Frailty and comorbidity in predicting community COVID-19 mortality in the UK Biobank: the effect of sampling

Frailty has been linked to increased risk of COVID-19 mortality, but evidence is mainly limited to hospitalized older individuals and analyses in community samples are scarce. This study aims to assess and compare the predictive abilities of different frailty measures – the frailty phenotype (FP), frailty index (FI), and Hospital Frailty Risk Score (HFRS), and comorbidity, measured using the Charlson Comorbidity Index (CCI), on COVID-19 mortality in a UK community sample of adults aged 52–86 years. We analyzed (i) the full sample of 428,754 UK Biobank participants and (ii) a subsample of 2,287 COVID-19 positive UK Biobank participants with data on COVID-19 outcomes between March 1 and September 21, 2020. COVID-19 positivity was confirmed by PCR, hospital records and/or death registers. Logistic regression models adjusted for age, sex, smoking, ethnicity, and socioeconomic variables with areas under the receiver operating characteristic curves (AUCs) were used in the modelling. Overall, 391 individuals died of COVID-19. In the full sample, all frailty measures and the CCI were associated with COVID-19 mortality but only the HFRS and CCI improved the predictive ability of a model including age and sex, yielding AUCs>0.80. However, when restricting analyses to the COVID-19 positive subsample, which had an over-representation of frail individuals, similar improvement in AUCs was not observed in which only the CCI was significantly associated with COVID-19 mortality. Our results suggest that HFRS and CCI can be used in COVID-19 mortality risk stratification at the population level, but they show limited added value in COVID-19 positive individuals.