How predictive of future healthcare utilisation and mortality is data-driven population segmentation based on healthcare utilisation and chronic condition comorbidity?

BACKGROUND: In recent years data-driven population segmentation using cluster analyses of mainly health care utilisation data has been used as a proxy of future health care need. Chronic conditions patterns tended to be examined after segmentation but may be useful as a segmentation variable which, in combination with utilisation could indicate severity. These could further be of practical use to target specific clinical groups including for prevention. This study aimed to assess the ability of data-driven segmentation based on health care utilisation and comorbidities to predict future outcomes: Emergency admission, A&E attendance, GP practice contacts, and mortality. METHODS: We analysed record-linked data for 412,997 patients registered with GP practices in 2018-19 in Cwm Taf Morgannwg University Health Board (CTM UHB) area within the Secure Anonymised Information Linkage (SAIL) Databank. We created 10 segments using k-means clustering based on utilisation (GP practice contacts, prescriptions, emergency and elective admissions, A&E and outpatients) and chronic condition counts for 2018 using different variable compositions to denote need. We assessed the characteristics of the segments. We employed a train/test scheme (80% training set) to compare logistic regression model predictions with observed outcomes on follow-up in 2019. We assessed the area under the ROC curve (AUC) for models with demographic variables, with and without the segments, as well as between segmentation implementations (with/without comorbidity and primary care data). RESULTS: Adding the segments to the model with demographic covariates improved the prediction for all outcomes. For emergency admissions this increased discrimination from AUC 0.65 (CI 0.64-0.65) to 0.73 (CI 0.73-0.74). Models with the segments only performed nearly as well as the full models. Excluding comorbidity showed reduced predictive ability for mortality (similar otherwise) but most pronounced reduction when excluding all primary care variables. CONCLUSIONS: This shows that the segments have satisfactory predictive ability, even for varied outcomes and a broad range of events and conditions used in the segmentation. It suggests that the segments can be a useful tool in helping to identify specific groups of need to target with anticipatory care. Identification may be refined with selected diagnoses or more specialised tools such as risk stratification.

Can cluster analyses of linked healthcare data identify unique population segments in a general practice-registered population?

BACKGROUND: Population segmentation is useful for understanding the health needs of populations. Expert-driven segmentation is a traditional approach which involves subjective decisions on how to segment data, with no agreed best practice. The limitations of this approach are theoretically overcome by more data-driven approaches such as utilisation-based cluster analysis. Previous explorations of using utilisation-based cluster analysis for segmentation have demonstrated feasibility but were limited in potential usefulness for local service planning. This study explores the potential for practical application of using utilisation-based cluster analyses to segment a local General Practice-registered population in the South Wales Valleys. METHODS: Primary and secondary care datasets were linked to create a database of 79,607 patients including socio-demographic variables, morbidities, care utilisation, cost and risk factor information. We undertook utilisation-based cluster analysis, using k-means methodology to group the population into segments with distinct healthcare utilisation patterns based on seven utilisation variables: elective inpatient admissions, non-elective inpatient admissions, outpatient first & follow-up attendances, Emergency Department visits, GP practice visits and prescriptions. We analysed segments post-hoc to understand their morbidity, risk and demographic profiles. RESULTS: Ten population segments were identified which had distinct profiles of healthcare use, morbidity, demographic characteristics and risk attributes. Although half of the study population were in segments characterised as 'low need' populations, there was heterogeneity in this group with respect to variables relevant to service planning - e.g. settings in which care was mostly consumed. Significant and complex healthcare need was a feature across age groups and was driven more by deprivation and behavioural risk factors than by age and functional limitation. CONCLUSIONS: This analysis shows that utilisation-based cluster analysis of linked primary and secondary healthcare use data for a local GP-registered population can segment the population into distinct groups with unique health and care needs, providing useful intelligence to inform local population health service planning and care delivery. This segmentation approach can offer a detailed understanding of the health and care priorities of population groups, potentially supporting the integration of health and care, reducing fragmentation of healthcare and reducing healthcare costs in the population.