The study was conducted using de-identified medical and prescription drug claims from Medicaid databases. The databases included claims of Medicaid beneficiaries from six US states (Iowa, Kansas, Mississippi, Missouri, New Jersey, Wisconsin) from July 2009 to March 2015. Detailed information on patient characteristics, physician visits, hospitalizations, long-term care services, prescription drugs, and other services reimbursed by Medicaid was available. Information was also available for Medicaid payments, out-of-pocket costs, and Medicare copayment and deductibles. Available cost data reflected the Medicaid payers’ perspective, while prescription drug claims were prior to any discounts or rebates. All data were de-identified and in compliance with the Health Insurance Portability and Accountability Act.
Study Design and Patient Selection
A retrospective cohort study design was used to compare patients with schizophrenia and CMCs, defined as patients with diabetes and/or CVD (including hyperlipidemia, hypertension, and other CVD-related conditions; see Table A of the Electronic Supplementary Material for a full list of the codes), who were initiated on either PP1M or an OAA. Outcomes were evaluated during the fixed 12-month observation period including and following the index date (date of AP initiation), while baseline demographics and clinical characteristics were evaluated during the 6-month period prior to the index date.
To be included in the study, patients were required to meet the following criteria: (1) at least two pharmacy and/or medical claims for PP1M or the same OAA (i.e., two claims for one of the following agents: aripiprazole, asenapine, iloperidone, lurasidone, olanzapine, paliperidone, quetiapine, risperidone, or ziprasidone) within 90 days in the year 2010 or after (the first of the two claims was defined as the index date), (2) no claim for the index agent (PP1M or OAA) in the 6-month period before the index date (baseline period), (3) at least two claims with a diagnosis for schizophrenia during the study period, with at least one claim occurring during the baseline period, (4) at least two claims with a diagnoses for diabetes or at least two claims with a diagnosis of CVD during the study period, with at least one of the claims occurring during the baseline period, (5) at least 6 months of continuous enrollment before the index date, and at least 12 months of follow-up after the index date (observation period), (6) 18–64 years of age as of the index date, and (7) no claim for long-acting injectable therapies during the baseline period (i.e., haloperidol, fluphenazine, aripiprazole, olanzapine, paliperidone, or risperidone long-acting injectable). Treatment groups were defined by the treatment initiated on the index date (i.e., PP1M or specific OAA) and data were analyzed using an intent-to-treat approach.
Outcome measures included persistence, healthcare costs, and HRU measured over the 12-month observation period for each patient and compared between PP1M and OAA treatment groups. Persistence to the index AP at 12 months was defined as not having any gap greater than 30, 60, or 90 days between days with available medication during the 12 months following treatment initiation.
All-cause, schizophrenia-related, and CMC-related costs and HRU measures were reported. Total healthcare costs were calculated and grouped into pharmacy and medical costs. Both medical costs and HRU were stratified into the following categories: outpatient visits, emergency room visits, inpatient visits, skilled nursing and long-term care admissions, mental health institute admissions, 1-day mental health institute outpatient visits, home care services, and other services. Schizophrenia and CMC-related medical costs and HRU were identified as the subset of claims with International Classification of Diseases, 9th revision, Clinical Modification diagnostic codes for schizophrenia, or for diabetes or CVD among patients with each respective comorbidity. Schizophrenia-related pharmacy costs were defined as the costs for AP pharmacy medication claims, while CMC-related pharmacy costs were defined as the costs for diabetes-related and CVD-related medication claims, among patients with each respective comorbidity.
Descriptive statistics for baseline characteristics included means, medians, standard deviations for continuous variables, and frequencies and proportions for categorical variables. Statistical comparisons between cohorts used standardized differences (Stdiffs), where a Stdiff ≥ 10% was considered clinically meaningful [29, 30]. Persistence to the index medication (i.e., PP1M or index OAA) at 12 months post-index date was compared using Chi-squared tests.
Multivariate adjustment (using age, sex, race, state, region, year of index date, presence of dual insurance eligibility, Quan–Charlson Comorbidity Index, comorbidities (CVD, diabetes), baseline use of atypical APs, presence of AP polypharmacy, total pharmacy costs, total medical costs, the number of mental health institute visits, the number of 1-day mental health institute visits, and the number of comorbidity-related inpatient visits) was conducted for cost and HRU outcome measures. Adjusted medical and pharmacy costs were compared using multivariate ordinary least-squares regressions and mean monthly cost differences (MMCDs), while adjusted HRU outcomes were compared using multivariate Poisson regressions and incidence rate ratios (IRRs). Confidence intervals and p values for both cost and HRU outcomes were obtained from a non-parametric bootstrap procedure (n = 499 replications). All cost outcomes were inflated to 2015 US dollars using the medical care component of the US Consumer Price Index.