Pharmacy and Therapeutics - January 2008 - (Page 16) Impact of Bipolar Disorder on the Family family members without a diagnosis of a serious mental illness (“control families”). We derived the data used for this study from the MarketScan Commercial Claims and Encounters Database, which reflects the combined use of health care services of more than two million privately insured individuals in the U.S. who were covered under fee-for-service, fully capitated, and partially capitated health plans. The database is constructed through employer-supplied records and does not contain information that allows for personal identification of individuals. We identified individuals with bipolar disorder through diagnostic codes for bipolar or manic disorder (296.4x–296.8x) from the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). Individuals with medical claims for these diagnoses from January 1998 through December 2002 were included in the analysis. Family members of the individuals with bipolar disorder were identified through eligibility codes that distinguished employees from dependents. We determined family position from eligibility files that included demographic information and the relationship to the policy beneficiary. The “bipolar families” were matched to control families without diagnosis codes for bipolar disorder or other serious mental illnesses (ICD-9-CM codes 290–319) in a 1:3 ratio based entirely on family characteristics: the number of family members, the type of insurance plan, and geographic region. We chose this matching ratio because taking more than one control per case would have increased the statistical confidence of the results. However, this effect declines considerably at a ratio of over five controls per case. In this situation, with families being matched, going beyond three controls per case would have been difficult and time-consuming. This ratio was also used in a similar study comparing family costs of migraine.15 We created a file containing inpatient, outpatient, and pharmaceutical claims for all individuals in both groups. Services and costs were considered to be related to bipolar disorder if an outpatient or emergency visit or a hospital stay was associated with a primary diagnosis code of 296.4x–296.8x; only prescriptions for lithium, valproate, lamotrigine (Lamictal, GlaxoSmithKline), and antipsychotic agents were considered bipolar disorder–related. We calculated health care use as the average number of inpatient and outpatient visits as well as the number of prescriptions per family over the five years of data. We determined the cost of health care services from the total payment received by the provider, including copayments and deductibles. The total costs included all “carve-out” claims such as behavioral health. We calculated the cost of care as an average over the five-year study period. The primary analyses comparing health care utilization and costs between families included the family member with bipolar disorder. We calculated per-member costs by dividing the total cost for the specific group by the total number of members in that group. Because we had concerns about extrapolating costs for services provided under capitated insurance plans, we included only families in health plans that reported the amount paid per service in calculating the costs of health care; by contrast, we included all families in calculating the utilization of health care resources. Although it is possible to assign a cost to a service provided under a capitated health plan, it was felt that the calculation could be biased. Therefore, we calculated resource utilization for everyone as an alternate way to assess the difference in health care usage between study groups. We used Major Diagnosis Categories (MDCs), a classification system developed by the Centers for Medicare and Medicaid Services, to group diagnosis codes into 26 major categories as proxies of medical and mental health comorbidities. We used t-tests to compare utilization of resources between groups. Wilcoxon rank sum tests were used to compare health care costs because of the non-normal distribution of cost data. A P value of 0.05 was considered statistically significant, and the Bonferroni method was used to adjust P values for multiple comparisons. We used regression analysis to test the impact of other demographic and health care variables on total health expenditures. We transformed the expenditures into logs before estimating using an ordinary least-squares model, as is typically done to normalize highly skewed distributions. Independent variables considered for the regression model were family size, the total number of MDCs, the presence of an individual with bipolar disorder in the family, area of residence, and type of insurance plan. RESULTS The MarketScan Database contained 1,868,968 families. From these, we identified 43,448 families, including at least one member with bipolar disorder, that were matched to 122,769 families without a serious mental illness (Table 1). Preferred provider organization (PPO) and point-of-service (POS) plans were the most prevalent types of health insurance; 50% of families (21,767 bipolar families and 61,139 control families) had this type of coverage. Most families resided in the southern and northern central regions of the U.S. The trends in insurance type and geographic region were similar for those individuals in the master MarketScan Database, with 39% and 28% residing in the southern and north central regions respectively; 63% were covered by PPO and POS health plans. In this sample, individuals with bipolar disorder were primarily between the ages of 18 and 65 years (37,204 family members, 82%) and female (27,322 family members, 60%). The mean number of MDCs for this population was 7.2. The most common bipolar diagnosis code for these individuals (10,881 members, 24%) was episodic affective disorder (ICD-9 code 296.8), followed closely by bipolar affective disorder–mixed (ICD-9 code 296.6) and bipolar affective disorder–unspecified (ICD-9 code 296.7), each with slightly more than 19% of diagnoses (8,779 and 8,787 family members, respectively). More comorbid conditions were observed among the bipolar families, according to the mean number of MDCs (Table 1). The distribution of diagnoses in each of the 26 MDCs (Table 2) was similar between the groups except for a lower percentage of pregnancy and newborn diagnoses, and a greater percentage of diagnoses of alcohol or drug use in the bipolar families. The category of “Injuries, Poison, and Toxic Effect of Drugs” also tended to be more common in the bipolar continued on page 23 16 P&T® • January 2008 • Vol. 33 No. 1
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.