A systematic review and meta-analysis of disease clusters in multimorbidity
Jennifer K. Ferris, Lean Fiedeldey, Boah Kim, Felicity Clemens, Michael A. Irvine, Sogol Haji Hosseini, Kate Smolina, Andrew Wister
Abstract
There is a growing body of research on disease clusters in multimorbidity. Here we systematically review clustering methodologies and perform a meta-analysis of disease cluster stability across the literature, searching Medline and EMBASE from inception to June 5th, 2024, for studies of disease clusters in multimorbidity (including network approaches). Here we include 79 articles. 30% of studies had high risk of bias. Hierarchical cluster analysis was the most used clustering methodology (25% of analyses), followed by latent class analysis (20%) and K-center clustering (15%). Network-based approaches were used in 19% of studies. We perform a meta-analysis of 1226 disease clusters across 73 studies. Strong relationships emerged between neurological, autoimmune, musculoskeletal, and cardiovascular diseases. We identify six meta-analytic disease clusters with moderate stability (Jaccard index ≥0.51), these largely featured cardiometabolic conditions. No disease clusters had high stability (Jaccard ≥0.75) and very few accounted for disease temporality. Multimorbidity disease clustering research is heterogeneous regarding disease definitions, the number of diseases included, and clustering methodologies. Despite this heterogeneity, moderately consistent disease clusters emerge. We provide suggestions to improve the performance and reporting of multimorbidity clustering research. There is a growing body of research on disease clusters in multimorbidity, necessitating a systematic review and meta-analysis of methods and findings. Here, the authors show the range of methods applied, and identify six disease clusters with moderate stability across the multimorbidity literature.