(#43) Use Of Machine Learning to Identify Variations in Clinical Characteristics, Healthcare Resource Utilization, and Treatment Adherence Among Patients With Schizophrenia Initiating Oral Olanzapine Treatment

Abstract: Oral antipsychotic treatments are associated with poor adherence and high disease/healthcare resource utilization (HCRU) burden in patients with schizophrenia. Long-acting injectable (LAI) antipsychotics may improve adherence and reduce HCRU. This analysis aimed to identify patient subtypes with distinct clinical profiles and unmet treatment needs that respond differently to oral olanzapine. This retrospective study using the MarketScan Multi-State Medicaid Database included adults with schizophrenia (ICD-10:F20) initiating oral olanzapine (2018ñ2022). Machine-learning approaches included automated feature engineering, k-prototype clustering (subgroup identification), and XGBoost (risk factor selection). Oral olanzapine adherence (proportion of days covered ?80% over 12 months) and HCRU were assessed for subgroup validation. Among 3668 patients, three subtypes were identified. Key differentiators were substance abuse, mental/psychiatric conditions, HCRU, medication usage, and aging/chronic conditions. Subtype 1 (53%) comprised younger patients (mean age: 36) with fewer psychiatric and chronic conditions and the lowest HCRU but poor adherence rates (19%). Subtype 2 (20%) included older patients (mean age: 53) with high chronic disease burden, complex medication use, and higher adherence rates (30%). Subtype 3 (27%) were the youngest (mean age: 35) and had severe psychiatric conditions, substance use, high HCRU, and the lowest adherence rate (15%) but better HCRU vs other subtypes (reduction in schizophrenia-related ER visits: 8% vs 1%; reduction in schizophrenia-related hospitalizations: 25% vs ?15%). This study underscores the importance of understanding complex clinical profiles in patients with schizophrenia to address diverse unmet needs and optimize treatment selections such as LAI intervention to maximize treatment benefits and improve outcomes.Short Description: This real-world data analysis used machine learning to identify subtypes of patients with schizophrenia initiating oral olanzapine. Three patient subtypes were identified with distinct clinical profiles, including substance abuse, age, psychiatric/chronic conditions, and medication use. Treatment adherence and HCRU were different across subtypes. Findings highlight the current unmet need and potential benefits from olanzapine LAI. These insights may support clinicians in tailoring interventions to improve outcomes for patients with schizophrenia.Name of Sponsoring Organization(s): Teva Branded Pharmaceutical Products R&D LLC