Non-communicable disease, sociodemographic factors, and risk of death from infection: a UK Biobank observational cohort study.

Background: Non-communicable diseases (NCDs) have been highlighted as important risk factors for COVID-19 mortality. However, insufficient data exist on the wider context of infectious diseases in people with NCDs. We aimed to investigate the association between NCDs and the risk of death from any infection before the COVID-19 pandemic (up to Dec 31, 2019).
Methods: For this observational study, we used data from the UK Biobank observational cohort study to explore factors associated with infection death. We excluded participants if data were missing for comorbidities, body-mass index, smoking status, ethnicity, and socioeconomic deprivation, and if they were lost to follow-up or withdrew consent. Deaths were censored up to Dec 31, 2019. We used Poisson regression models including NCDs present at recruitment to the UK Biobank (obesity [defined by use of body-mass index] and self-reported hypertension, chronic heart disease, chronic respiratory disease, diabetes, cancer, chronic liver disease, chronic kidney disease, previous stroke or transient ischaemic attack, other neurological disease, psychiatric disorder, and chronic inflammatory and autoimmune rheumatological disease), age, sex, ethnicity, smoking status, and socioeconomic deprivation. Separate models were constructed with individual NCDs replaced by the total number of prevalent NCDs to define associations with multimorbidity. All analyses were repeated with non-infection-related death as an alternate outcome measure to establish differential associations of infection death and non-infection death. Associations are reported as incidence rate ratios (IRR) accompanied by 95% CIs.
Findings: After exclusion of 9210 (1·8%) of the 502 505 participants in the UK Biobank cohort, our study sample comprised 493 295 individuals. During 5 273 731 person-years of follow-up (median 10·9 years [IQR 10·1-11·6] per participant), 27 729 deaths occurred, of which 1385 (5%) were related to infection. Advancing age, male sex, smoking, socioeconomic deprivation, and all studied NCDs were independently associated with the rate of both infection death and non-infection death. Compared with White ethnicity, a pooled Black, Asian, and minority ethnicity group was associated with a reduced risk of infection death (IRR 0·64, 95% CI 0·46-0·87) and non-infection death (0·80, 0·75-0·86). Stronger associations with infection death than with non-infection death were observed for advancing age (age 65 years vs 45 years: 7·59, 95% CI 5·92-9·73, for infection death vs 5·21, 4·97-5·48, for non-infection death), current smoking (vs never smoking: 3·69, 3·19-4·26, vs 2·52, 2·44-2·61), socioeconomic deprivation (most vs least deprived quintile: 2·13, 1·78-2·56, vs 1·38, 1·33-1·43), class 3 obesity (vs non-obese: 2·21, 1·74-2·82, vs 1·55, 1·44-1·66), hypertension (1·36, 1·22-1·53, vs 1·15, 1·12-1·18), respiratory disease (2·21, 1·96-2·50, vs 1·28, 1·24-1·32), chronic kidney disease (5·04, 4·28-7·31, vs 2·50, 2·20-2·84), psychiatric disease (1·56, 1·30-1·86, vs 1·23, 1·18-1·29), and chronic inflammatory and autoimmune rheumatological disease (2·45, 1·99-3·02, vs 1·41, 1·32-1·51). Accrual of multimorbidity was also more strongly associated with risk of infection death (five or more comorbidities vs none: 9·53, 6·97-13·03) than of non-infection death (5·26, 4·84-5·72).
Interpretation: Several NCDs are associated with an increased risk of infection death, suggesting that some of the reported associations with COVID-19 mortality might be non-specific. Only a subset of NCDs, together with the accrual of multimorbidity, advancing age, smoking, and socioeconomic deprivation, were associated with a greater IRR for infection death than for other causes of death. Further research is needed to define why these risk factors are more strongly associated with infection death, so that more effective preventive strategies can be targeted to high-risk groups.
Funding: British Heart Foundation.
(Copyright © 2021 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. Published by Elsevier Ltd.. All rights reserved.)