For the millions of adults managing two or more chronic conditions simultaneously, research into how those conditions cluster together holds genuine clinical promise — pointing toward shared biological mechanisms, tailored treatment pathways, and more efficient care models. But a methodological fault line running through this entire field may be quietly undermining its findings before they even reach clinicians.
A Personal View published in The Lancet Healthy Longevity argues that multimorbidity clustering studies are frequently built on an unacknowledged ambiguity: researchers rarely make explicit whether they are clustering conditions — grouping diseases that tend to co-occur — or clustering individuals — grouping patients who share similar disease profiles. These are fundamentally different analytical objects, yet dominant techniques such as latent class analysis and hierarchical clustering can be applied to either without the distinction being clearly flagged. The result is a literature in which studies nominally addressing the same question are, in practice, answering different ones, with incompatible outputs being compared as if they were equivalent. Additional undisclosed choices — including how conditions are selected for inclusion, how comorbidity is encoded, and how cluster solutions are validated — compound the interpretive confusion.
This critique arrives at a critical moment. Multimorbidity research has expanded rapidly over the past decade, producing hundreds of clustering studies that now inform everything from healthcare resource allocation to genome-wide association analyses seeking genetic drivers of disease co-occurrence. If the foundational taxonomy of what is being clustered remains inconsistent, downstream conclusions about sociodemographic drivers, biological mechanisms, or intervention targets may rest on unstable ground. The limitation here is not computational but conceptual — a problem no larger dataset or more sophisticated algorithm will automatically resolve. For a field aspiring to reshape how complex patients are managed, achieving definitional consensus on the unit of analysis is arguably a prerequisite for cumulative scientific progress. This piece reads as a necessary, if overdue, methodological corrective.