Exploring Schizophrenia Treatment Pattern with AI: TAK (TREATMENT SEQUENCES ANALYSIS THROUGH K-CLUSTERING) Application in a Japan Retrospective Claims Database
Author(s)
ABSTRACT WITHDRAWN
Presentation Documents
OBJECTIVES:
Schizophrenia is affecting 0.56% of the Japanese population. The estimated annual burden of disease exceeds 3.5 million yen per patient (approximately US$30,000). Prescribing patterns are complex and characterized by treatment switching, discontinuation, relapse and poor adherence to antipsychotic medication. This study aims to improve the knowledge in real word of schizophrenia treatment patterns based on Japan retrospective claims.METHODS:
Patients with at least one antipsychotic drug were selected from JMDC claims, an employment-based administrative database containing the medical and pharmacy claims from ten different payers between 2009 and 2013. A TAK method was applied in 3 steps. A vector analysis of each patient and the treatment pattern to cluster patient treatment based on a similarity score was applied. The second step was to rank the treatment pattern based on hierarchical Agglomerative Clustering technique. Finally, an image processing was applied to eliminate noise and improve interpretability.RESULTS:
Among 12,066 patients, the mean (SD) age was 35 (14) years and mean(SD) follow-up duration was 1.9 (1.2) years. Among 24 identified treatments, Sulpiride was the most frequently first line treatment (50.8% of patients). Sulpiride patients profile presented a low persistence and high discontinuation rate. In subsequent lines, Sulpiride was prescribed in combination with Olanzapine. Risperidone was the second most frequent first line treatment (9.6% of patients). Patients initiating Risperidone presented a low persistence characterized by a short time to discontinuation. In subsequent lines, switching from Risperidone to Levomepromazine Maleate was observed. Overall, patients presented high discontinuation and low persistence which confirms poor psychiatric response along with worsening symptoms in most of real-world cases. CONCLUSION: The major advantage of this approach is to provide information on the temporal correlations and reveal hidden patterns in data. This AI method support a better RWE understanding of antipsychotic treatment sequences in schizophrenic patients.Conference/Value in Health Info
2022-05, ISPOR 2022, Washington, DC, USA
Value in Health, Volume 25, Issue 6, S1 (June 2022)
Code
RWD102
Topic
Clinical Outcomes, Methodological & Statistical Research
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics, Clinical Outcomes Assessment
Disease
Mental Health