.0001), were more likely to be full-time, and were less likely to be black. Furthermore, HCV employees earned nearly $5,000 less in annual salary than the control employees. Other demographic variables were not significantly different (P > 0.05) between the HCV and control cohorts.

Table 1. Demographic and Baseline Characteristics
VariableEmployees With HCVEmployees Without HCVDifference in MeansP-Value for Difference2
NMeanStandard ErrorNMeanStandard Error
  • 1

    For employees with HCV treatment, the index date is the date of the first prescription for ribavirin, interferon or peginterferon in the study period. For employees with HCV and without treatment, the index date is the average index date (by company) of employees with HCV and treatment. For employees without HCV, the index date is the average index date based on the group of employees with HCV and treatment.

  • 2

    Differences are considered significant if P < 0.05 (based on t-tests for continuous variables and chi-square tests for binary variables and are highlighted in bold).

Age (at index date1)1,66445.730.18337,77440.520.025.20<0.0001
Annual salary1,660$49,689$633334,285$54,695$145−$5,006<0.0001
Female1,66437.4%1.2%337,79240.5%0.1%−3.1%0.0112
Married1,55452.2%1.3%311,85751.7%0.1%0.5%0.7098
White1,27956.0%1.4%260,48955.1%0.1%0.9%0.5190
Black1,27913.2%0.9%260,48915.5%0.1%−2.2%0.0267
Hispanic1,2799.3%0.8%260,4898.7%0.1%0.6%0.4576
Full time1,66493.3%0.6%337,79289.7%0.1%3.6%<0.0001
Charlson index1,6640.6750.036337,7920.2010.0010.474<0.0001

The majority (67.3%, n = 1120) of subjects had chronic HCV infection without mention of hepatic coma (ICD-9 070.54); 28.1% (n = 468) of the subjects had acute HCV without mention of hepatic coma (ICD-9 070.51), and 2.9% (n = 48) had unspecified HCV (ICD-9 070.7x). Only 1.7% (n = 28) had chronic HCV infection with hepatic coma. In the HCV population, 73.5% (n = 1223) were HCV-infected subjects without treatment, and 26.5% (n = 441) were HCV-infected subjects with antiviral treatment.

Table 2 provides a comparison of adjusted annual health-related work absence days for employees with and without HCV infection. Health-related absence days fell into four categories: sick leave, short-term disability, long-term disability, and workers' compensation with varying Ns by category. Using regression modeling to control for demographic differences between the cohorts, there were significantly (P < 0.05) higher annual sick leave (2.85 versus 2.21), short-term disability (4.17 versus 1.93), and long-term disability (1.36 versus 0.34) absence days for employees with HCV in comparison with employees without HCV.

For the total health benefit costs, the cost of illness for employees was determined using healthcare, prescription drug, sick leave, disability (short-term and long-term), and workers' compensation information. Table 2 compares annual health benefit costs per employee for the HCV cohort versus the non-HCV cohort. All healthcare, prescription drugs, sick leave, short-term disability, long-term disability, and workers' compensation costs among HCV employees were significantly higher than these costs among the control cohort. Overall, the total difference was $8352 per year, including $490 in indirect (absence) costs.

Table 2. Annual Adjusted Absence Days1,2 and Health Benefit Costs2,3 per Employee for Employees With HCV Versus Employees Without HCV (Control)
Annual Absence DaysEmployees With HCVEmployees Without HCVDifferenceP-Value4
NAdjusted Mean DaysNAdjusted Mean Days
  • 1

    Absence days were adjusted using regression modeling and controlling for age, gender, marital status, race, exempt status, full-time/part-time status, salary, location, the Charlson Comorbidity Index, and the number of months of eligibility. Only employees eligible for each specific benefit were included in the regression models for that benefit. Lost days include all days from claims begun at some point during the eligibility period following the index date.

  • 2

    For employees with HCV treatment, the index date is the date of the first prescription for ribavirin, interferon or peginterferon in the study period. For employees with HCV and without treatment, the index date is the average index date (by company) of employees with HCV and treatment. For employees without HCV, the index date is the average index date based on the group of employees with HCV and treatment.

  • 3

    Costs were adjusted using regression modeling and controlling for age, gender, marital status, race, exempt status, full-time/part-time status, salary, location, the Charlson Comorbidity Index, and the number of months of eligibility. Costs are also inflation adjusted to 2007 dollars. Only employees eligible for each specific benefit were included in the regression models for that benefit. Lost time costs include all costs from claims begun at some point during the eligibility period following the index date.

  • 4

    Differences are considered significant if P < 0.05 and are highlighted in bold.

Annual absence days      
Sick leave7112.85150,4252.210.64<0.0001
Short-term disability9404.17191,9071.932.24<0.0001
Long-term disability1,3631.36257,6100.341.020.0049
Worker compensation1,4940.63292,6310.380.250.0597
 Total absence days 9.02 4.874.15 
Annual Health Benefit CostsNAdjusted Mean CostNAdjusted Mean CostDifferenceP-Value4
Indirect costs      
 Sick leave711$445150,425$364$81<0.0001
 Short-term disability940$530191,907$256$273<0.0001
 Long-term disability1,363$90257,610$28$620.0118
 Workers, compensation1,494$359292,631$286$740.0490
 Sub-total indirect (absence) costs $1,424 $934$490 
Direct costs      
 Health care1,664$4,885337,792$2,281$2,604<0.0001
 Prescription drug1,664$5,801337,792$543$5,258<0.0001
 Sub-total direct costs $10,686 $2,824$7,862 
 Total Costs $12,111 $3,758$8,352 

Healthcare costs were further examined by point of service (Fig. 1). Among those with point of service information, HCV employees (n = 900) had higher medical costs than control employees (n = 152,011) in all six point-of-service categories (Doctor's Office, Inpatient Hospital, Outpatient Hospital or Clinic, Emergency Department, Laboratory, and Other). Among those with point of service information, the annual prescription drug cost for employees with HCV and the control cohort were $4932 and $600, respectively. All comparisons were significant (P < 0.0001).

Figure 1.

Adjusted direct annual cost of healthcare per employee by point of service (HCV versus control).

Table 3 compared hourly productivity (units of work processed per hour worked) for a subset of employees for whom productivity data were available. Employees in the control cohort with productivity data (n = 27,401) processed 7.5% more units per hour worked than the employees in HCV cohort with productivity data (n = 92). The difference in hourly productivity between cohorts, however, was not statistically significant.

A cohort of 26,580 controls with 1 year of health plan enrollment was propensity score matched 20:1 to 1329 subjects with HCV and 1 year of health plan enrollment for the comparisons of annual prevalence of comorbidities by major diagnostic category (Table 4). Apart from pregnancy childbirth, employees with HCV had higher prevalence of comorbidities than the control group, with infectious and parasitic disease, and digestive system having the top two differences.

Using the same comorbidity subset, comparisons of annual prevalence of comorbidities by selected specific categories are presented in Table 5. Almost all of these categories had significantly higher prevalences for the HCV cohort except “Other Psychoses” and ”Mental Behav/Observ/Screening,” both of which trended higher but were not significant.

Table 3. Real Productivity Comparison of Units Processed per Hour1,2 (HCV Versus Control)
 Employees with HCVEmployees Without HCV  
N = 92N = 27,401DifferenceP-Value4
  • 1

    Productivity output measurements come from real worker output data. These data provide the number of units processed (units of work performed) for each employee on a daily basis.

  • 2

    Productivity output measurements are taken during the eligibility period following the employee's index date. For employees with HCV treatment, the index date is the date of the first prescription for ribavirin, interferon or peginterferon in the study period. For employees with HCV and without treatment, the index date is the average index date (by company) of employees with HCV and treatment. For employees without HCV, the index date is the average index date based on the group of employees with HCV and treatment. Outliers (≤1st percentile and ≥ 99th percentile) were removed.

  • 3

    Productivtiy output measurements shown are adjusted by using regression modeling and by controlling for age, gender, marital status, race, exempt status, full-time/part-time status, salary, location, the Charlson Comorbidity Index, and the number of months of eligibility.

  • 4

    Differences in adjusted units processed per hour are statistically significant if P<0.05.

Adjusted3units processed per hour17.1918.48−1.29 (7.5%)0.1220
Standard error0.780.84  
95% Lower confidence limit15.6616.84  
95% Upper confidence limit18.7220.13  
Table 4. Annual Prevalence of Comorbidities by Major Diagnostic Category1 (HCV Versus Control)
MDC CategoryEmployees With HCV2Matched Employees Without HCV2DifferenceP-Value3
(N = 1,329)(N = 26,580)
  • 1

    For employees with HCV treatment, the index date is the date of the first prescription for ribavirin, interferon or peginterferon in the study period. For employees with HCV and without treatment, the index date is the average index date (by company) of employees with HCV and treatment. For employees without HCV, the index date is the average index date based on the group of employees with HCV and treatment.

  • 2

    Employees without HCV were matched 20:1 to employees with HCV by matching on propensity scores built from the demographic data.

  • 3

    Differences are considered significant if P < 0.05 and are highlighted in bold.

Infectious & parasitic disease73%13%60.46%<0.0001
Neoplasms19%13%5.88%<0.0001
Endocrine, nutritional, metabolic, immunity disorders34%27%6.16%<0.0001
Blood & blood forming organs12%4%8.47%<0.0001
Mental disorders20%10%9.92%<0.0001
Nervous system & sense organs31%24%6.69%<0.0001
Circulatory system36%28%8.32%<0.0001
Respiratory system42%32%10.64%<0.0001
Digestive system42%18%23.82%<0.0001
Genitourinary system35%28%6.66%<0.0001
Pregnancy, childbirth, puerperium2%3%−0.29%0.5233
Skin & subcutaneous tissue23%17%6.32%<0.0001
Musculoskeletal & connective tissue41%31%10.66%<0.0001
Congenital anomalies2%1%0.98%0.0035
Perinatal period1%0%0.22%0.2192
Injury and poisoning25%17%7.23%<0.0001
Other conditions68%49%19.37%<0.0001
Table 5. Annual Prevalence of Comorbidities for Selected AHRQ Specific Categories1 (HCV Versus Control)
Specific CategoryEmployees With HCV2)Matched Employees Without HCV2DifferenceP-Value3
(N = 1,329)(N = 26,580)
  • 1

    For employees with HCV treatment, the index date is the date of the first prescription for ribavirin, interferon or peginterferon in the study period. For employees with HCV and without treatment, the index date is the average index date (by company) of employees with HCV and treatment. For employees without HCV, the index date is the average index date based on the group of employees with HCV and treatment.

  • 2

    Employees without HCV were matched 20:1 to employees with HCV by matching on propensity scores built from the demographic data.

  • 3

    Differences are considered significant if P < 0 05 and are highlighted in bold.

Alcohol-related mental disorders2.03%0.42%1.61%<0.0001
Substance-related mental disorders3.31%1.27%2.04%<0.0001
Affective disorders7.00%3.04%3.95%<0.0001
Schizophrenia and related disorders0.30%0.08%0.23%0.0114
Other psychoses0.23%0.08%0.15%0.0889
Anxiety, somatoform, dissociative, and personality disorders5.72%3.55%2.17%<0.0001
Other mental conditions7.98%3.86%4.11%<0.0001
Personal history of mental disorder0.30%0.12%0.18%0.0839
Liver disease, alcohol-related1.28%0.07%1.21 %<0.0001

Discussion

Several studies have reported that HCV treatment is associated with absenteeism and reduced productivity,8, 9, 19 but none have outlined the workplace burden for HCV infection. Additionally, published studies focus on the amount of work time lost. When presented, absenteeism costs in other conditions use proxies such as self-reported costs,20 median costs from the age- and sex-adjusted full-time wage and salary workers in the United States population,21 or an average cost per unit of work time lost.22, 23 This study used actual payments for absences paid by the employers, and although similar to the other studies, it did not quantify or report the impact of the additional staffing required to cover these absences.

The current study controlled for a broad array of differences between employees with and without HCV. The factors with significant differences include age, marital status, race, full-time/part-time status, region, and Charlson comorbidity index.10 These analyses controlled for such differences using two-part regression methods that account for the nonnormal distributions of the absence and cost data18, 24 or by matching (in the comorbidity prevalence comparisons).

The HCV cohort had 1.8 times more absence days, driven by short-term disability claims (54%). Of note is the almost four-fold increase in long-term disability absences. The HCV cohort had 1.5 times more indirect costs than the controls. The difference in the impact between absence days and absence costs is likely due to differences in salary among the employees.

Overall, the HCV cohort was 3.2 times more costly than the controls. Direct costs were 88% of the HCV cohort, 75% of the controls, and 94% of the difference. In the HCV cohort, 34.7% the costs were incurred at the outpatient hospital or clinic, 31.8% at the inpatient hospital, and 28.5% in the physician's office. Not surprisingly, the comorbidity prevalence differences were led by infectious and parasitic disease (the category that includes HCV). Despite the incremental differences in prevalences of alcohol-related mental disorders, alcohol-related liver diseases, and substance-related mental disorders appearing low, the differences were significant and, compared with controls, the rates were 2.6, 4.8, and 18.3 times more likely in the HCV cohort.

The current study has several limitations. First, all subjects in this study were employees and must have visited the doctor and received an HCV-specific ICD-9 diagnosis during the study timeframe to be included in the cohort of employees with HCV. However, some employees with HCV symptoms may not have visited a doctor in response to those symptoms or were incorrectly diagnosed or not coded on their claims data. These employees would be included in the cohort of employees without HCV. This same challenge exists in the assessment of all comorbidities and severity of disease. It is difficult to assess the proportion of disorders (similar to those for HCV) that required a doctor visit and correct diagnosis due to the coding limitations. Also, because all subjects were employees, they may not be reflective of the overall HCV population (for example, lower prevalence of drug-related comorbid conditions). Furthermore, given the postindex timeframe for all study outcomes, preindex diagnoses and conditions might not be identified.

Additionally, there are limitations in the productivity results. The objective measures on productivity at work used in this analysis were from a subset of employees who worked in task-oriented positions and may not be easily generalized to other employee populations. Thus, measurements of the association of HCV on at-work productivity in the current study will help to establish that HCV is associated with the objective productivity output of employees, but should not necessarily be used as an estimate of the level of that impact for any given employee. The current study only addresses productivity losses at work, not impairments to other activities outside of work, and it only addressed the quality of work performed to the extent that employers require employees to redo poor-quality work. Therefore, further studies should be conducted to assess the impact of HCV on the quality of work performed in addition to the quantity. Lastly, the objective productivity findings may be understated compared with self-reported information. Lerner et al.25 reported a 2:1 relationship between illness-related self-reported productivity loss while at work and objectively measured productivity loss. That study found that for each 10% increase in work limitations employees reported on each of three Work Limitations Questionnaire scales, there was an approximate 4% to 5% reduction in objectively measured productivity.

In a retrospective claims-based design, there are certain factors in the study that cannot be controlled. Social factors such as work demands, job satisfaction, and job security were not available, and these omissions may lead to some selection bias. Another challenge associated with claims data is that severity measures such as disease stage or genotype are generally not available unless they are represented by a specific ICD-9. Future research should explore the impact of the comorbid conditions in Table 5 on study outcomes and should also attempt to control for nonlinear forms of the continuous control variables (age, salary, Charlson comorbidity index, and number of months of plan enrollment). Additionally, research examining the impact of treatment on outcomes and costs may shed new light on the value of therapy26 and whether any of the comorbid conditions identified in this study are caused by HCV or the therapies.

In conclusion, this study examined absences, productivity, and health benefit costs between HCV-infected and noninfected control patients. Within the relatively large population of employees studied in this analysis, HCV infection is associated with significant incremental health-related work absences, health benefit costs, and comorbidity. There is substantial work loss that contributes to the direct and indirect burden of illness for employees with HCV infection.

Acknowledgements

We thank Uchenna H. Iloeje (Bristol-Myers Squibb, Wallingford, CT) for input on the study design; Arthur K. Melkonian (Yerevan, Armenia) for assistance with the analysis of the data; and Jim Smeeding (the JeSTARx Group, Dallas, TX) for input on study design and the manuscript.

Ancillary