Employee Costs Before and After Treatment Initiation for Bipolar Disorder

, ,
The American Journal of Managed Care, April 2007, Volume 13, Issue 4

Objective: To examine pretreatment and posttreatment total medical costs and overall mental healthcare costs for patients with bipolar disorder (BPD) treated with different medication regimens (alone and in combination) vs an untreated (UnTx) cohort.

Study Design: Retrospective employer-based administrative database analysis of costs before and after the start of therapy for BPD from 2001 through 2004.

Methods: Patients were grouped into 3 cohorts based on type of therapy vs the UnTx cohort. Total medical and mental health-specific healthcare costs were compared between the 6-month preindex period and the 6-month postindex period. A mean index date of the treated cohorts was assigned to the UnTx cohort. Regression models were used to calculate cost differences.

P

P

P

Results: Reductions in direct medical costs among 1284 patients were largest for the cohort receiving atypical antipsychotics (ATYP) only (-$2886 [n = 55]), followed by the UnTx cohort (-$365 [n = 306]) and the cohort receiving ATYP plus other BPD medications (BOTH) (-$78 [n = 369]). In the cohort receiving other BPD medications (OTHR), costs increased by $168 (n = 554). Differences between the ATYP cohort and the OTHR cohort were significant (= .04). For specific direct mental health-related costs, the cost changes were -$1523 for the ATYP cohort, -$441 for the OTHR cohort, -$38 for the BOTH cohort, and -$704 for the UnTx cohort. Differences between the ATYP cohort and the OTHR and BOTH cohorts were significant (= .02 and = .002, respectively).

Conclusions: Patients using ATYP for BPD seem to have the largest cost reductions. Additional investigation is needed to identify whether the UnTx cohort had the least severe BPD, had nonadherent prescription fill behavior, or both.

(Am J Manag Care. 2007;13:179-186)

Employees with bipolar disorder (BPD) have been found to have annual total direct and indirect medical costs approximately 3 times as high as those for employees without BPD. Employers should be aware that:

  • The prevalence of BPD among 1284 employees in this study was 0.3%.
  • Patients using atypical antipsychotics for BPD had the largest direct medical cost reductions during the 6-month posttreatment period.

Bipolar disorder (BPD) extracts a heavy economic burden in the United States, with a 1991 cost estimate for direct inpatient and outpatient care placed at $7.6 billion and at $38 billion for indirect costs.1 In previous investigations exploring costs in managed care settings, employees with BPD have been found to have annual total direct and indirect medical costs that were approximately 3 times as high as those for employees without BPD, and these costs were greater for every component of healthcare studied, including medical, prescription, and absence costs.2 Total medical costs have been found to be higher for mental health-related and nonmental health-related care in managed care settings,3 with the greatest proportion of mental healthcare costs being related to inpatient admissions.4

There is little research available to examine the effect of treatment initiation on costs in managed care settings and in particular on total medical costs. Most research has compared persons with BPD with those without BPD,5 some research has compared persons with BPD with persons with other diseases,6 and other research is based on trials conducted in populations treated at Veterans Affairs hospitals7 and at mental health clinics.8 In addition, there has been little research to investigate the effect of the increasing introduction of new psychotropic drug classes to treat BPD on these cost variables. The primary aim of this study was to examine the demographic differences among employees treated with different classes of psychotropic drugs. We then assessed changes before and after treatment initiation in the following: (1) total medical costs (excluding prescription costs), (2) costs in the subcategories of nonmental health-related and mental health-related medical care, and (3) total prescription costs. In addition, the study aimed to determine differences in the likelihood of mental health-related emergency department (ED) visits and inpatient admissions before and after treatment initiation among patients taking different classes of psychotropic drugs.

METHODS

Data Source

International Classification of Diseases, Ninth Edition

ICD-9-CM

Data for this retrospective analysis were taken from the Human Capital Management Services Research Reference Database. The database contained health insurance claims information, including those for demographics, payroll, healthcare (with prescription drug data), disability, absence, and workers' compensation, for a population of more than 300 000 employees and their covered dependents. These data were compiled from several large national US employers and represented retail, service, manufacturing, and financial industries. Employees and spouses with a diagnosis of BPD (type I or type II) were identified from the overall database if they were noted to have a primary, secondary, or tertiary () diagnostic code for BPD (code 296.0x, 296.1x, 296.4x, 296.5x, 296.6x, 296.7x, or 296.8x) from 2001 through 2003.

Cohort Classification

Four cohorts of patients with BPD were identified based on their treatment patterns. The first cohort was defined as employees and spouses with BPD who received only atypical antipsychotics (ATYP). The second cohort was defined as employees and spouses with BPD who received at least 1 medication from a category labeled as OTHR (taking other BPD medications) that comprised conventional antipsychotics, primary mood stabilizers (including carbamazepine, valproic acid, divalproex sodium, and lithium carbonate), potential mood stabilizers (including additional anticonvulsants used in the treatment of BPD), zonisamide, and tiagabine hydrochloride. Individuals in the ATYP and OTHR cohorts could receive more than 1 drug from their designated drug category sequentially or simultaneously. The third treatment cohort took an atypical antipsychotic and an OTHR sequentially or simultaneously (BOTH). Assignment to the ATYP, OTHR, or BOTH cohorts was based on 1 or more dispensed prescriptions. The fourth untreated (UnTx) cohort had a BPD diagnosis but no BPD medication claims. The cohort classification is given in Table 1.

Index Date

The analysis plan established index dates for each cohort based on the start of therapy. An index prescription date was defined for patients in the ATYP, OTHR, and BOTH cohorts as the date of the patient's first BPD-related prescription following his or her BPD diagnosis. Patients in these 3 cohorts were required to have health insurance enrollment for 6 months before and after the index prescription date. To reduce confounding of results, there could be no record of any BPD-related medication prescription in the 6-month period before this index date. The pretreatment period was defined as the 6-month period before the index prescription date, and the posttreatment period was defined as the 6-month period that followed. Individuals in the ATYP cohort received no prescriptions from the OTHR classification during the posttreatment period, and individuals in the OTHR cohort received no ATYP during the posttreatment period.

On average, the index prescription date was found to be 121 days after the index diagnosis date (defined as the date of the first diagnosis) for the 3 cohorts that were prescribed medications. Therefore, the index prescription dates for the UnTx cohort were defined to be 121 days after their index diagnosis dates. Patients were excluded from this cohort if they did not have health insurance enrollment for at least 6 months before and after the newly created index prescription date or if they received a BPD-related prescription during this period.

Statistical Analysis

t

P

ICD-9-CM

ICD-9-CM

ICD-9-CM

ICD-9-CM

The mean values for the demographic data of the 4 cohorts were calculated and compared using tests for continuous variables and ?2 tests for discrete variables. Results were considered significant at < .05. Unadjusted semiannual cost differences before and after the index prescription date were calculated for all 4 cohorts. Total pretreatment medical costs (excluding prescription costs) were defined as all medical claims insurance payments related to any code during the pretreatment period, with the total posttreatment medical costs defined similarly. Prescription costs were defined as all pharmacy and mail-order prescription drug claim insurance payments during similar pretreatment and posttreatment periods. Mental health-related medical costs (again excluding prescription drug costs) were calculated from claims related to any code from the Agency for Healthcare Research Quality major diagnostic category of mental disorder,9 while medical costs related to all other codes not included in the mental disorder classification were classified as nonmental health-related medical costs. The likelihood of a mental health-related ED visit was calculated during a 6-month period by including all ED visits related to an code from the Agency for Healthcare Research Quality mental disorder major diagnostic category, with a similar method used to calculate the likelihood of a mental health-related inpatient admission.

Comparisons were made for the following: (1) total medical costs, excluding prescription costs; (2) total prescription costs; (3) mental health-related medical costs; (4) nonmental health-related medical costs; (5) the likelihood of a mental health-related ED visit; and (6) the likelihood of a mental health-related inpatient admission.

To calculate adjusted values of semiannual cost and likelihood changes, linear regression models were used to measure the effect of drug treatment class on the following dependent variables: change in medical costs, change in prescription drug costs, change in mental health-related medical costs, change in nonmental health-related medical costs, change in the likelihood of a mental health-related ED visit, and change in the likelihood of a mental health-related inpatient admission. The models were designed to control for the effects of other confounding factors (where applicable and after examining collinearity). These factors included age, sex, pretreatment BPD-specific healthcare costs, other pretreatment healthcare costs, pretreatment Charlson Comorbidity Index scores,10 region (as designated by the first digit in the subject's ZIP code), and an employee versus spouse indicator. Variables representing the interaction between the employee versus spouse indicator and marital status, race/ethnicity, full-time or part-time status, and annual salary were also included in the models. The interaction variables were included because some information was only available for employees and not for spouses. Version 9.1 of SAS System for Windows (SAS Institute, Inc, Cary, NC) was used to generate all statistical analyses.

RESULTS

Overall, 1542 patients (employees and spouses) were identified with BPD. The prevalence of the condition among employees was found to be 0.3%.11 The mean age of the overall patient population with BPD was 41.60 years (95% confidence interval, 41.15-42.05 years), and the mean percentage of employees (vs spouses) in the patient population with BPD was 60.8% (95% confidence interval, 58.4%-63.3%).

P

P

P

P

P

P

P

Table 2 gives the overall descriptive demographic characteristics of the 4 cohorts by cohort. Patients in the BOTH cohort were more likely to be female than patients in the UnTx cohort (< .05). Table 2 reflects that certain elements, such as tenure, annual salary, marital status, race/ethnicity, exempt status, and full-time status, were only available for employees and not for spouses. The UnTx cohort had a significantly higher percentage of employees (vs spouses) than the ATYP and BOTH cohorts (< .05). Patients in the OTHR cohort were also more likely to be employees than patients in the BOTH cohort (< .05). In addition, the employees in the OTHR and UnTx cohorts were more likely to be married than the employees in the BOTH cohort (< .05). In the UnTx cohort, employees also had higher annual salaries than employees in the BOTH and OTHR cohorts ($56 089 compared with $45 725 and $50 075, respectively; < .05). On the whole, employees in the OTHR and UnTx cohorts had higher annual salaries than employees in the ATYP and BOTH cohorts. In addition, employees in the OTHR cohort were more likely to be in salaried positions (exempt) than patients in the BOTH cohort (< .05), and employees in the UnTx cohort were more likely to work full-time than employees in the ATYP and BOTH cohorts (< .05).

P

P

P

P

P

The adjusted incremental changes in total medical costs before and after treatment initiation are shown as blue bars in Figure 1. Using these adjusted values, the incremental change for the ATYP cohort was significantly lower than for the OTHR cohort ($3054 difference, = .04) and approached significance when compared with the BOTH cohort (-$78, = .07), bringing the costs incurred during the 6 months after treatment initiation with ATYP down to a value similar to that found for patients in the OTHR cohort. The adjusted incremental changes in prescription drug treatment costs before and after treatment initiation are shown as red bars in Figure 1 and indicate that the value calculated for the UnTx cohort ($59) was significantly lower than for all other cohorts (< .001 for all comparisons). The comparison between the OTHR and BOTH cohorts was also significant at < .001. The difference between the ATYP and OTHR cohorts approached significance (= .07).

P

P

P

P

P

The adjusted changes in nonmental health-related medical costs are shown as blue bars in Figure 2. None of the differences in adjusted changes achieved significance. However, the 2 cohorts that experienced reductions were taking ATYP. The adjusted changes in mental health-related medical costs are shown as red bars in Figure 2 and indicate that the ATYP cohort had significantly greater decreases than the OTHR and BOTH cohorts (= .02 and = .002, respectively). The adjusted change for the UnTx cohort was significantly lower than for the BOTH cohort (= .01). The ATYP cohort's lower adjusted change value approached significance compared with the value for the UnTx cohort (= .09). The comparison between the BOTH and OTHR cohorts approached significance (= .08), with the OTHR cohort having the lower value.

P

P

P

The adjusted changes in semiannual mental health-related ED visit rates per patient are shown as blue bars in Figure 3. The adjusted change in the likelihood of mental health-related ED visits for the BOTH cohort was significantly lower than for the OTHR (= .008) and UnTx (= .03) cohorts. The adjusted changes in the likelihood of mental health-related inpatient admissions are shown as red bars in Figure 3. The rate decrease for the ATYP cohort was more pronounced than for the OTHR cohort (= .08).

DISCUSSION

Retrospective database analyses such as this one that evaluate pretreatment and posttreatment health benefit claims among patients with BPD allow for an examination of differential rates of resource use and costs associated with different classes of psychotropic medications (or no medication). Based on demographics, costs, and the likelihood of ED visits and inpatient admissions, the results show that the 4 cohorts seem to cluster in the following 2 larger groups: (1) employees and spouses in the OTHR cohort and the UnTx cohort and (2) those in the ATYP cohort and the BOTH cohort. These demographic results suggest that patients in the OTHR and UnTx cohorts may fall into categories of BPD that were less severe or that they are more likely to be adherent with their treatment regimens.

The largest number of patients (n = 554) was in the OTHR cohort, which had lower total medical costs than any other cohort (before and after treatment initiation) and a low likelihood of ED visits and inpatient admissions. Although individuals in the UnTx cohort also had a low likelihood of ED visits and inpatient admissions, they had high total medical costs before and after treatment initiation, which seemed to be primarily related to nonmental health-related diagnostic categories.

Individuals in the ATYP and BOTH cohorts had the highest likelihood of inpatient admissions. Individuals in the ATYP cohort had percentages of mental health-related medical costs that were similar to those of individuals in the OTHR cohort, but the decrease in costs after treatment was significantly greater than that found for the OTHR cohort. In fact, individuals in the ATYP cohort had the largest adjusted decreases in total medical costs and in nonmental health-related and mental health-related medical costs of all cohorts analyzed. Therefore, the total medical and mental health-related decreases in costs seem to be secondary, in part because of the large decrease in ED visits and inpatient admissions that occurred after treatment initiation for the ATYP cohort. The BOTH cohort had the highest total medical costs of all cohorts, and these costs remained high after treatment initiation. Although there were large decreases in ED visits and inpatient admissions in this cohort, adjusted values showed little change in mental health-related medical costs before and after treatment.

Several explanations can be postulated retrospectively to explain the differences found in this study. As noted, the demographics suggest that patients in the ATYP and BOTH cohorts tended to be more severely ill than individuals in the OTHR and UnTx cohorts, and this hypothesis is substantiated by the analysis of ED visits and inpatient admissions. Because the ATYP and BOTH cohorts included patients treated with ATYP, an evaluation of treatment guidelines at the time of data collection (2001-2003) for the suggested use of this class of medication was warranted.

The American Psychiatric Association12 treatment guidelines published in 2002 recommended lithium plus an antipsychotic or valproate sodium as first-line therapy, with less severely ill patients receiving monotherapy with lithium, valproate, or an antipsychotic agent. Second-line therapy included a combination of 2 first-line medications or an addition of an antipsychotic agent. Maintenance therapy included treatment with lithium and valproate, with alternatives of lamotrigine, carbamazepine, or oxcarbazepine. It was suggested that antipsychotics be discontinued unless they were required for persistent psychotic symptoms. The Texas Medication Algorithm13 published in 1998 did not recommend atypical antipsychotics in the acute phase until stage 2 and then as combination therapy. Maintenance therapy in this treatment guideline included an antimanic agent for all patients. The guideline also suggested that patients could gradually discontinue treatment 6 months after full remission. These guidelines may partially help explain whether the patients in the UnTx cohort were in full remission.

Among the treated cohorts, all subjects had 1 or more outpatient diagnoses of BPD and at least 1 BPD-related prescription, or they had an inpatient hospital diagnosis of BPD. In the UnTx cohort, 38.6% had a single outpatient diagnosis of BPD. Given the challenge of an accurate diagnosis of BPD14-16 and the lengthy time reported to diagnosis,14,15,17 it was believed that a single BPD diagnosis was sufficient.

The most recent edition of the Algorithm for Treatment of Bipolar I Disorder18 recommended ATYP as potential monotherapy in stage 1A (patients presenting with euphoric, irritable hypomanic, or manic symptoms or with mixed symptoms), including aripiprazole, olanzapine, quetiapine fumarate, risperidone, and ziprasidone hydrochloride. Combination treatment with ATYP was recommended in stage 2, with the suggestion that aripiprazole, olanzapine, quetiapine, risperidone, ziprasidone, and other drugs in this class be paired with valproate or lithium. For stages 3 and 4, the algorithm recommends the introduction of larger sets of medications, 3-drug combinations, and electroconvulsive therapy as treatment options. The recommendation of ATYP for maintenance after hypomania, mania, or mixed episodes has had limited study, and this class of medications was recommended more frequently at later stages than more tested treatment regimens such as lithium or valproate. Most patients in clinical practice receive combination therapy for maintenance, but the role of combination therapy versus monotherapy has not been well researched.18 As reflected by the changes noted in the most recent Algorithm for Treatment of Bipolar I Disorder, the small ATYP cohort in this study suggests that the use of such agents was being expanded, with further research ongoing at the time of data collection.19

The use of ATYP in the small ATYP cohort resulted in large decreases in total medical costs, a large portion of which seems to be secondary to the decrease in mental health-related costs. This cohort also had a lower likelihood of ED visits and inpatient admissions that were mental health related, findings that were also seen for the patients in the BOTH cohort. Because the largest contributor to direct medical costs has been found to be inpatient hospitalization,20 these results are promising in suggesting improvement of this costly component of care of patients with BPD. Further research into the types of variables that were found to foster this improvement should be examined, because these results can be used to identify the types of patients that most benefit from the use of ATYP.

The range of total medical costs for patients with BPD during the 6-month posttreatment period was $2385 to $4187 (or $4770-$8374 per year). These values were consistent with the overall annual medical costs of $6250 calculated by Bryant-Comstock et al.3 These authors found that mental health-related care accounted for 22% of total healthcare costs.3 A similar percentage was found during the 6-month posttreatment period for the ATYP and OTHR cohorts in this study. As noted, patients in the BOTH cohort had higher mental health-related medical costs, which may be explained by factors such as disease severity.

On the other hand, the large value for nonmental healthcare costs observed in the UnTx cohort is not easily explained and requires further analysis. Patients with BPD have been found to have increased costs for nonmental healthcare compared with individuals without BPD.21 Further research is required to determine if there is a correlation between the fact that these patients were not taking BPD medications and the underlying nonmental health-related medical conditions. In particular, analysis of mental well-being as it relates to diagnosis and treatment of the patient as a whole, as well as its effect on nonmental health-related conditions, is needed. Furthermore, the role of medication nonadherence, one of the most common reasons for relapse of this condition, needs further investigation.22

ICD-9-CM

This study has multiple limitations, with the first and foremost being the retrospective database analysis design. The findings are also limited to applicability among managed care and other indemnity plans. Another potential limitation of this research is that entry into the study was restricted to individuals with diagnosis codes of BPD, and the findings may not be representative of persons with BPD who are not diagnosed, who are misdiagnosed, or who do not have a diagnosis in their medical records.23 In addition, the cohorts studied were categorized by psychotropic medication class treatment, with identification by prescription claims. Assignment of these cohorts may introduce biases to the study secondary to current managed care prescription policies that include components of formulary design, step therapy, higher copayments, or prior authorization.

Despite the limitations, the study allows for interesting insights into the treatment of BPD in managed care settings. It also suggests areas of research that can be pursued to further pinpoint areas to improve treatment of patients with BPD, with one of the most important research needs being the development of methods to determine which groups of patients with BPD respond best to which types of medication treatment.

Acknowledgments

We recognize Harold H. Gardner, MD, president of the Human Capital Management Services Group (HCMS), Suzanne Novak, MD, PhD, University of Texas at Austin, James E. Smeeding, RPh, MBA, president, the JeSTARx Group, and the Center for Pharmacoeconomic Studies at the University of Texas at Austin for contributions to this research.

Author Affiliations:

From Retrospective Analysis, the JeSTARx Group, Newfoundland, NJ (RAB); Analysis and Research, HCMS, Cheyenne, Wyo (NLK); and Health Economics/Outcomes Research, AstraZeneca Pharmaceuticals, Wilmington, Del (KR). This research was supported by AstraZeneca Pharmaceuticals.

Correspondence Author:

Richard A. Brook, MBA, Retrospective Analysis, the JeSTARx Group, 18 Hirth Dr, Newfoundland, NJ 07435-1710. E-mail: rbrook@jestarx.com.