The Coronavirus Health and Impact Survey (CRISIS) reveals reproducible correlates of pandemic-related mood states across the Atlantic

The researchers did not test an intervention. Instead, they tested a set of **predictors** (independent variables) to see which ones best explained **negative mood states** (the outcome) during the COVID-19 pandemic. The predictors were:

**Pre-existing mood states:** How the person felt emotionally *before* the pandemic (retrospectively reported).

**Perceived COVID-19 risk:** How much the person worried about catching the virus, getting seriously ill, or having family members affected.

**Lifestyle changes:** Self-reported disruptions to daily routines, including changes in sleep, physical activity, social contact, work/study patterns, and time spent outdoors.

The outcome measures were:

**Negative mood states:** A composite score capturing anxiety, depression, irritability, and loneliness during the pandemic.

**Positive mood states:** A separate composite score capturing calmness, happiness, and energy.

The study also tested whether the same predictors worked equally well in the US and UK, and whether the CRISIS survey itself (a new questionnaire) was reliable and valid.

The study included two separate samples:

**US sample:** 1,035 adults (mean age 41.2 years, range 18–87; 68% female; 79% White, 8% Black, 7% Asian, 6% other). Recruited via online platforms (Amazon Mechanical Turk and Prolific) between April and May 2020.

**UK sample:** 1,004 adults (mean age 39.8 years, range 18–80; 70% female; 90% White, 4% Asian, 3% Black, 3% other). Recruited via Prolific between May and June 2020.

Additionally, the study included **parent reports on children** (ages 2–17) from both countries:

US: 1,035 parents reporting on one child each.

UK: 1,004 parents reporting on one child each.

All participants were required to be fluent in English and living in their respective country during the pandemic lockdowns. The study explicitly excluded anyone who had already been diagnosed with COVID-19.

The researchers developed a new survey called the **CoRonavIruS health and Impact Survey (CRISIS)**. It is a self-report questionnaire that takes approximately 15–20 minutes to complete online. Key scales within CRISIS included:

**Negative mood state (NMS):** 6 items (e.g., "felt anxious," "felt depressed," "felt lonely") rated on a 5-point scale from "not at all" to "extremely" over the past 2 weeks. Score range: 6–30, higher = worse mood.

**Positive mood state (PMS):** 4 items (e.g., "felt calm," "felt energetic") rated on the same 5-point scale. Score range: 4–20, higher = better mood.

**Pre-existing mood state (retrospective):** Same 6-item NMS scale, but participants were asked to recall how they felt "before the pandemic" (i.e., January 2020). Score range: 6–30.

**Perceived COVID-19 risk:** 4 items (e.g., "How likely do you think it is that you will get infected?" "How worried are you about a family member getting infected?") rated on 5-point scales. Score range: 4–20, higher = greater perceived risk.

**Lifestyle change index:** 8 items covering changes in sleep, physical activity, social contact, work/study, time outdoors, screen time, diet, and substance use. Each item rated on a 5-point scale from "much less" to "much more" compared to before the pandemic. The researchers calculated a "disruption score" based on the absolute deviation from "no change."

The CRISIS survey also included demographic questions, COVID-19 exposure history, and a brief measure of resilience (the Brief Resilience Scale, 6 items, 1–5 scale).

**Study design:** This is a **cross-sectional observational study** with a replication component. The researchers collected data at a single time point from two independent samples (US and UK) during the first wave of the pandemic (April–June 2020). They then tested whether the same statistical models (predictors of mood) worked in both samples.

**Why this design matters:** Cross-sectional designs are fast, cheap, and can identify strong associations between variables. The replication across two countries is a major strength — if a predictor works in both the US and UK, it is less likely to be a fluke or a cultural artifact. However, because data are collected at one time point, **causality cannot be established**. For example, if people with worse mood also report more lifestyle disruption, we cannot tell whether the disruption caused the bad mood, the bad mood caused the disruption, or a third factor (e.g., job loss) caused both.

**Statistical approach:** The researchers used **multiple linear regression** to predict negative mood state (NMS) from pre-existing mood, perceived COVID risk, and lifestyle changes, while controlling for age, sex, and education. They reported **standardized beta coefficients** (which allow comparison of predictor strength) and **R² values** (the proportion of variance explained). They also used **confirmatory factor analysis** to test whether the CRISIS survey items grouped into the expected subscales (e.g., negative mood, positive mood, risk perception).

**What this design can prove:**

That certain factors are **strongly associated** with mood during a pandemic.

That these associations are **reproducible** across two different populations.

That the CRISIS survey is a **reliable and valid** measurement tool.

**What this design cannot prove:**

**Causation:** It cannot tell us whether changing lifestyle would improve mood, or whether improving mood would change lifestyle.

**Directionality:** It cannot tell us whether pre-existing mood caused later mood, or whether people with bad mood simply remember their past as worse (recall bias).

**Long-term effects:** It captures only a single snapshot during the first lockdown; we do not know if these relationships held up months later.

**Major methodological weaknesses:**

1. **Retrospective recall of pre-existing mood:** People were asked to remember how they felt months earlier (January 2020). Memory is notoriously unreliable, especially during a stressful event. People with current bad mood may overestimate how good they felt before (contrast effect) or underestimate it (consistency bias).

2. **Self-report only:** No objective measures of mood, activity, or health. Social desirability bias and common method variance (using the same survey for predictors and outcomes) can inflate correlations.

3. **Convenience sampling:** Participants were recruited from online platforms, not randomly selected from the general population. The samples were predominantly female, White, and well-educated, limiting generalizability.

4. **Single time point:** Cannot assess change over time or within-person dynamics.

5. **No pre-pandemic baseline:** The "pre-existing mood" measure is retrospective, not a true baseline collected before the pandemic.

All results below are from the adult samples unless noted. The parent-report child data showed similar patterns but are not detailed here for brevity.

**Primary outcome: Negative mood state (NMS) during the pandemic**

**Pre-existing mood was the strongest predictor** of pandemic negative mood in both samples:

- US: β = 0.46, p < 0.001 (standardized beta; a 1-standard-deviation increase in pre-existing negative mood was associated with a 0.46-SD increase in pandemic negative mood).

- UK: β = 0.44, p < 0.001.

- This means that about 20% of the variance in pandemic mood was explained by pre-pandemic mood alone.

**Lifestyle changes were the second strongest predictor:**

- US: β = 0.22, p < 0.001.

- UK: β = 0.24, p < 0.001.

- Specifically, greater disruption to routines (especially reduced physical activity, increased screen time, and reduced social contact) was associated with worse mood.

**Perceived COVID-19 risk was a weaker but still significant predictor:**

- US: β = 0.12, p < 0.001.

- UK: β = 0.10, p < 0.01.

- Worrying about infection was associated with worse mood, but the effect was small compared to pre-existing mood and lifestyle.

**Combined model (all predictors + demographics):**

- US: R² = 0.38 (38% of variance in pandemic negative mood explained).

- UK: R² = 0.36 (36% explained).

- This means that even with all predictors, over 60% of the variation in mood was unexplained.

**Secondary outcome: Positive mood state (PMS) during the pandemic**

Pre-existing positive mood was the strongest predictor (US: β = 0.42, UK: β = 0.40, both p < 0.001).

Lifestyle changes were also significant but in the opposite direction: more disruption predicted less positive mood (US: β = -0.18, UK: β = -0.20, both p < 0.001).

Perceived COVID risk was not significantly associated with positive mood in either sample.

**Reliability and validity of CRISIS:**

Internal consistency (Cronbach's alpha) for the negative mood scale: 0.89 (US) and 0.90 (UK) — excellent.

Internal consistency for the positive mood scale: 0.82 (US) and 0.84 (UK) — good.

Confirmatory factor analysis supported the intended factor structure (CFI > 0.90, RMSEA < 0.08 in both samples).

**Reproducibility across countries:**

The researchers explicitly tested whether the regression coefficients differed between the US and UK samples. They found **no significant differences** for any predictor (all interaction p-values > 0.05). This means the same model worked equally well in both countries.

To translate these numbers into plain English:

**Pre-existing mood:** If you were already feeling anxious or depressed before the pandemic, you were likely to feel *even more* anxious or depressed during it. The effect was large: a person who scored 1 standard deviation above average on pre-existing negative mood (roughly the top 16% of the population) would score about 0.45 standard deviations above average on pandemic negative mood. In practical terms, on a 6–30 scale, this translates to about a 3–4 point difference — the difference between "a little" anxious and "moderately" anxious.

**Lifestyle disruption:** The effect was about half as strong as pre-existing mood. A person with major lifestyle disruption (e.g., stopped exercising, lost social contact, increased screen time) would score about 1.5–2 points higher on the negative mood scale compared to someone with minimal disruption. This is roughly equivalent to the difference between "not at all" and "a little" depressed.

**Perceived COVID risk:** The effect was small. A person who was extremely worried about infection would score only about 0.5–1 point higher on negative mood compared to someone not worried at all. This suggests that worrying about the virus itself was less important for mood than the *secondary effects* of the pandemic (disrupted routines, lost social connection).

**Overall model:** The combined predictors explained about 37% of mood variation. This is a moderate-to-large effect for psychological research (typical R² values for mood prediction are 10–25%). However, it also means that **63% of why people felt bad was not captured by these factors** — things like genetics, personality, social support, financial stress, and random daily events matter a lot.

**What the authors acknowledge:**

Cross-sectional design prevents causal inference.

Retrospective recall of pre-pandemic mood is subject to memory bias.

Convenience sampling limits generalizability to the broader population.

Self-report measures may be influenced by social desirability and common method variance.

The study was conducted during the first lockdown (April–June 2020); results may not generalize to later pandemic phases.

**What a critical reader would add:**

**No objective measures:** The study relies entirely on self-report. We do not know if people who reported "more lifestyle disruption" actually changed their behavior, or if they just perceived it that way. Objective measures (e.g., smartphone GPS for mobility, wearable activity trackers) would strengthen the findings.

**Single time point:** Even within the cross-sectional design, collecting data at only one point means we cannot assess within-person change. A person who was already depressed before the pandemic might have been depressed regardless.

**Recall bias is severe:** Asking people to remember their mood from 3–5 months earlier (January to April/May 2020) is highly unreliable. People's current mood strongly colors their memory of past mood. This could artificially inflate the correlation between "pre-existing" and "current" mood.

**No control for pandemic exposure:** The study did not measure actual COVID-19 infection, job loss, or bereavement. People who lost loved ones or their jobs would likely have worse mood, and this confound is not accounted for.

**Demographic homogeneity:** The samples were predominantly White, female, and well-educated. Results may not apply to men, non-White populations, or people with lower education/income.

**No blinding or randomization:** Not applicable to an observational study, but worth noting that the design cannot rule out reverse causation or third variables.

**Parent-report data on children is even weaker:** Parents may not accurately know their child's internal mood state, especially for teenagers. The child data should be interpreted with caution.

For someone running their own n=1 experiment on mood and lifestyle during a stressful period (e.g., a pandemic, a major life transition, or even a personal challenge like exam season):

### What to test

**The intervention:** Deliberately maintain or restore a specific lifestyle routine. Based on this study, the most promising targets are:

- **Physical activity:** Aim for at least 30 minutes of moderate exercise daily (e.g., brisk walking, cycling, bodyweight exercises).

- **Social contact:** Schedule at least one meaningful social interaction per day (phone call, video chat, or in-person if safe).

- **Screen time:** Limit recreational screen use (social media, streaming) to no more than 2 hours per day outside of work/study.

- **Time outdoors:** Aim for at least 20 minutes outside in daylight each day.

**The comparator:** Your own baseline. Track your mood and routines for 1–2 weeks *before* making any changes, then compare.

**The outcome:** Your daily negative mood (anxiety, depression, irritability) and positive mood (calm, energy, happiness).

### Minimum meaningful duration

**At least 3 weeks:** This study looked at mood over the past 2 weeks. For a self-experiment, you need at least 1 week of baseline tracking, then 2 weeks of the intervention to see a stable effect. Longer is better (4–6 weeks total) because mood fluctuates with weekly cycles (e.g., weekend vs. weekday).

### What to measure (specific metrics)

**Daily negative mood:** Rate on a 1–5 scale (1 = not at all, 5 = extremely) for: anxious, depressed, irritable, lonely. Average these 4 items for a daily score (range 1–5).

**Daily positive mood:** Rate on a