Bagalman, Erin MSW; Yu-Isenberg, Kristina S. PhD, RPh; Durden, Emily PhD; Crivera, Concetta PharmD, MPH; Dirani, Riad PhD; Bunn, William B. III MD, JD, MPH
* Review the characteristics and treatments for bipolar disorder (BPD), along with research data on its prevalence in the population and cost burden on employers.
* Summarize the new findings on BPD in the workplace, including rates of treatment nonadherence and its effects on different cost categories.
* Identify factors associated with treatment nonadherence among workers with BPD.
Bipolar disorder (BPD) is a psychiatric illness characterized by alternating episodes of mania (or less severe hypomania) and depression, separated by periods of normal moods.1 Interviews of US adults conducted between February 2001 and April 2003 as part of the National Comorbidity Survey Replication found a lifetime prevalence of 4.4% (SD 24.3) and a 12-month prevalence of 2.8% (SD 18.9) for any bipolar spectrum disorder, including type I, type II, and subthreshold BPD.2 Wyatt and Henter3 estimated the annual direct and indirect costs of BPD to be $45 billion (in 1991 US dollars), attributing $2.4B to inpatient care, $0.3B to outpatient care, and $0.13B to medication. Begley et al4 defined indirect costs more narrowly when they estimated lifetime costs for persons with BPD onset in 1998 to be $24B (in 1998 US dollars), attributing $13B to direct costs.
BPD imposes a substantial cost burden on employers.5 In a retrospective analysis of employee health benefits data from multiple large employers, Gardner et al6 found that employees with BPD annually cost $6836 more than employees without BPD and were significantly more costly in every health benefit cost category. A study of health and productivity data from six large corporations found BPD to be the most expensive psychiatric disorder, with half of the cost attributable to absence and short-term disability (STD).7 In a study using the National Comorbidity Survey Replication and the World Health Organization Composite International Diagnostic Interview to assess the prevalence of both major depressive disorder and BPD in the workforce, Kessler et al8 found that BPD was associated with 65.5 lost workdays per ill worker per year. Other studies have found that patients with BPD have higher costs for absence, STD, and workers' compensation (WC) relative to matched controls without psychiatric illness1 or employees with major depression.9
Several studies have compared direct or indirect costs between patients with and without BPD, but only a few have compared costs among patients with BPD. In one such study evaluating direct health care costs of BPD employees and dependents before and after treatment initiation, Brook et al found that prescription costs increased, whereas medical costs decreased for most patients. Among patients who commenced on atypical antipsychotic treatment, the adjusted incremental change in prescription costs was $656 and the adjusted incremental change in total medical costs was −$2886. The study did not address indirect costs of work loss, nor did it address the effect of medication adherence.10
Adherence to treatment is a major determinant of outcome in BPD; nonadherence and partial adherence both play a significant role in relapse.11 The prevalence of nonadherence with mood stabilizers ranges from ∼18% to 52%.12,13 In a retrospective analysis of electronic prescription and medical claims representing ∼1.4 million managed-care commercial health plan members with mental health benefits, Lew et al14 found that reduced adherence to traditional mood-stabilizing therapy in patients with BPD was associated with significantly greater risk of mental health-related emergency room visits (odds ratio [OR] 1.98; 95% confidence interval [CI] = 1.38–2.84) and inpatient hospitalizations (OR 1.71; 95% CI = 1.27–2.32), after adjusting for age, gender, and comorbidities. In a study examining the risk of relapse after hospitalization for BPD, Hassan and Lage15 found that greater medication adherence (≥75%) was associated with significantly lower odds of any rehospitalization (OR 0.730; 95% CI = 0.580–0.919) or psychiatric rehospitalization (OR 0.759; 95% CI = 0.603–0.955).
Although several existing studies address direct and indirect costs of BPD versus non-BPD comparisons, and some address direct costs of adherent versus nonadherent BPD patients, to the best of our knowledge, no study has addressed work-loss costs of adherent versus nonadherent BPD employees. This study aims to examine the association between treatment adherence and work-loss outcomes among a sample of employees with BPD.
MATERIALS AND METHODS
This retrospective analysis used productivity loss data linked to medical claims for employees with both medical and pharmacy benefits obtained from the MarketScan® Research Databases of Thomson Reuters. The MarketScan medical claims databases are derived from employer and government-funded health insurance plans in the United States, providing de-identified medical and pharmacy claims for several million employees and their dependents, covered under a variety of fee-for-service and capitated health plans. Medical and pharmacy claims are linked to enrollment files, which include basic demographic data, by a unique identification number assigned to each enrollee. Productivity loss data are available for a subset of employees in the larger MarketScan health claims database, allowing for the assessment of the relationship between adherence to treatment (from the medical claims) and productivity loss. The types of productivity loss captured in the MarketScan databases include paid absence or paid time off (PTO), STD, and WC.
Adults aged 18 to 64 years having at least two medical claims with a diagnosis of BPD (see Appendix A for diagnoses), one prescription claim for a mood stabilizer or an atypical antipsychotic (see Appendix B for drugs), continuous medical and prescription coverage (including mental health benefits) for at least 6 months before and 12 months subsequent to the prescription from January 1, 2001, through December 31, 2004, were initially identified for analysis. To ensure that all behavioral health services were captured (even if a particular plan had a behavioral health carve-out), only data contributed by employers (as opposed to health plan contributors) were used in this analysis. Patients with evidence of pregnancy or childbirth during the 18-month period were excluded. Finally, patients with data tracking workplace absence, whether through PTO, STD, or WC, were retained for analysis.
The index date was the date of the first prescription claim for a mood stabilizer or atypical antipsychotic with at least 6 months of prior continuous medical and prescription coverage and at least 12 months of subsequent continuous medical and prescription coverage. Patients were followed for 18 months, beginning 6 months prior to index (baseline) and ending 12 months after index (study period), to evaluate adherence to treatment and productivity loss.
Patients were categorized as adherent or nonadherent based on the medication possession ratio (MPR), calculated as the number of ambulatory days (when available) supplied of any bipolar treatment (ie, at least one of any type of bipolar treatment) divided by the number of ambulatory days remaining in the period after the first index medication (ie, the length of the follow-up period = 365 days if no hospital days). Patients were considered adherent if their MPR was ≥0.8, which is a commonly used threshold found in the adherence to treatment literature.16–20
Demographic and Clinical Characteristics
Patient demographic characteristics included age at index, gender, region of US residence, population density (urban vs rural residence), salaried versus hourly pay, full-time versus part-time employment status, and insurance plan type (those with capitated payment arrangements vs other). A proxy measure of socioeconomic status, median household income within the ZIP code of residence, derived from the Census Bureau (2000) data. Clinical variables calculated during the baseline period included the Charlson Comorbidity Index (CCI, Deyo version), number of psychiatric diagnoses other than BPD, baseline emergency department visits, and baseline hospital admissions. In addition, flags indicating diagnoses of depression, substance abuse, anxiety disorder, epilepsy, diabetes, hypertension, and high cholesterol levels during the study period were created.
As noted above, adherence was measured using the MPR. This was calculated for all bipolar treatments as a class, ie, ambulatory days supplied were counted for mood stabilizers and atypical antipsychotics together. In addition, MPR was coded into a dichotomous variable, with values of adherent (MPR ≥0.80) and nonadherent (MPR = 0.00–0.79).
Productivity Loss and Associated Costs
Workplace productivity loss was captured by absence or PTO, STD, and WC. For each category of work loss, the number of patients reporting data, the number of patients reporting loss/use, the average number of days lost, and the average associated indirect costs were evaluated. The cost of absenteeism was imputed using 2005 hourly wage data from the Bureau of Labor Statistics, matched by age group, sex, and census region; for patients with no match, the mean hourly rate for the sample with data was used ($18.22).21 The average hourly wage rate was multiplied by 8 to arrive at an estimated cost per day associated with absence or PTO. The costs associated with STD days were calculated using the same method, but the wage value was multiplied by 0.60, with the assumption that a typical STD program benefit compensates for 60% of an employee's pay and benefits.7 The costs associated with WC reflect the actual WC payments made to recipients during the follow-up period.
Frequency distributions, means, and standard deviations were calculated for all variables. The descriptive analyses were stratified by adherence status (adherent versus nonadherent). Bivariate statistical testing (ie, χ2 tests for categorical variables and t tests for continuous variables) was used to evaluate differences between the adherent and the nonadherent groups. Values of P < 0.05 were considered statistically significant.
The adjusted indirect costs associated with work loss of the adherent and nonadherent groups were assessed by means of two-part multiple regression analyses. The two-part models estimating costs due to work loss adjusted for age, sex, geographic region, median neighborhood household income, hourly versus salaried employee status, health plan type (capitated health plan or other), CCI, CCI-squared, substance abuse, diabetes, hypertension, high cholesterol levels, and adherence status. Each two-part model consisted of a logistic regression model that estimated the probability of incurring positive dollars in each category of work loss (ie, PTO, STD, or WC) and a general linear model with a log link function that predicted the amount of expenditure among those with expenditure. The predicted probability of having any expenditure (obtained from the first model) was multiplied by the predicted magnitude of the expenditure (obtained from the second model) to obtain the total predicted work-loss expenditure. The expected level of costs is given by
Equation (Uncited)Image Tools
The study population comprised 1258 patients, of whom 50.9% were women and 29.6% were younger than 35 years (Table 1). Nearly two thirds of the sample (64.7%) was nonadherent to bipolar treatment during the follow-up period. Half of the sample was paid hourly and 89.9% resided in an urban area. More than half of the patients (57.6%) included in the sample participated in a health plan without capitated payment arrangements. Tests of differences in patient demographic characteristics showed several significant differences between patients in the adherent and nonadherent groups. Adherent patients were slightly older than their nonadherent counterparts (41.2 years, SD 9.0 vs 39.4 years, SD 9.4, respectively) and were more likely to be male (53.8% vs 46.6%, respectively) and salaried (41.7% vs 33.2%, respectively).
On the index date, the majority of patients (62.7%) were taking an anticonvulsant medication, 30.6% were taking an atypical antipsychotic, and 14.5% were treated with lithium (Table 2). Less than one tenth of patients (8.9%) were treated with both a mood stabilizer and an atypical antipsychotic at index. Nearly two thirds of patients (61.2%) were treated with a selective serotonin reuptake inhibitor antidepressant medication during the follow-up period. The average (SD) CCI score of the sample was 0.18 (SD 0.57). Most patients (61.6%) had a psychiatric diagnosis other than BPD during the pre-period, including depression (46.6%), anxiety (15.7%), and substance abuse (6.1%). The average MPR for the sample was 0.58 (SD 0.31).
Several significant differences between patients in the adherent and nonadherent groups were observed. Relative to adherent patients, the nonadherent were less likely to be taking lithium at index (20.3% for the adherent and 13.0% for the nonadherent) and were less likely to be treated with a combination regimen of a mood stabilizer and an atypical antipsychotic (12.2% for the adherent and 7.1% for the nonadherent). Relative to adherent patients, the nonadherent were more likely have a recent diagnosis of BPD (46.8% for the adherent and 55.9% for the nonadherent) and were also more likely to be new users of their index medication(s) (72.3% for the adherent and 88.0% for the nonadherent). A greater proportion of adherent patients had a visit with a mental health professional before their index medication than did the nonadherent (20.0% vs 15.1%, respectively).
Productivity loss due to PTO, STD, and WC was evaluated separately for each patient cohort (ie, the adherent and the nonadherent). As shown in Table 3, slightly higher proportions of nonadherent patients had at least 1 day of work loss due to PTO (86.3% for the adherent and 88.9% for the nonadherent), STD (21.4% for the adherent and 27.6% for the nonadherent), and WC (4.0% for the adherent and 8.1% for the nonadherent). Nonadherent patients incurred greater average unadjusted indirect costs than adherent patients due to workplace absence ($3534.11, SD $3377.81 for the adherent and $3997.15, SD $5725.65 for the nonadherent), STD ($1861.29, SD $5331.70 for the adherent and $2208.26, SD $5167.54 for the nonadherent), and WC ($344.55, SD $1961.56 for the adherent and WC $1004.91, SD $5791.50 for the nonadherent) Furthermore, nonadherent patients with work loss due to absence and WC incurred more time off than adherent patients with these types of work loss.
Figure 1 presents estimates of indirect costs due to each type of work loss for the adherent and the nonadherent cohorts, adjusted for differences in the covariates in the two-part models. The average adjusted indirect costs associated with PTO are estimated at $3226.56 for adherent patients and $3997.97 for the nonadherent. The adjusted average indirect costs associated with STD are estimated at $1848.13 for adherent patients and $2133.85 for the nonadherent. Finally, the average adjusted indirect costs associated with WC are estimated at $401.16 for adherent patients and $761.79 for the nonadherent.
Potential Indirect Savings Related to Treatment Adherence
Results of the analysis reported above were used to calculate the potential annual cost savings to employers associated with adherence. The 334 employees reporting absence data who were nonadherent had an absence incidence rate of 88.9% (n = 297) in the 1-year follow-up period. The odds of having an absence event among the nonadherent group are 31.6% higher than the odds of absence use among the adherent (based on the 1.316 OR after adjusting for confounding factors [95% CI = 0.726–2.381]). Converting odds into incidence rates and vice versa, we calculate that the absence event rate for the adherent group of employees could have been 85.9% (calculated using
, which translates to 287 employees with any absence incidence (0.859 × 334 employees).
Equation (Uncited)Image Tools
The potential number of absence cases saved would be 10 (297 employees − 287 employees). The average number of absence days per year among those with absence incidence is 29, so potential absence days saved would be 290 (10 employees × 29 days). Assuming a wage rate of $145.76 per employee per day, the total potential indirect savings in this study population (N = 1258) would be $42,270.40 (290 days × $145.76 per day). The average savings for one employee taking bipolar treatment is $33.60 ($42,270.40/1258). In the MarketScan Commercial database, we find that 3.3% of employees with pharmacy benefits fill at least one prescription for a medication used to treat BPD in a year. In a company with 70,000 employees, that translates to 2100 people taking any bipolar treatment. If all employees adhere to treatment, there is a potential savings in absence costs of $70,560 ($33.60 × 2100 employees).
With respect to STD, the 497 employees reporting STD data who were nonadherent had an STD incidence rate of 27.6% (n = 137) in the 1-year follow-up period. The odds of STD use among the nonadherent group are 31.0% higher than the odds of STD use among the adherent (based on the 1.305 OR after adjusting for confounding factors [95% CI = 0.912–1.869]). Therefore, the STD event rate for this group of employees could have been 22.6% (calculated using
, which translates to 112 employees with any STD incidence (0.226 × 497 employees).
Equation (Uncited)Image Tools
The potential number of STD cases saved would be 25 (137 employees − 112 employees). The average number of STD days per year among those with any incidence is 57, so potential STD days saved would be 1425 (25 employees × 57 days). Assuming a 60% STD benefit for an average wage rate of $145.76 per employee per day, the total potential indirect savings in this study population (N = 1258) would be $124,630.50 (1425 days × $87.46 per day). The average savings for one employee taking bipolar treatment is $99.07 ($124,630.50/1258). As above, in a company with 70,000 employees, a 3.3% rate translates into 2100 people taking any BPD treatment. If all employees adhere to treatment, there is a potential savings in STD costs of $208,048 ($99.07 × 2100 employees).
Finally, with respect to WC, the 418 employees reporting WC data who were nonadherent had a WC incidence rate of 8.1% (n = 34) in the 1-year follow-up period. The odds of a WC event for the nonadherent group are 99.0% higher than the odds of WC use among the adherent (based on the 1.992 OR after adjusting for confounding factors [95% CI = 0.904–4.405]). Therefore, the WC event rate for this group of employees could have been 4.3% (calculated using
), which translates to 18 employees with any WC incidence (0.043 × 418 employees).
Equation (Uncited)Image Tools
The potential number of WC cases saved would be 16 (34 employees − 18 employees). The average number of WC days per year is 77, so potential WC days saved would be 1232 (16 employees × 77 days). Assuming a wage rate of $145.76 per employee per day, the total potential indirect savings in this study population (N = 1258) would be $179,576 (1232 days × $145.76 per day). The average savings for one employee taking bipolar treatment is $142.75 (179,576/1258). Again, in a company with 70,000 employees, a 3.3% rate translates into 2100 people taking any BPD treatment. If all employees adhere to treatment, there is a potential savings in WC costs of $299,775 ($142.75 × 2100 employees).
This retrospective database analysis used commercial health plan claims and employee productivity data to evaluate the indirect costs associated with adherence to treatment for BPD. Consistent with previous research indicating nonadherence to be related to negative outcomes,14,18 our analysis revealed that the indirect costs due to lost workplace productivity were higher for workers who were nonadherent than for those who were adherent. For each type of productivity loss considered—absence or PTO, STD, and WC—the adjusted indirect costs were higher for workers with BPD who were nonadherent to treatment than for those who were adherent.
We used the results of our analysis to extrapolate potential costs associated with nonadherence to bipolar treatment to a hypothetical employer with 70,000 employees, assuming an incidence rate for bipolar treatment of 3.3%. These hypothetical scenarios indicate potential annual cost savings of $70,560 in indirect costs due to absence, $208,048 in indirect costs due to STD, and $299,770 in indirect costs due to WC if all employees adhere to bipolar treatment.
It should be noted that the 18-month continuous enrollment requirement limited the analysis to individuals with BPD capable of maintaining long-term employment; this may have excluded employees with more severe BPD and understated the true work-loss costs associated with nonadherence in BPD.
Although the MPR is frequently used as a measure of adherence, it is operationalized differently in different studies. As a result, adherence may not be comparable across studies. The average MPR in this study (0.58, SD 0.31) was substantially lower than the average MPR found by Gianfrancesco et al22 in a commercially insured population (>0.90 for all atypical antipsychotics evaluated). The average MPR found in this study is more consistent with the average MPR found by Hassan et al23 in a study of adherence within a Medicaid population (0.63, SD 0.32). Hassan and Lage, however, found a lower average MPR (0 0.46, SD 0.32) among patients recently hospitalized for BPD.15
Bipolar treatment may also be categorized differently across studies. One classification system used in this study (categorizing index treatment as atypical antipsychotic only, mood stabilizer only, or both) is similar to the classification used by Brook et al (atypical antipsychotic only, other class only, or both). This study found that 21.7% of employees commenced on atypical antipsychotics alone. In contrast, Brook et al found that only 4.7% of employees initiating pharmacologic treatment for BPD received atypical antipsychotics alone. The difference may be partially explained by an increase in the use of atypical antipsychotic medications over time: Brook et al used 2001–2004 data, whereas this study included 2005 data.10
This analysis has several limitations. First, the analysis is based on a convenience sample of employees of large, self-insured corporations for whom workplace data were available in the MarketScan database. Second, because this study analyzed administrative claims data, several assumptions were made, including that clinical diagnoses are accurate, that medical codes are entered correctly, and that a prescription filled is a prescription used. A third limitation of our study is the reliance on absence, STD, and WC as the sole measures of productivity. Additional measures of productivity, such as presenteeism, may provide a more comprehensive understanding of the indirect costs associated with adherence to bipolar treatment, representing a potentially fruitful avenue for future research.
In summary, this study finds that the indirect costs associated with three types of productivity loss—absence, STD, and WC—are lower among employees with BPD who are adherent to bipolar treatment than for those who are nonadherent to bipolar treatment.
The authors wish to acknowledge the technical and editorial support provided by Dr Matthew Grzywacz, PhD, and Helix Medical Communications.
This work is supported by funding from Ortho-McNeil Janssen Scientific Affairs, LLC.
1. Montejano LB, Goetzel RZ, Ozminkowski RJ. Impact of bipolar disorder on employers: rationale for workplace intervention. Dis Manag Health Outcomes. 2005;13:267–280.
2. Merikangas KR, Akiskal HS, Angst J, et al. Lifetime and 12-month prevalence of bipolar spectrum disorder in the National Comorbidity Survey Replication. Arch General Psychiatry. 2007;64:543–552.
3. Wyatt RJ, Henter I. An economic evaluation of manic-depressive illness—1991. Soc Psychiatry Psychiatr Epidemiol. 1995;30:213–219.
4. Begley CE, Annegers JF, Swann AC, et al. The lifetime cost of bipolar disorder in the US: an estimate for new cases in 1998. Pharmacoecononmics. 2001;19(5 Part 1):483–495.
5. Sajatovic M. Bipolar disorder: disease burden. Am J Manag Care. 2005; 11(3 suppl):S80–S84.
6. Gardner HH, Kleinman NL, Brook RA, Rajagopalan K, Brizee TJ, Smeeding JE. The economic impact of bipolar disorder in an employed population from an employer perspective. J Clin Psychiatry. 2006;67:1209–1218.
7. Goetzel RZ, Hawkins K, Ozminkowski RJ, Wang S. The health and productivity cost burden of the “top 10” physical and mental health conditions affecting six large U.S. employers in 1999. J Occup Environ Med. 2003;45:5–14.
8. Kessler RC, Akiskal HS, Ames M, et al. Prevalence and effects of mood disorders on work performance in a nationally representative sample of U.S. workers. Am J Psychiatry. 2006;163:1561–1568.
9. Bowden CL. Bipolar disorder and work loss. Am J Manag Care. 2005; 11(3 suppl):S91–S94.
10. Brook RA, Kleinman NL, Rajagopalan K. Employee costs before and after treatment initiation for bipolar disorder. Am J Manag Care. 2007;13:179–186.
11. El-Mallakh RS. Medication adherence and the use of long-acting antipsychotics in bipolar disorder. J Psychiatr Pract. 2007;13:79–85.
12. Scott J, Pope M. Nonadherence with mood stabilizers: prevalence and predictors. J Clin Psychiatry. 2002;63:384–390.
13. Sajatovic M, Valenstein M, Blow F, Ganoczy D, Ignacio R. Treatment adherence with lithium and anticonvulsant medications among patients with bipolar disorder. Psychiatr Serv. 2007;58:855–863.
14. Lew KH, Chang EY, Rajagopalan K, Knoth RL. The effect of medication adherence on health care utilization in bipolar disorder. Manag Care Interface. 2006;19:41–46.
15. Hassan M, Lage MJ. Risk of rehospitalization among bipolar disorder patients who are nonadherrent to antipsychotic therapy after hospital discharge. Am J Health Syst Pharm. 2009;66:358–365.
16. Ahn J, McCombs JS, Jung C, et al. Classifying patients by antipsychotic adherence patterns using latent class analysis: characteristics of nonadherent groups in the California Medicaid (Medi-Cal) program. Value Health. 2008;11:48–56.
17. Lage MJ, Hassan MK. The relationship between antipsychotic medication adherence and patient outcomes among individuals diagnosed with bipolar disorder: a retrospective study. Ann Gen Psychiatry. 2009;8:7.
18. Gilmer TP, Dolder CR, Lacro JP, et al. Adherence to treatment with antipsychotic medication and health care costs among Medicaid beneficiaries with schizophrenia. Am J Psychiatry. 2004;161:692–699.
19. Sajatovic M, Valenstein M, Blow FC, Ganozy D, Ignacio RV. Treatment adherence with antipsychotic medications in bipolar disorder. Bipolar Disord. 2006;8:232–241.
20. Woltmann EM, Valenstein M, Welsh DE, et al. Using pharmacy data on partial adherence to inform clinical care of patients with serious mental illness. Psychiatr Serv. 2007;58:864–867.
21. Bureau of the Census. Current Population Survey, 2005 Annual Social and Economic (ASEC) Supplement [machine-readable data file]. Washington, DC: Bureau of the Census; 2005.
22. Gianfrancesco FD, Rajagopalan K, Sajatovic M, Wang RH. Treatment adherence among patients with bipolar or manic disorder taking atypical and typical antipsychotics. J Clin Psychiatry. 2006;67:222–232.
23. Hassan M, Madhavan SS, Kalsekar ID, et al. Comparing adherence to and persistence with antipsychotic therapy among patients with bipolar disorder. Ann Pharmacother. 2007;41:1812–1818.
APPENDIX B: MEDICATIONS USED TO TREAT BIPOLAR DISORDER
TABLE A1 ICD-9-CM Co...Image Tools
Conventional Antipsychotic Agents
Atypical Antipsychotic Agents
Risperidone (including injectable)
Ziprasidone mesylate Cited Here...