Bipolar Medication Use and Adherence to Antiretroviral Therapy Among Patients With HIV-AIDS and Bipolar Disorder

This study utilized Medicaid Analytic Extract data sets from 2001 to 2004 from eight states, which contained data for 4,568,010 individuals who were enrolled in Medicaid for at least one month. The eight states were California, New York, New Jersey, Georgia, Ohio, Illinois, Texas, and Florida. Person- and claims-level information was used, including ICD-9-CM diagnosis codes, National Drug Codes, and pharmacy dispense dates for every medication for which a prescription was filled. ICD-9-CM codes used to identify bipolar disorder were 296.0, 296.1, and 296.4–296.8; for HIV they were 042–044; for substance use disorders they were 291, 292, and 303–305; and for schizophrenia the code was 295.

Using a person-month as the unit of analysis, we classified each person-month as covered or not covered by antiretroviral therapy and by medication for bipolar disorder. Medications for bipolar disorder included lithium, antipsychotics, and the anticonvulsants carbamazepine, oxcarbazepine, valproate, and lamotrigine. We excluded gabapentin and topiramate, which do not demonstrate efficacy in randomized clinical trials. Because most Medicaid prescriptions for antiretroviral therapy and bipolar medications were for 30 days, with only 2% of antiretroviral therapy and 3% of bipolar medication prescriptions exceeding 31 days, we classified each calendar month as covered if a prescription for the medication was filled in that month.

Participants were aged 18 to 65 years and during 2001–2004 had at least one inpatient or two outpatient claims for bipolar disorder, one inpatient or outpatient claim for HIV, plus at least one pharmacy claim for antiretroviral therapy and one for bipolar medications. We analyzed treatment episodes, defined as starting with the first month during which prescriptions for both bipolar and antiretroviral medications were filled. Only patients who remained enrolled in Medicaid for at least one year after this start date were included. Patients who also had a diagnosis of schizophrenia were excluded. Finally, we included only fee-for-service beneficiaries who were continuously enrolled in Medicaid and who did not receive long-term care or have nursing home days. Covariates included the state and the beneficiary's gender, age, race-ethnicity, and diagnosed substance use disorders.

Using multivariate logistic regression, we treated a refill for antiretroviral therapy in a given month as a dependent variable and treated a refill for bipolar medication in the prior month as a predictor variable. We chose this “lagged” strategy because same-month correlations between bipolar medication refills and antiretroviral refills would have included, for example, simultaneous refills occurring in a single pharmacy visit and would therefore have been less reflective of a causal direction running from bipolar treatment to antiretroviral adherence. Independent variables included patient characteristics, as well as the interaction of each covariate with the indicator of lagged bipolar medication possession, in order to delineate whether the effect of bipolar medication use varied by patient characteristics. Analyses took into account statistical dependence of repeated observations within individuals by using the cluster option within the logistic regression procedure in Stata, version 9.2. We utilized SAS, version 9.2, for data management. This research was approved by the Rutgers University Institutional Review Board.

We identified 10,971 beneficiaries with both bipolar disorder and HIV infection. The criteria regarding age, long-term and managed care, and one year of continuous enrollment after the start of the treatment episode reduced the sample to 2,631. Eliminating beneficiaries with schizophrenia reduced the sample to 1,687. The mean±SD observation period for the 1,687 persons was 32±4.6 months (N=53,971 person-months). Antiretroviral therapy medications were refilled in 72% of the months, and bipolar medications were refilled in 55%.

Confirming our primary hypothesis, we found that adherence to bipolar medication was associated with higher rates of adherence to antiretroviral therapy in the following month. As shown in Table 1, when bipolar medication was not filled in the prior month, the rate of antiretroviral fill in the subsequent month was 65%, compared with an antiretroviral fill rate of 78% when bipolar medication was filled in the previous month. The association remained significant after the analysis controlled for demographic characteristics and diagnoses of comorbid substance use disorders and their interactions. Adherence to antiretroviral therapy was higher for men than women. Fill rates for antiretroviral medications were highest for beneficiaries aged ≥50 and lowest for the youngest beneficiaries. As hypothesized, diagnosis of a comorbid substance use disorder was associated with a significantly lower antiretroviral adherence rate (adjusted odds ratio=.62).

We found a temporal association between receipt of bipolar medication and subsequent receipt of antiretroviral therapy in an analysis that controlled for several potentially confounding background demographic and clinical variables. Limitations should be noted. By requiring one year of observation, we may have missed less stable patients. Clinical status was observed by diagnoses, which lack valuable details (for example, a diagnosis of a substance use disorder does not necessarily indicate actual use during the observation period). It cannot be determined from prescription claims whether a patient ingests a medication. Experimental designs—particularly randomized controlled trials—are required to isolate a specific medication effect, although practical and ethical difficulties in this research area make randomized controlled trials challenging. Nevertheless, future investigations that include longitudinal collection of data on symptoms would allow researchers to assess the role of symptom control, mood stabilization, active substance use, and phase of disorder in supporting adherence to antiretroviral therapy for patients with bipolar illness and HIV-AIDS. The study was also limited to adult Medicaid beneficiaries in various phases of bipolar disorder who had already initiated bipolar medications and antiretroviral therapy. It is not clear whether similar results would have been obtained for other patient populations.

Our findings converge with those of previous research in suggesting that among persons with HIV and serious mental illness, adherence to a psychoactive medication regimen can improve adherence to antiretroviral therapy (8). Because of the central clinical importance of antiretroviral adherence, it seems appropriate that patients with bipolar disorder and HIV have individualized and ongoing assessments of their capacity to adhere to antiretroviral therapy. Adherence may fluctuate with clinical status. The findings underline the importance of psychopharmacological treatment in this population, both to benefit bipolar symptoms and to promote antiretroviral adherence.

This work was supported by grants R01-MH076206 and R01-MH058984 from the National Institute of Mental Health (NIMH), with additional support from the Agency for Healthcare Research and Quality (AHRQ) through cooperative agreement U18HS016097 for the Center for Research and Education on Mental Health Therapeutics at Rutgers, as part of AHRQ's Centers for Education and Research on Therapeutics Program. The content is solely the responsibility of the authors and does not necessarily reflect the official views of NIMH or AHRQ.

Dr. Olfson has received research funding from Eli Lilly and Company and Bristol-Myers Squibb. The other authors report no competing interests.