K-means clustering to identify high risk of early revisits in patients with drug-related problems attending the emergency department.

Abstract

Objective: Drug-related problems (DRPs) are a frequent reason for visits to the emergency department (ED). However, data about the characteristics associated with early revisits are limited. We aimed to identify clinical phenotype clusters of patients admitted to emergency rooms due DRPs to identify those patients with the highest risk of new visits.

Methods: We included consecutive patients admitted to EDs due DRPs (February 2021 to December 2022), including DRP admissions in 2023 as validation cohort. We employed K-means clustering to group patients according to adjusted morbidity groups (GMA), age, and number of drugs at admission. To determine the optimal number of cluster centres, we used the elbow method. The impact of 30-day revisits in each cluster was assessed.

Results: 1611 patients (mean (SD) age 75.0 (15.1) years) were included. We identified six clusters, with 30-day revisits rates ranging from 14.8% to 24.5%. The main groups of drugs implicated in the DRP episodes were diuretics (190 patients; 11.8%). The most common DRP diagnoses were constipation (191; 11.9%) and gastrointestinal bleeding (158; 9.8%). Six clusters of patients were identified. Significantly higher 30-day revisits in patients identified in cluster 4 (24.5% vs 17.5%; p=0.007). The highest revisit rate was observed in the cluster including patients with a higher number of drugs and GMA status.

Conclusions: Patients admitted to the ED due DRPs exhibit varying revisit rates across different clinical phenotypes. K-means clustering aids in identifying patients who derive the greatest rates of readmission, and is a useful tool to prioritise interventions in these units.