Multimorbidity patterns among patients hospitalized with prostate cancer in Portugal: a cluster analysis approach

Abstract

Abstract Multimorbidity is a common condition among cancer patients, resulting in increased complexity of care and risk of negative outcomes. This study aims to use clustering analysis to identify and characterize multimorbidity patterns among hospitalized prostate cancer patients in Portugal. This is a retrospective observational study using inpatient data from the Portuguese National Hospital Morbidity Database. Data on hospital admissions with a diagnosis of prostate cancer occurring in all public hospitals in mainland Portugal during 2011–2017 were considered. Partitioning clustering algorithms, namely K-modes, PAM (Partitioning Around Medoids), and hierarchical clustering, were used to identify multimorbidity clusters. Results obtained from the different clustering approaches were compared and assessed in terms of clinical relevance. A total of 10394 inpatient episodes were analyzed, with 6091 (58%) reporting multimorbidity. Similar clusters were obtained through the different partitioning approaches, with PAM presenting a higher stability and the best quality results in terms of average silhouette. The analysis of the 6 clusters obtained with PAM reveals groups with a pattern of hypertension co-occurring with diabetes, obesity, and arrhythmia, in addition to cancer itself. In this study, the validity of cluster analysis as an exploratory method for identifying clusters of multimorbid conditions among prostate cancer patients in Portugal was demonstrated, identifying relevant patterns of disease co-occurrence, with potential impact on treatment decisions and outcomes. The identified clusters revealed conditions that typically co-occur with prostate of cancer and that can be controlled throughout all phases of cancer survivorship by means of healthier behaviors aligned with integrated and coordinated care.