Identification and Analysis of Patient Community Patterns with Advanced Clustering Model

This research is to identify patient community patterns, and enable hospitals to effectively classify the communities from which their patients come into meaningful segments, and provide a strategy for customizing healthcare services.

In this study, 320 lifestyle variables were selected and cleaned out of 3,000 plus raw variables from the EASI database by exploratory data analysis.  These variables mainly cover demographics, consumer expenditure, life stage characteristics, and health conditions at nationwide zip+4 level. To facilitate clustering and avoid the curse of dimensionality, 47 representative lifestyle composite scores were extracted by factor analysis. Based on these scores, 8 community segments were generated by k-means clustering. To further validate the segments, the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey data of 1,942 U.S. hospitals in 2012 were integrated to profile each segment. Finally we visualized these community segments on different GIS maps. Our results indicate that the identified community segments can significantly differentiate patient population in terms of community characteristics, demographics as well as patient experience scores. For example, we categorized a segment “wealthy families” as it could be profiled by educational white collar, busy lifestyle, and wealthiest households. In Massachusetts, the patient satisfaction scores of this segment are about 20-30% lower than those of the national benchmark, which may be a signal for hospitals to adjust and improve services to patients in this community segment. With such strategic solution, hospitals will better understand patients and predict their needs, and further provide them with high quality customized care.