Data-driven patient segmentation using K-means clustering

Machine learning continues to forge the future of decision making in a broad diversity of domains including healthcare. Data-driven methods are increasingly geared towards leveraging evidence-based insights from large volumes of patient data. In this context, this paper embraces a mere data-driven approach for the segmentation of patients with application to hip fracture care in Ireland. Using K-Means clustering, elderly patients are grouped based on the similarity of age, length of stay (LOS) and elapsed time to surgery. We utilise a dataset retrieved from the Irish Hip fracture Database (IHFD) covering the period of two years (2013--2014). Our results suggest the presence of three coherent clusters of patients. Through cluster analysis, possible correlations are explored in relation to patient characteristics, care-related factors, and patient outcomes. For instance, the study inspects the potential impact of time to surgery on patient outcomes (e.g. LOS) within the discovered clusters of patients. Furthermore, the clusters are visually interpreted in a demographic context with respect to the structure of the healthcare system in Ireland. Broadly, the study is claimed to serve useful purposes for healthcare executives in Ireland for developing more patient-centred care strategies.