The nature of clusters of multimorbid patients in the UK: a latent class analysis.

Objectives: To identify, describe and validate clusters of patients based on their multimorbidity, to allow better design of health services and highlight groups that may require tailored interventions. Design: Retrospective cohort study. Patients with multimorbidity were stratified by four age strata and clustered using latent class analysis. Associations between multimorbidity clusters, demographics and outcomes were quantified using generalised linear models. Setting: 382 general practices in England contributed primary care health record data to the Clinical Practice Research Datalink (CPRD). Participants: All multimorbid adults (18 years old or more, with two or more long-term conditions) whose diagnoses are defined in 2012 (N=113,211), from a random sample of CPRD-GOLD (N=391,669). Cluster identification used a random set of 80% of the multimorbid patients (N=90,571), with consistency of results checked in the remaining 20% of multimorbid patients (N=22,640). Main outcome measures: NHS service utilisation was measured by three variables: primary care consultations, hospitalisations and repeat prescriptions in one year after January 2012. All-cause mortality was recorded within two and five years. Results: Clinically distinct and meaningful clusters were identified using robust latent class analysis for 38 long-term conditions within each age strata. Associated patient profiles and outcomes were derived. 9Physical-mental health co-morbidity9 and 9Respiratory disease and multimorbidity9 clusters were common across all age strata. In under 65 year olds, 9Substance misuse and mental illness co-morbidity9 cluster (