Multimorbidity in a selected cohort compared to a representative sample: Does selection bias influence outcomes?

Context: UK Biobank is increasingly used to study causes, associations, and implications of multimorbidity. However, criticised for lack representativeness ‘healthy volunteer bias’. Selection bias can lead spurious or biased estimates associations between exposures outcomes. Objectives: To compare association multimorbidity adverse health outcomes in a nationally representative sample. Design: Cohorts identified from linked routine healthcare data the Secure Anonymised Information Linkage (SAIL) databank. Setting: Community. Participants: participants (n=211,597, age 40-70) with primary care sample source (n=852,055, 40-70). Main outcome measures: Multimorbidity (n=40 long-term conditions [LTCs]) was Read codes quantified using simple count weighted score. Individual LTCs LTC combinations were also assessed. Associations all-cause mortality, unscheduled hospitalisation, major cardiovascular events (MACE) assessed Weibull Poisson models adjusted age, sex, socioeconomic status. Results: less common than SAIL. This difference attenuated, but persisted, after standardising by sex The effect increasing on MACE similar SAIL at counts ≤3, however above this level underestimated risk associated Absolute hospitalisation MACE, all levels multimorbidity, lower (adjusting status). Both cohorts produced hazard ratios some (e.g. hypertension coronary heart disease) others alcohol problems mental conditions). Similarly cardiovascular, respiratory conditions), including pain conditions. Conclusions: accurately ≤3. ≥4 magnitude are likely be conservative.