Exploring the impact of selection bias in observational studies of COVID-19: a simulation study.

BACKGROUND: Non-random selection of analytic subsamples could introduce selection bias in observational studies. We explored the potential presence and impact of selection in studies of SARS-CoV-2 infection and COVID-19 prognosis. METHODS: We tested the association of a broad range of characteristics with selection into COVID-19 analytic subsamples in the Avon Longitudinal Study of Parents and Children (ALSPAC) and UK Biobank (UKB). We then conducted empirical analyses and simulations to explore the potential presence, direction and magnitude of bias due to this selection (relative to our defined UK-based adult target populations) when estimating the association of body mass index (BMI) with SARS-CoV-2 infection and death-with-COVID-19. RESULTS: In both cohorts, a broad range of characteristics was related to selection, sometimes in opposite directions (e.g. more-educated people were more likely to have data on SARS-CoV-2 infection in ALSPAC, but less likely in UKB). Higher BMI was associated with higher odds of SARS-CoV-2 infection and death-with-COVID-19. We found non-negligible bias in many simulated scenarios. CONCLUSIONS: Analyses using COVID-19 self-reported or national registry data may be biased due to selection. The magnitude and direction of this bias depend on the outcome definition, the true effect of the risk factor and the assumed selection mechanism; these are likely to differ between studies with different target populations. Bias due to sample selection is a key concern in COVID-19 research based on national registry data, especially as countries end free mass testing. The framework we have used can be applied by other researchers assessing the extent to which their results may be biased for their research question of interest.

The UK Coronavirus Job Retention Scheme and smoking, alcohol consumption and vaping during the COVID-19 pandemic: evidence from eight longitudinal population surveys.

BACKGROUND: Employment disruptions can impact smoking and alcohol consumption. During the COVID-19 pandemic, many countries implemented furlough schemes to prevent job loss. We examine how furlough was associated with smoking, vaping and alcohol consumption in the UK. METHODS: Data from 27,841 participants in eight UK adult longitudinal surveys were analysed. Participants self-reported employment status and current smoking, current vaping and alcohol consumption (>4 days/week or 5+ drinks per typical occasion) both before and during the early stages of the pandemic (April-July 2020). Risk ratios were estimated within each study using modified Poisson regression, adjusting for a range of potential confounders, including pre-pandemic behaviour. Findings were synthesised using random effects meta-analysis. RESULTS: Compared to stable employment and after adjustment for pre-pandemic characteristics, furlough was not associated with smoking (ARR = 1.05; 95% CI: 0.95-1.16; I2: 10%), vaping (ARR = 0.89; 95% CI: 0.74-1.08; I2: 0%) or drinking (ARR = 1.03; 95% CI: 0.94-1.13; I2: 48%). There were similar findings for no longer being employed, and stable unemployment, though this varied by sex: stable unemployment was associated with smoking for women (ARR = 1.35; 95% CI: 1.00-1.82; I2: 47%) but not men (0.84; 95% CI: 0.67-1.05; I2: 0%). No longer being employed was associated with vaping among women (ARR = 2.74; 95% CI: 1.59-4.72; I2: 0%) but not men (ARR = 1.25; 95% CI: 0.83-1.87; I2: 0%). CONCLUSIONS: We found no clear evidence of furlough or unemployment having adverse impacts on smoking, vaping or drinking behaviours during the early stages of the COVID-19 pandemic in the UK. Differences in risk compared to those who remained employed were largely explained by pre-pandemic characteristics.

Bias from questionnaire invitation and response in COVID-19 research: an example using ALSPAC.

Background: Longitudinal studies are crucial for identifying potential risk factors for infection with, and consequences of, COVID-19, but relationships can be biased if they are associated with invitation and response to data collection. We describe factors relating to questionnaire invitation and response in COVID-19 questionnaire data collection in a multigenerational birth cohort (the Avon Longitudinal Study of Parents and Children, ALSPAC). Methods: We analysed online questionnaires completed between the beginning of the pandemic and easing of the first UK lockdown by participants with valid email addresses who had not actively disengaged from the study. We assessed associations of pre-pandemic sociodemographic, behavioural, anthropometric and health-related factors with: i) being sent a questionnaire; ii) returning a questionnaire; and iii) item response (for specific questions). Analyses were conducted in three cohorts: the index children born in the early 1990s (now young adults; 41 variables assessed), their mothers (35 variables) and the mothers' partners (27 variables). Results: Of 14,849 young adults, 41% were sent a questionnaire, of whom 57% returned one. Item response was >95%. In this cohort, 78% of factors were associated with being sent a questionnaire, 56% with returning one, and, as an example of item response, 20% with keyworker status response. For instance, children from mothers educated to degree-level had greater odds of being sent a questionnaire (OR=5.59; 95% CI=4.87-6.41), returning one (OR=1.60; 95% CI=1.31-1.95), and responding to items (e.g., keyworker status OR=1.65; 95% CI=0.88-3.04), relative to children from mothers with fewer qualifications. Invitation and response rates and associations were similar in all cohorts. Conclusions: These results highlight the importance of considering potential biases due to non-response when using longitudinal studies in COVID-19 research and interpreting results. We recommend researchers report response rates and factors associated with invitation and response in all COVID-19 observational research studies, which can inform sensitivity analyses.

Bias from questionnaire invitation and response in COVID-19 research: an example using ALSPAC [version 2;peer review: 2 approved]

Background: Longitudinal studies are crucial for identifying potential risk factors for infection with, and consequences of, COVID-19, but relationships can be biased if they are associated with invitation and response to data collection. We describe factors relating to questionnaire invitation and response in COVID-19 questionnaire data collection in a multigenerational birth cohort (the Avon Longitudinal Study of Parents and Children, ALSPAC). Methods: We analysed online questionnaires completed between the beginning of the pandemic and easing of the first UK lockdown by participants with valid email addresses who had not actively disengaged from the study. We assessed associations of pre-pandemic sociodemographic, behavioural, anthropometric and health-related factors with: i) being sent a questionnaire;ii) returning a questionnaire;and iii) item response (for specific questions). Analyses were conducted in three cohorts: the index children born in the early 1990s (now young adults;41 variables assessed), their mothers (35 variables) and the mothers’ partners (27 variables). Results: Of 14,849 young adults, 41% were sent a questionnaire, of whom 57% returned one. Item response was >95%. In this cohort, 78% of factors were associated with being sent a questionnaire, 56% with returning one, and, as an example of item response, 20% with keyworker status response. For instance, children from mothers educated to degree-level had greater odds of being sent a questionnaire (OR=5.59;95% CI=4.87-6.41), returning one (OR=1.60;95% CI=1.31-1.95), and responding to items (e.g., keyworker status OR=1.65;95% CI=0.88-3.04), relative to children from mothers with fewer qualifications. Invitation and response rates and associations were similar in all cohorts. Conclusions: These results highlight the importance of considering potential biases due to non-response when using longitudinal studies in COVID-19 research and interpreting results. We recommend researchers report response rates and factors associated with invitation and response in all COVID-19 observational research studies, which can inform sensitivity analyses.

The UK Coronavirus Job Retention Scheme and diet, physical activity, and sleep during the COVID-19 pandemic: evidence from eight longitudinal population surveys.

BACKGROUND: In March 2020, the UK implemented the Coronavirus Job Retention Scheme (furlough) to minimise job losses. Our aim was to investigate associations between furlough and diet, physical activity, and sleep during the early stages of the COVID-19 pandemic. METHODS: We analysed data on 25,092 participants aged 16-66 years from eight UK longitudinal studies. Changes in employment, including being furloughed, were based on employment status before and during the first lockdown. Health behaviours included fruit and vegetable consumption, physical activity, and sleep. Study-specific estimates obtained using modified Poisson regression, adjusting for socio-demographic characteristics and pre-pandemic health and health behaviours, were statistically pooled using random effects meta-analysis. Associations were also stratified by sex, age, and education. RESULTS: Across studies, between 8 and 25% of participants were furloughed. Compared to those who remained working, furloughed workers were slightly less likely to be physically inactive (RR = 0.85; [95% CI 0.75-0.97]; I 2 = 59%) and did not differ overall with respect to low fruit and vegetable consumption or atypical sleep, although findings for sleep were heterogenous (I 2 = 85%). In stratified analyses, furlough was associated with lower fruit and vegetable consumption among males (RR = 1.11; [1.01-1.22]; I 2 = 0%) but not females (RR = 0.84; [0.68-1.04]; I 2 = 65%). Considering changes in quantity, furloughed workers were more likely than those who remained working to report increases in fruit and vegetable consumption, exercise, and hours of sleep. CONCLUSIONS: Those furloughed exhibited similar health behaviours to those who remained in employment during the initial stages of the pandemic. There was little evidence to suggest that adoption of such social protection policies in the post-pandemic recovery period and during future economic crises had adverse effects on population health behaviours.

Mental and social wellbeing and the UK coronavirus job retention scheme: Evidence from nine longitudinal studies.

BACKGROUND: The COVID-19 pandemic has led to major economic disruptions. In March 2020, the UK implemented the Coronavirus Job Retention Scheme - known as furlough - to minimize the impact of job losses. We investigate associations between change in employment status and mental and social wellbeing during the early stages of the pandemic. METHODS: Data were from 25,670 respondents, aged 17-66, across nine UK longitudinal studies. Furlough and other employment changes were defined using employment status pre-pandemic and during the first lockdown (April-June 2020). Mental and social wellbeing outcomes included psychological distress, life satisfaction, self-rated health, social contact, and loneliness. Study-specific modified Poisson regression estimates, adjusting for socio-demographic characteristics and pre-pandemic mental and social wellbeing, were pooled using meta-analysis. Associations were also stratified by sex, age, education, and household composition. RESULTS: Compared to those who remained working, furloughed workers were at greater risk of psychological distress (adjusted risk ratio, ARR = 1.12; 95%CI: 0.97, 1.29), low life satisfaction (ARR = 1.14; 95%CI: 1.07, 1.22), loneliness (ARR = 1.12; 95%CI: 1.01, 1.23), and poor self-rated health (ARR = 1.26; 95%CI: 1.05, 1.50). Nevertheless, compared to furloughed workers, those who became unemployed had greater risk of psychological distress (ARR = 1.30; 95%CI: 1.12, 1.52), low life satisfaction (ARR = 1.16; 95%CI: 0.98, 1.38), and loneliness (ARR = 1.67; 95%CI: 1.08, 2.59). Effects were not uniform across all sub-groups. CONCLUSIONS: During the early stages of the pandemic, those furloughed had increased risk of poor mental and social wellbeing, but furloughed workers fared better than those who became unemployed, suggesting that furlough may have partly mitigated poorer outcomes.

Continuing inequalities in COVID-19 mortality in England and Wales, and the changing importance of regional, over local, deprivation.

BACKGROUND: Observational studies have highlighted that where individuals live is far more important for risk of dying with COVID-19, than for dying of other causes. Deprivation is commonly proposed as explaining such differences. During the period of localised restrictions in late 2020, areas with higher restrictions tended to be more deprived. We explore how this impacted the relationship between deprivation and mortality and see whether local or regional deprivation matters more for inequalities in COVID-19 mortality. METHODS: We use publicly available population data on deaths due to COVID-19 and all-cause mortality between March 2020 and April 2021 to investigate the scale of spatial inequalities. We use a multiscale approach to simultaneously consider three spatial scales through which processes driving inequalities may act. We go on to explore whether deprivation explains such inequalities. RESULTS: Adjusting for population age structure and number of care homes, we find highest regional inequality in October 2020, with a COVID-19 mortality rate ratio of 5.86 (95% CI 3.31 to 19.00) for the median between-region comparison. We find spatial context is most important, and spatial inequalities higher, during periods of low mortality. Almost all unexplained spatial inequality in October 2020 is removed by adjusting for deprivation. During October 2020, one standard deviation increase in regional deprivation was associated with 20% higher local mortality (95% CI, 1.10 to 1.30). CONCLUSIONS: Spatial inequalities are greatest in periods of lowest overall mortality, implying that as mortality declines it does not do so equally. During the prolonged period of low restrictions and low mortality in summer 2020, spatial inequalities strongly increased. Contrary to previous months, we show that the strong spatial patterning during autumn 2020 is almost entirely explained by deprivation. As overall mortality declines, policymakers must be proactive in detecting areas where this is not happening, or risk worsening already strong health inequalities.

Psychological Distress Before and During the COVID-19 Pandemic Among Adults in the United Kingdom Based on Coordinated Analyses of 11 Longitudinal Studies.

Importance: How population mental health has evolved across the COVID-19 pandemic under varied lockdown measures is poorly understood, and the consequences for health inequalities are unclear. Objective: To investigate changes in mental health and sociodemographic inequalities from before and across the first year of the COVID-19 pandemic in 11 longitudinal studies. Design, Setting, and Participants: This cohort study included adult participants from 11 UK longitudinal population-based studies with prepandemic measures of psychological distress. Analyses were coordinated across these studies, and estimates were pooled. Data were collected from 2006 to 2021. Exposures: Trends in the prevalence of poor mental health were assessed in the prepandemic period (time period 0 [TP 0]) and at 3 pandemic TPs: 1, initial lockdown (March to June 2020); 2, easing of restrictions (July to October 2020); and 3, a subsequent lockdown (November 2020 to March 2021). Analyses were stratified by sex, race and ethnicity, education, age, and UK country. Main Outcomes and Measures: Multilevel regression was used to examine changes in psychological distress from the prepandemic period across the first year of the COVID-19 pandemic. Psychological distress was assessed using the 12-item General Health Questionnaire, the Kessler 6, the 9-item Malaise Inventory, the Short Mood and Feelings Questionnaire, the 8-item or 9-item Patient Health Questionnaire, the Hospital Anxiety and Depression Scale, and the Centre for Epidemiological Studies-Depression across different studies. Results: In total, 49 993 adult participants (12 323 [24.6%] aged 55-64 years; 32 741 [61.2%] women; 4960 [8.7%] racial and ethnic minority) were analyzed. Across the 11 studies, mental health deteriorated from prepandemic scores across all 3 pandemic periods, but there was considerable heterogeneity across the study-specific estimated effect sizes (pooled estimate for TP 1: standardized mean difference [SMD], 0.15; 95% CI, 0.06-0.25; TP 2: SMD, 0.18; 95% CI, 0.09-0.27; TP 3: SMD, 0.21; 95% CI, 0.10-0.32). Changes in psychological distress across the pandemic were higher in women (TP 3: SMD, 0.23; 95% CI, 0.11, 0.35) than men (TP 3: SMD, 0.16; 95% CI, 0.06-0.26) and lower in individuals with below-degree level education at TP 3 (SMD, 0.18; 95% CI, 0.06-0.30) compared with those who held degrees (SMD, 0.26; 95% CI, 0.14-0.38). Increased psychological distress was most prominent among adults aged 25 to 34 years (SMD, 0.49; 95% CI, 0.14-0.84) and 35 to 44 years (SMD, 0.35; 95% CI, 0.10-0.60) compared with other age groups. No evidence of changes in distress differing by race and ethnicity or UK country were observed. Conclusions and Relevance: In this study, the substantial deterioration in mental health seen in the UK during the first lockdown did not reverse when lockdown lifted, and a sustained worsening was observed across the pandemic period. Mental health declines have been unequal across the population, with women, those with higher degrees, and those aged 25 to 44 years more affected than other groups.

Pre-pandemic mental health and disruptions to healthcare, economic and housing outcomes during the COVID-19 pandemic: evidence from 12 UK longitudinal studies.

BACKGROUND: The COVID-19 pandemic has disrupted lives and livelihoods, and people already experiencing mental ill health may have been especially vulnerable. AIMS: Quantify mental health inequalities in disruptions to healthcare, economic activity and housing. METHOD: We examined data from 59 482 participants in 12 UK longitudinal studies with data collected before and during the COVID-19 pandemic. Within each study, we estimated the association between psychological distress assessed pre-pandemic and disruptions since the start of the pandemic to healthcare (medication access, procedures or appointments), economic activity (employment, income or working hours) and housing (change of address or household composition). Estimates were pooled across studies. RESULTS: Across the analysed data-sets, 28% to 77% of participants experienced at least one disruption, with 2.3-33.2% experiencing disruptions in two or more domains. We found 1 s.d. higher pre-pandemic psychological distress was associated with (a) increased odds of any healthcare disruptions (odds ratio (OR) 1.30, 95% CI 1.20-1.40), with fully adjusted odds ratios ranging from 1.24 (95% CI 1.09-1.41) for disruption to procedures to 1.33 (95% CI 1.20-1.49) for disruptions to prescriptions or medication access; (b) loss of employment (odds ratio 1.13, 95% CI 1.06-1.21) and income (OR 1.12, 95% CI 1.06 -1.19), and reductions in working hours/furlough (odds ratio 1.05, 95% CI 1.00-1.09) and (c) increased likelihood of experiencing a disruption in at least two domains (OR 1.25, 95% CI 1.18-1.32) or in one domain (OR 1.11, 95% CI 1.07-1.16), relative to no disruption. There were no associations with housing disruptions (OR 1.00, 95% CI 0.97-1.03). CONCLUSIONS: People experiencing psychological distress pre-pandemic were more likely to experience healthcare and economic disruptions, and clusters of disruptions across multiple domains during the pandemic. Failing to address these disruptions risks further widening mental health inequalities.

Time-varying selection bias in analyses of COVID-19 in UK Biobank

RationaleAssociations between COVID-19 risk factors and COVID-19 outcomes change over time, likely due to selection into who receives a COVID-19 test. When studies do not account for the changes in testing criteria, the association between a risk factor and outcome is a joint estimate across time. The transportability of a joint estimate aggregated over multiple testing periods may be limited. To improve generalisability, it is desirable to estimate effects net of time-varying selection.Aim1) Demonstrate variation in the association between covariates expected to associate with testing, and those which would not, on COVID-19 at different timepoints. 2) Apply methods to mitigate biases in empirical estimates.MethodsAnalyses will be carried out on up to 421,037 UK Biobank participants residing in England at baseline (mean age at baseline = 56;55% female). Risk factors will be determined at baseline (from 2006 to 2010), and COVID-19 outcomes will be ascertained from linked Public Health England COVID-19 test data and mortality statistics.Univariate cox proportional hazard models will be used to explore how associations between time-varying and time-stable variables change over time with;i) having a test for COVID-19, ii) testing positive for COVID-19 and iii) dying with COVID-19.Time-varying risk factors will be based on measures of socioeconomic position (SEP) including education, Townsend deprivation index and income. ABO blood group will be considered as a time-stable risk factor. Distinct time periods will be defined based on changes in testing definitions and changes in lockdown restrictions.Inverse probability weights will then be calculated for each time period. These weights will then be applied to models estimating risk across all time periods.Expected ResultsPreliminary analyses show that the size of the association between SEP and i) COVID-19 testing and ii) testing positive for COVID-19, changes across the course of the pandemic. These differences may be due to differential testing and not time-varying causal effects of the risk factor. We expect inverse probability weights will provide estimates closer to the true value for the association between each risk factor and outcome, independent of selection pressures on receiving a COVID-19 test. Population Health Relevance. Where studies do not account for time-varying selection pressures, the causal interpretations and the validity of results may be distorted. Where these findings are to be translated into developing population level or pharmaceutical interventions to mitigate against COVID-19 outcomes, efforts may be diverted away from more important risk factors.

Interrogating structural inequalities in COVID-19 mortality in England and Wales.

BACKGROUND: Numerous observational studies have highlighted structural inequalities in COVID-19 mortality in the UK. Such studies often fail to consider the hierarchical, spatial nature of such inequalities in their analysis, leading to the potential for bias and an inability to reach conclusions about the most appropriate structural levels for policy intervention. METHODS: We use publicly available population data on COVID-19-related mortality and all-cause mortality between March and July 2020 in England and Wales to investigate the spatial scale of such inequalities. We propose a multiscale approach to simultaneously consider three spatial scales at which processes driving inequality may act and apportion inequality between these. RESULTS: Adjusting for population age structure and number of local care homes we find highest regional inequality in March and June/July. We find finer grained within region inequality increased steadily from March until July. The importance of spatial context increases over the study period. No analogous pattern is visible for non-COVID-19 mortality. Higher relative deprivation is associated with increased COVID-19 mortality at all stages of the pandemic but does not explain structural inequalities. CONCLUSIONS: Results support initial stochastic viral introduction in the South, with initially high inequality decreasing before the establishment of regional trends by June and July, prior to reported regionality of the 'second-wave'. We outline how this framework can help identify structural factors driving such processes, and offer suggestions for a long-term, locally targeted model of pandemic relief in tandem with regional support to buffer the social context of the area.

Collider bias undermines our understanding of COVID-19 disease risk and severity.

Numerous observational studies have attempted to identify risk factors for infection with SARS-CoV-2 and COVID-19 disease outcomes. Studies have used datasets sampled from patients admitted to hospital, people tested for active infection, or people who volunteered to participate. Here, we highlight the challenge of interpreting observational evidence from such non-representative samples. Collider bias can induce associations between two or more variables which affect the likelihood of an individual being sampled, distorting associations between these variables in the sample. Analysing UK Biobank data, compared to the wider cohort the participants tested for COVID-19 were highly selected for a range of genetic, behavioural, cardiovascular, demographic, and anthropometric traits. We discuss the mechanisms inducing these problems, and approaches that could help mitigate them. While collider bias should be explored in existing studies, the optimal way to mitigate the problem is to use appropriate sampling strategies at the study design stage.