Impact of COVID-19 pandemic on mental healthAn international study

Here's what you would predict about a global lockdown and mental health. On one hand, there is the destruction of routine, enforced isolation, economic shock, and an invisible lethal threat — all the ingredients for a mental health catastrophe. On the other hand, when everyone is locked down together, perhaps the shared experience softens the blow. Maybe the collective nature of it changes something. Both predictions are plausible, but both can't be fully right. A team led by Andrew Gloster surveyed nine thousand five hundred sixty-five people across seventy-eight countries and eighteen languages during the early months of 2020 to determine which prediction held true. What they found was more specific and more actionable than either prediction suggested. The study ran from April seventh to June seventh in 2020, when most participating countries had declared states of emergency. Recruitment was deliberately broad, including university email lists, local press, social media, professional networks, hospitals, churches, and schools. The sample skewed female, with seventy-seven point seven percent, and had a mean age of thirty-six point nine years. Participants had been in lockdown for a median of five weeks at the time they responded. Outcomes measured included stress using the Perceived Stress Scale, depressive symptoms, positive and negative affect using the Positive and Negative Affect Schedule — or PANAS — and overall wellbeing using the Mental Health Continuum Short Form. Predictors were grouped into four categories: sociodemographic factors, lockdown characteristics, social factors, and psychological factors. The team used linear mixed-effects models with a random effect for country, a statistical approach that considers the fact that people in Austria and Hong Kong are not drawing from the same baseline. So what did the population actually look like? Not collapsed, but not thriving either. Using the Mental Health Continuum Short Form, ten point one percent of the sample was classified as languishing — genuinely low mental health. About fifty percent fell into moderate mental health, and thirty-nine point nine percent were flourishing. The average wellbeing score was forty-one out of a possible seventy. Regarding stress, fifty-five point nine percent of respondents reported moderate stress, and eleven point one percent reported high stress. Average depressive symptoms scored six point six on their scale. About a third of participants reported significant boredom, and nearly half said they felt they were wasting time. Financially, thirty-three point three percent of respondents indicated that their situation had worsened during lockdown. That middle-ground distribution is striking. Most people were neither thriving nor in crisis. The majority occupied a kind of suspended moderate state — functional, but depleted. A consistent, nontrivial minority of roughly one in ten was genuinely struggling. Country differences existed but were mostly modest. Austria, Finland, and Portugal reported higher wellbeing relative to the reference, while Hong Kong remained notably lower, with a coefficient of negative six point eighty-four on the wellbeing scale. However, the bulk of the variation wasn't between countries; it was within them. And that's where the predictors come in. Three factors consistently showed up across every outcome measure: social support, education level, and psychological flexibility. Social support was the most powerful factor. In the multivariate models, high versus low social support predicted a stress reduction of three point thirty-five points — a standardized effect size of approximately negative zero point forty-four. For positive affect, the effect was even larger: a coefficient of five point sixty-nine with an effect size of zero point seventy-one. For wellbeing, the difference between high and low social support corresponded to a coefficient of thirteen point twenty, with an effect size approaching one point zero. Gloster and colleagues describe these as very large effects. In a study of this size, across this many countries, that consistency is hard to dismiss. Education was also protective. Higher education levels were associated with lower stress and higher positive affect, with coefficients around negative two point one to negative two point eight compared to primary school completion. The paper characterizes these as moderate to large effects, and they persisted after adjusting for country. Psychological flexibility, measured with the Psyflex scale, is worth explaining because it's not a term most people typically use. Gloster and colleagues define it as the tendency to hold one's thoughts lightly, accept difficult experiences rather than rigidly avoiding them, and continue acting in line with what matters to you even when conditions are hard. Think of it as the opposite of being dominated by your own mental weather. In the models, psychological flexibility predicted lower stress with a coefficient of negative zero point sixty-five and an effect size of negative zero point thirty-six, lower negative affect at negative zero point sixty-two, and higher positive affect at zero point seventy-seven. For wellbeing, the effect size was zero point forty-two. These numbers may not be the largest in the table, but they appear consistently across every outcome. Critically, psychological flexibility is trainable; it can be built through scalable interventions, including digital and online tools. Now, here's the counterintuitive part. You might assume that the concrete features of lockdown — how long you had been confined, how much space you had, and whether you could leave for work — would be major drivers of mental health. They weren't. Weeks in quarantine showed a standardized effect size of essentially zero for positive affect. Indoor living space had a near-zero coefficient for stress. Leaving home for work had small positive associations with affect — going out more than three times a week predicted a modest increase in positive affect and a small drop in negative affect. However, those effects were dwarfed by social and psychological factors. The number of weeks someone had been locked down barely moved the needle. Two lockdown features did matter. Worsening finances predicted higher stress, with a coefficient of two point thirty-two and an effect size of about zero point thirty-one, as well as worse depression. Not being able to obtain basic supplies predicted higher stress at one point eighty-two, worse depression, and lower positive affect. These aren't psychological abstractions. They're concrete material conditions — money and food — and when those deteriorated, mental health followed. So, the picture that emerges is this: the rules of lockdown mattered far less than the social and psychological environment people inhabited inside it. Someone who lost income or couldn't reliably get groceries was at meaningful risk. Someone without social support was at an even greater risk. Someone who was psychologically rigid, unable to adapt and driven by avoidance, was vulnerable across every measure. Meanwhile, the person sitting in a smaller apartment who had been locked down longer but retained social connections and some psychological flexibility tended to do better than those objective circumstances might predict. The public health implications follow directly from the predictors. Gloster and colleagues argue that interventions should target people without social support and those whose finances worsen as a result of lockdown. Ensuring access to basic supplies is not just a logistical concern — it shows up in the data as a mental health variable. Promoting psychological flexibility through scalable programs that are online, virtual, and widely distributed is a legitimate tool for reducing population-level psychological harm during large disruptions. The study has real limitations. It is cross-sectional, meaning it captures a snapshot rather than tracking people over time, so causal claims aren't warranted from this design alone. Recruitment was online, which means it likely undersampled people without internet access and those on the front lines of physical work. The sample skews female and relatively educated. Country-level differences in COVID incidence and lockdown stringency weren't systematically modeled. What makes the findings endure despite those limits is the consistency. The same predictors — social support, education, psychological flexibility, financial deterioration, access to basic needs — appeared across seventy-eight countries, eighteen languages, and five different outcome measures. That's not an artifact of one culture or one instrument. In a crisis that touched everyone, those who struggled most were not random. The patterns were there to be found, and this study uncovered them. The lesson is uncomfortable in its specificity. A global pandemic might feel like an equalizer, with everyone locked down and everyone at risk. But the data show that risk was not distributed evenly, and that distribution wasn’t mysterious. It tracked social connection, economic stability, and psychological adaptability. Those are things that vary predictably across populations, and they are things that policy can, at least in principle, address. Knowing that is the starting point for doing better next time. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.