Emotional adaptation during a crisis: decline in anxiety and depression after the initial weeks of COVID-19 in the United States

The study investigates how depression and anxiety changed over time during the first wave of the COVID-19 pandemic in the United States and identifies demographic and dynamic factors associated with these trajectories. Prior work shows pandemics and economic crises worsen mental health, but humans also exhibit resilience and emotional adaptation. The authors hypothesized that depressive and anxiety symptoms would be elevated early in the pandemic and decline over time, with individual differences (e.g., age, sex, income, prior psychiatric diagnoses) and dynamic, pandemic-related factors (economic impact, perceived duration, informedness, social media use, and COVID-19 case trends) explaining variation in levels and change over time. Understanding these patterns is important for targeting mental health resources during ongoing crises.

The paper situates the work within evidence that pandemics and recessions heighten mental health problems, citing increased suicide during the 1918 influenza and mental health deterioration reported early in COVID-19 across several countries. Demographic risk factors for depression and anxiety include being female and lower socioeconomic status, while social and educational support confer resilience. Behavioral factors such as outdoor activity and limited social media exposure have been linked to better wellbeing during COVID-19. The literature also discusses overlap and distinctions between depression and anxiety symptomatology, comorbidity, and neurobiological and pharmacological differences. Prior recession research links financial insecurity and unemployment to increased suicide, reduced wellbeing, and higher psychiatric service use, suggesting socioeconomic stressors may causally contribute to worse mental health during crises.

Design: A 10-week, web-based longitudinal survey study conducted weekly from April 2 to June 4, 2020, using Prolific to recruit US adults. Participants: 1512 eligible US residents aged 18–64 with >90% Prolific approval enrolled within 24 hours starting April 2. After excluding duplicate/corrupt entries (60; 0.4% of observations) and failed attention checks (140; 0.93% of observations), valid data at time 1 were obtained from 1456 participants (49.2% female; mean age 35.04 ± 13.08). Weekly dropout was 3.90–11.72%; 743 completed all 10 weeks. Procedure: Weekly surveys within a 24-hour window assessed mental health and COVID-19-related perceptions/behaviors; demographics collected at baseline. Participants completed additional decision-making tasks (reported elsewhere). Compensation varied by week, with completion bonuses at weeks 5 and 10. Measures: Depression via Zung Self-Rating Depression Scale (SDS); anxiety via State-Trait Anxiety Inventory, state subscale (STAI-S). Dynamic variables collected weekly included: perceived economic impact of COVID-19 (−50 to +50, scaled to −0.5 to 0.5), informedness about COVID-19 (0–100, scaled to 0–1), social media use (0–100, scaled to 0–1), and subjective projection of pandemic duration (categorical: 1–2 months to >1 year). COVID-19 severity was computed as the 7-day running average of new daily US cases and the week-over-week change in that average. Preprocessing: Sex binarized (male vs female/other), race (white vs non-white), marital status (married vs not), income binned into 12 pre-COVID categories, and indicators for past/present mood disorder (MDD or bipolar) and anxiety disorder (GAD, panic disorder, social anxiety, OCD). Dynamic variables were scaled as noted. Statistical analysis: Two linear mixed-effects models were fit separately for depression and anxiety: outcome ~ 1 + age + sex + race + income + diagnosis + time + COVID-19 severity + economic impact + informedness + social media + COVID-19 future + (1 + time | participant). Marital status was later added where it improved fit. Time-by-demographic interactions were tested for improved fit. Model comparisons used likelihood ratio tests (anova in R). Repeated-measures ANOVAs (case-wise deletion; n=657) described overall time trends. Analyses used MATLAB R2018b for data handling and plotting, and R 4.0.5 (lme4, lmerTest) with FIML estimation, assuming missing at random conditional on covariates.

- Both depression and anxiety were highest at the beginning of April 2020 and declined over the 10 weeks (repeated-measures ANOVA: depression F(9,504)=18.10, p<0.001; anxiety F(9,504)=26.00, p<0.001). At week 1, mean STAI-S anxiety was 41.41 ± 13, exceeding a clinical cutoff of 40 (t≈4.1, p<0.001), while mean depression remained below a standard clinical cutoff. - Demographics: Younger age, female sex, and lower income were associated with higher depression and anxiety across time (all p<0.001). Past/present mood disorder diagnosis predicted higher depression; past/present anxiety disorder predicted higher anxiety (both p<0.001). Adding marital status improved only the depression model; being married was associated with lower depression (β≈−2.19, p<0.001) but not anxiety. - Dynamic factors: Worse perceived economic impact predicted higher depression (p<0.001) and anxiety (p<0.001). Longer projected pandemic duration predicted increases in both depression (p=0.001) and anxiety (p<0.001). Higher informedness predicted lower depression (p<0.001) but did not affect anxiety. Greater social media use predicted higher anxiety (p=0.012) but not depression. The week-over-week change in national COVID-19 cases (but not absolute case counts) predicted higher anxiety (p<0.001). - Interactions with time: Only the age-by-time interaction improved the depression model fit (χ²=9.55, p=0.002), with a positive interaction coefficient (β≈0.004, p=0.002), indicating that although older adults were less depressed overall, their depressive symptoms decreased less over time compared to younger adults. No time interactions significantly improved the anxiety model. - Trends in the full sample closely matched those of the 10-week completer subsample.

The findings support the hypothesis of emotional adaptation during a prolonged crisis: depression and anxiety peaked early in the first COVID-19 wave in the US and declined over subsequent weeks. Both stable demographic factors and changing pandemic-related factors explained variability in symptoms. Financial hardship and longer subjective projections of pandemic duration were linked to higher depression and anxiety, underscoring socioeconomic drivers of mental health burden during crises. Distinct patterns emerged between disorders: anxiety tracked with social media use and week-over-week increases in national cases—factors plausibly linked to uncertainty and immediate threat—while depression decreased with greater informedness, consistent with a role for information in fostering hope or perceived control. Marriage buffered depression but not anxiety, and younger individuals showed larger declines in depression over time than older adults, though younger participants were more distressed overall. These results refine understanding of overlapping and distinct determinants of depression and anxiety during widespread societal stress and highlight potential targets—economic support, information dissemination, and media consumption moderation—for mitigation efforts.

This study demonstrates that US adults exhibited a decline in depression and anxiety over the first 10 weeks of the COVID-19 pandemic, indicative of emotional adaptation. It identifies key demographic (younger age, female sex, lower income, prior psychiatric diagnoses) and dynamic factors (economic impact, projected pandemic duration, informedness, social media use, and change in case counts) associated with symptom levels and trajectories, with partially distinct patterns for depression versus anxiety. Findings can inform clinicians and policymakers in allocating mental health resources and designing interventions during ongoing or future crises. Future research should establish causal pathways (e.g., effects of financial assistance, media exposure interventions), explore mechanisms underlying disorder-specific associations, and examine generalizability in more representative samples and across longer time spans.

- No pre-pandemic baseline measures; comparisons relied on previously published community norms rather than within-sample baselines. - Observational and correlational design limits causal inference for most associations (e.g., social media use and anxiety; economic impact and depression). - Sample not fully representative of the US population (younger and more White than census estimates), which may limit generalizability of demographic effects. - Missing data managed under MAR assumptions; repeated-measures ANOVAs used case-wise deletion for time-trend description.