Comparative analysis of methods for identifying multimorbidity patterns: A study of 'real-world' data

Albert Roso-Llorach, Concepción Violán, Quintí Foguet-Boreu, Teresa Rodriguez-Blanco, Mariona Pons-Vigués, Enriqueta Pujol-Ribera, Jose Maria Valderas

    Research output: Contribution to journalArticleResearchpeer-review

    20 Citations (Scopus)

    Abstract

    © 2018 Article author(s). Objective The aim was to compare multimorbidity patterns identified with the two most commonly used methods: Hierarchical cluster analysis (HCA) and exploratory factor analysis (EFA) in a large primary care database. Specific objectives were: (1) to determine whether choice of method affects the composition of these patterns and (2) to consider the potential application of each method in the clinical setting. Design Cross-sectional study. Diagnoses were based on the 263 corresponding blocks of the International Classification of Diseases version 10. Multimorbidity patterns were identified using HCA and EFA. Analysis was stratified by sex, and results compared for each method. Setting and participants Electronic health records for 408 994 patients with multimorbidity aged 45-64 years in 274 primary health care teams from 2010 in Catalonia, Spain. Results HCA identified 53 clusters for women, with just 12 clusters including at least 2 diagnoses, and 15 clusters for men, all of them including at least two diagnoses. EFA showed 9 factors for women and 10 factors for men. We observed differences by sex and method of analysis, although some patterns were consistent. Three combinations of diseases were observed consistently across sex groups and across both methods: Hypertension and obesity, spondylopathies and deforming dorsopathies, and dermatitis eczema and mycosis. Conclusions This study showed that multimorbidity patterns vary depending on the method of analysis used (HCA vs EFA) and provided new evidence about the known limitations of attempts to compare multimorbidity patterns in real-world data studies. We found that EFA was useful in describing comorbidity relationships and HCA could be useful for in-depth study of multimorbidity. Our results suggest possible applications for each of these methods in clinical and research settings, and add information about some aspects that must be considered in standardisation of future studies: Spectrum of diseases, data usage and methods of analysis.
    Original languageEnglish
    Article numbere018986
    JournalBMJ Open
    Volume8
    Issue number3
    DOIs
    Publication statusPublished - 1 Mar 2018

    Keywords

    • cluster analysis
    • diseases
    • electronic health records
    • factor analysis
    • multimorbidity
    • primary health care

    Fingerprint

    Dive into the research topics of 'Comparative analysis of methods for identifying multimorbidity patterns: A study of 'real-world' data'. Together they form a unique fingerprint.

    Cite this