Cluster Analysis and Its Utility in Healthcare Claims Data Analysis: Results of a Retrospective Observational Study of Patients With Metastatic Urothelial Cancer (mUC) in Germany

OBJECTIVES: Cluster analysis (CA) is a frequently used applied statistical technique that helps to reveal hidden structures and homogenous clusters or subgroups within large datasets. The purpose of this study was to identify clinically relevant segments of patients with mUC that did not receive systemic anticancer treatment using CA.

METHODS: This nonexperimental, retrospective CA used 2 statutory health insurance claims databases (2013-2020; ≈8 million insured) to identify adult patients with an incident mUC diagnosis from 2015-2019 in Germany. Patients with other malignant tumors were excluded. Patients were observed for ≥12 months after incident mUC diagnosis (index) or until death. Patient characteristics were analyzed descriptively. A CA application was used to identify patient characteristics and healthcare system factors related to nontreatment.

RESULTS: Of 3,226 patients with mUC, 70.8% were male, mean (SD) age was 73.8 (10.8) years, mean (SD) Charlson Comorbidity Index (CCI) score was 6.3 (3.8), and mean (SD) Elixhauser Comorbidity Index score was 17.6 (11.4). A total of 1,892 patients (58.6%) with mUC did not receive systemic treatment within 12 months of diagnosis. After identifying outliers, CA indicated that a 2-cluster solution was the most appropriate option for both databases. Clusters with the highest proportion of untreated patients also had the largest number of patients who were older, had higher CCI scores, required home care, and were more likely to receive their index diagnosis in smaller hospitals.

CONCLUSIONS: This analysis of healthcare claims data using clustering methodologies identified meaningful subgroups of patients with mUC that did not receive systemic treatment. The results corroborate findings from previous studies where older patients with mUC who had several comorbidities were most likely to remain untreated. By identifying subgroups with increased risk of nontreatment such as those diagnosed in smaller hospitals, healthcare decision-makers may design personalized intervention programs according to their unique needs.