Investigators from the COVID-19 Risks Across the Lifespan (CORAL) Study reported that, on average, alcohol frequency and quantity did not change markedly across three assessment points spanning May 2020 to April 2021. The cohort did not show a uniform increase in drinking over the first pandemic year; instead, reported use remained broadly similar to pre-pandemic reports when averaged across the sample. The study team assessed pre-pandemic drinking at baseline using the first two AUDIT items and asked participants at each timepoint whether their current drinking frequency and quantity were “lot less” to “a lot more” relative to pre-pandemic levels. Models accounting for time and age found no overall linear upward trend in frequency or quantity across T1–T3.
However, investigators observed non-linear age associations with reported changes. Model comparisons retained a cubic age term as the best fit, indicating complex, age-dependent variation in who reported increasing consumption. Specifically, the youngest cohort (ages 11–20) and a mid-aged group (approximately 50–70 years) had the highest probability of reporting increased frequency and quantity compared with other ages. The authors noted these age patterns rather than a simple monotonic relationship between age and change in drinking.
The study team examined several binary drinking behaviours and motivations relative to pre-pandemic habits, including whether participants began drinking earlier in the day, drank alone more often, or had trouble stopping once they started. Overall, these behaviours and the most commonly endorsed motivations (for example, drinking to destress or to reduce boredom) remained largely stable across the three waves. Mixed-effects logistic models with timepoint and baseline age as predictors did not show sweeping shifts in these behaviours for the whole sample.
Country- and age-specific differences did emerge, however. The authors reported variations by country in some behaviours and motivations, a pattern they attributed in part to differing public policies—such as restrictions on on-premise drinking or off-premise alcohol availability—and varying lockdown measures across Australia, the United Kingdom, and the United States. The CORAL protocol did not attempt to model policy exposure directly; instead, country was included as a categorical covariate in all analyses. The team also modelled age effects and tested interactions with time where feasible, noting that non-linear age effects informed several findings.
Cross-lagged panel analyses tested bidirectional associations between alcohol frequency and quantity and multiple mental health constructs: general perceived stress (PSS), COVID-19-related stress (PAS), loneliness (UCLA scale), depressive symptoms (PHQ-8), and anxiety symptoms (GAD-7). Country was entered as a covariate and robust maximum likelihood estimation was used.
Data indicated a unidirectional association in which higher loneliness predicted subsequent increases in alcohol quantity. The study team reported that loneliness at an earlier timepoint was associated with later higher reported quantity of drinking, whereas the reverse pathway—higher alcohol use predicting subsequent loneliness—was not supported in these models. The authors therefore emphasised loneliness as an important public health factor with implications for risk of elevated alcohol consumption during the pandemic period.
The analytic sample included 2,369 participants after excluding those missing age or the baseline alcohol-drinker item. The mean age was 38.81 years (SD reported in the source); ages ranged from 11 to 100 years. Recruitment was via convenience sampling online, and 90% of the retained sample were female. The CORAL Study assessed participants at three waves roughly three months apart: baseline (T1), a 3-month follow-up (T2), and a 6-month follow-up (T3), spanning May 2020 to April 2021.
Alcohol measures comprised a baseline retrospective pre-pandemic frequency and quantity assessment using AUDIT items, and repeated relative-change items comparing current use to pre-pandemic levels. Behaviours and motivations were captured with binary items. Mental health instruments included validated scales with reported internal consistency: the PAS for COVID-related stress, the PSS for general stress, the UCLA Loneliness Scale, PHQ-8 for depressive symptoms, and GAD-7 for anxiety.
Analyses were conducted in R using cumulative link ordinal regressions for change outcomes, mixed-effects ordinal and logistic models for longitudinal change and behaviours, and cross-lagged panel models for directional associations. Complete-case analysis handled missing data; country was modelled as a covariate throughout. The authors compared linear, quadratic, and cubic age terms and retained cubic terms where model fit supported them.
The authors documented several limitations that affect interpretation. Recruitment relied on convenience online sampling, creating potential selection bias and limiting representativeness—particularly given the sample’s 90% female composition. Pre-pandemic alcohol use was assessed only at baseline via recall, and participants with missing data on alcohol items were excluded via complete-case approaches, which may bias results. The study did not report direct measures of local alcohol policy exposure or objective sales data, so inferences linking country differences to policy remain speculative within the dataset. Cross-lagged modelling addressed temporal ordering but cannot establish definitive causality.
Remaining questions noted by the investigators include whether observed age-specific increases persist beyond the first pandemic year, how policy-level differences in alcohol availability quantitatively influenced individual consumption, and how interventions addressing loneliness might modify downstream alcohol risk. These points, along with the sampling limitations, frame the next steps that the authors identify for research and public health monitoring.