Latent Variable Analysis of Antipsychotic Adherence among South Carolina Medicaid Beneficiaries with Schizophrenia

Abstract: Schizophrenia, a chronic mental illness, requires long-term treatment with antipsychotics, where adherence is crucial for favorable clinical and economic outcomes. Adherence, however, is a multifaceted concept, not fully captured by a single measure. Thus, adherence can be considered a latent variable that cannot be directly observed but is inferred from other measured variables such as the proportion of days covered (PDC), medication possession ratio (MPR), persistence, and gaps in medication refill. To uncover the complexity of medication adherence patterns, this study applied a latent variable approach to assess medication adherence among 3,994 schizophrenia patients aged 18-65 treated with antipsychotic medications, using South Carolina Medicaid data from 2012-2019. Four distinct adherence groups were identified: "best adherence” (consistent use, 58%), "intermittent adherence” (sporadic use, 17%), "early drop-off" (brief utilization then discontinuation, 9%), and "worst adherence” (minimum/no use, 16%). The patient’s membership in an adherence class was determined by the highest conditional probability using latent profile analysis. Further analysis using a multinomial logistic regression model assessed the impact of demographic and treatment-related factors on these adherence groups. Key findings include a higher likelihood of nonadherence among patients on oral antipsychotics compared to long-acting injectables, with notable disparities based on age, race, and treatment history. Younger patients and black patients exhibited higher rates of “intermittent adherence”, while new users showed a significantly higher risk of being in the "worst adherence" category. These insights may help customize interventions based on patients’ adherence membership which in turn can help improve patient outcomes.Short Description: This study employed a latent variable framework to assess medication adherence among South Carolina Medicaid beneficiaries with schizophrenia and identified four categories: "best"(consistent use), "intermittent"(sporadic use), "early drop-off"(brief engagement then discontinuation), and "worst"(minimal/no use). Adherence classifications were determined through latent variable analysis and multinomial logistic regression was used to assess how demographic and treatment factors influence adherence patterns. The study highlights the complexity of adherence, underscoring the need for personalized interventions to improve treatment outcomes.Name of Sponsoring Organization(s): Janssen Scientific Affairs, LLC