The heterogeneous health state profiles of high-risk healthcare utilizers and their longitudinal healthcare service utilization and mortality patterns
Patients at high risk of hospital readmissions are vulnerable during transitions of care. Segmentation of such heterogenous patients into relatively homogenous, distinct subgroups help facilitate healthcare resource planning and the development of effective interventions. We aimed to apply a data-driven latent class analysis (LCA) to segment a transitional care program population and validate its discriminative ability on 30- and 90-day longitudinal hospital readmission and mortality data.
We extracted data from the H2H program for all adult patients enrolled from June to November 2018. LCA was used to determine the number and characteristics of latent subgroups that best represented these data. The derived models were assessed based on model fit and clinical interpretability. 30- and 90-day hospital readmission and all-cause mortality from date of enrolment were compared across classes. Regression analysis was performed to assess predictive ability of class membership on hospital readmission and mortality.
We included 752 patients in total, with 68% of the population above 65 years old and 52.4% female. A 3-class best fit model was selected with the following classes: Class 1 “Frail, cognitively impaired and physically dependent”, Class 2, “Pre-frail, but physically independent” and Class 3 “Physically independent”. The three classes had significantly distinct demographics, medical and socioeconomic characteristics (p< 0.05) as well as 30- and 90-day all-cause readmission (p< 0.05) and mortality (p< 0.01). Class 1 patients were the most vulnerable, having the highest age-adjusted 90-day readmissions (OR= 2.2, 95% CI: 1.34-3.61, p= 0.0077), and 30- (OR= 7.06, 95% CI: 1.94-25.71, p= 0.0056) and 90-day mortality (OR= 11.57, 95% CI: 4.88-27.46, p< 0.0001).
We demonstrated the applicability of LCA in identifying 3 unique patient subgroups with distinct longitudinal hospital readmission patterns and mortality risk amongst high-risk patients. This provides important information for tailoring future post-acute care interventions. Class 1 in particular, necessitates intensive intervention in order to ameliorate its high healthcare burden.