Tobacco smoking by disability status before and after COVID-19 onset: a repeated cross-sectional analysis of 1 087 678 adults

Abstract

Smoking remains a leading cause of preventable illness and death in the United Kingdom, but little is known about recent trends in smoking disparities between disabled and non-disabled adults, particularly in the context of the coronavirus disease 2019 (COVID-19) pandemic. We analysed UK Annual Population Survey data from 2017 to 2023 for adults aged ≥18 years. Smoking status was classified as current, ex-, or never smoker, and disability status was defined according to the Equality Act. Multinomial logistic regression was used to estimate pooled adjusted relative risk ratios (RRRs) for smoking outcomes by disability status, and then separately for periods before (2017-19) and after (2020-23) the pandemic onset. To test whether the pandemic had an effect beyond underlying trends, we fitted a joint model including survey year and a post-2020 indicator. Adjusted average marginal effects quantified absolute percentage point (pp) differences in predicted probabilities between disabled and non-disabled adults. The analytic sample comprised 1 087 678 adults, of whom 26.7% (n = 290 536) reported a disability. In pooled adjusted analyses covering all survey years, and controlling for age, sex, education, ethnicity, marital status, and region of residence, disabled adults had higher relative risk ratios of being current smokers (RRR = 1.78; 95% CI: 1.75-1.80; P < .001) and ex-smokers (RRR = 1.44; 95% CI: 1.42-1.45; P < .001) compared with never smokers. Period-stratified analyses (not adjusted for temporal trends) showed adjusted RRRs for current smoking of 1.65 (95% CI: 1.62-1.68; P < .001) before and 1.95 (95% CI: 1.91-1.99; P < .001) after the pandemic onset. In the fully adjusted joint model accounting for temporal trends (survey year) and pandemic period, the disability × pandemic period interaction was not statistically significant (χ2 = 3.11; P = .21). Adjusted average marginal effects from the trend-adjusted model showed that disabled adults had a higher predicted probability of current smoking both before (+5.63 percentage points; 95% CI: 5.30-5.96; P < .001) and after (+4.60 pp; 95% CI: 4.25-4.96; P < .001) the pandemic onset, representing a modest narrowing of the absolute gap (difference = -1.03 pp; P = .001). Disabled adults remained substantially more likely to smoke than their non-disabled counterparts. After accounting for underlying temporal trends, the onset of the COVID-19 pandemic did not independently change this association, highlighting the continued need for disability-inclusive cessation strategies.