Multimorbidity and adverse outcomes in UK Biobank: are findings biased by lack of representativeness?

Problem

UK Biobank is increasingly used to study the causes, associations, and implications of multimorbidity. However, UK Biobank is criticised for its lack of representativeness and ‘healthy volunteer bias’. Selection bias such as this can lead to spurious or biased estimates of associations between exposures and outcomes. Therefore, we aimed to compare the association between multimorbidity and adverse health outcomes in UK Biobank and an unselected, representative sample identified from SAIL databank.

Approach

Analysis of cohorts identified from linked routine healthcare data from the UK Biobank cohort and from the Secure Anonymised Information Linkage (SAIL) databank. Long-term conditions were identified at baseline using primary care data (Read codes for diagnoses and prescriptions) with linkage to hospital and mortality data. We included UK Biobank participants (n=211,597, aged 40-70) with linked primary care data and a similar aged sample from a nationally representative routine data source (SAIL) (n=852,055, age 40-70). Multimorbidity (n=40 LTCslong-term conditions [LTCs]) was quantified using a simple count and a weighted score. Individual LTCs and LTC combinations were also assessed. Associations with all-cause mortality, unscheduled hospitalisation, and major adverse cardiovascular events (MACE) were assessed using Weibull or Poisson models and adjusted for age, sex, and socioeconomic status.

Findings

Multimorbidity was less common in UK Biobank than in SAIL. The difference was attenuated, but persisted, after standardising by age, sex and socioeconomic status. The relative effect of increasing multimorbidity on all-cause mortality, unscheduled hospitalisation, and MACE was similar between UK Biobank and SAIL at smaller LTC counts (between 0 and 3 LTCs), however above this level UK Biobank underestimated the risk associated with multimorbidity. Absolute risk of mortality, hospitalisation and MACE, at all levels of multimorbidity, was lower in UK Biobank than in SAIL (adjusting for age, sex, and socioeconomic status). UK Biobank and SAIL produced similar hazard ratios for mortality for some LTCs (e.g. hypertension and coronary heart disease) but underestimated the risk for others (e.g. alcohol problems, mental health conditions, or pain). The hazard ratios for clusters of LTC including pain or mental health conditions were also lower in UK Biobank than SAIL.

Consequences

UK Biobank accurately represents the increased risks associated with moderate multimorbidity and therefore can be a useful resource for multimorbidity research. However, its lack of representativeness limits inferences about the minority of people with the highest degree of multimorbidity. The rich lifestyle, environmental, phenotypic and genetic data available in UK Biobank offers unique opportunities to understand factors associated with development of multimorbidity as well as relationships between multimorbidity and various associated phenomena. Nonetheless, where long-term conditions are the exposure, UK Biobank results are likely to be more conservative than in a general population cohort. UK Biobank is likely to underestimate the impact of the highest levels of multimorbidity.