Loneliness in the UK during the COVID-19 pandemicCross-sectional results from the COVID-19 Psychological Wellbeing Study

Picture the first weeks of 2020 in the United Kingdom. On January thirty-first, the first confirmed SARS-CoV-2 case. By March twenty-third, a nationwide lockdown. Overnight, the fabric of daily life tightened: doors shut, commutes vanished, hugs and handshakes disappeared. Even before the pandemic, policymakers had been calling loneliness a public health problem in its own right. Now the worry was immediate and concrete: if in-person contact evaporated, would loneliness spike, and if so, for whom? That’s the question Groarke and colleagues set out to answer. They didn’t wait for hindsight. On the very day lockdown began, they launched an online study, the UK COVID-19 Psychological Wellbeing Study, to take a real-time snapshot of how people were doing. The aim was simple but urgent: measure how common loneliness was in those first weeks and map out who was most at risk and what might protect people. The logic was practical. If you know which levers matter—age, living situation, mental health, sleep—you can target support where it will land. Let me set the stage for how they did it. Between March twenty-third and April twenty-fourth, 2020, nearly two thousand adults completed a baseline survey. The final sample for the loneliness analysis was one thousand nine hundred sixty-four people, with an average age of about thirty-seven, and roughly seventy percent identifying as female. Recruitment came through two channels: a UK panel on Prolific, where participants received a small payment, and social media, where participants entered a prize draw. This is not a random sample. Older adults and men were underrepresented compared to census figures, which means we should be careful about generalizing. But it is a timely, detailed look at a critical moment. How did they measure loneliness? With the short, three-item University of California, Los Angeles Loneliness Scale, the version that asks, in plain language, how often you feel left out, isolated, or lacking companionship. It’s a blunt tool by design, but reliable and widely used. People scoring seven or higher out of nine were counted as lonely. Around that core, Groarke’s team layered in other pieces: perceived social support; how many adults lived in your household; relationship status; sleep quality in general and sleep specifically affected by COVID; emotional regulation difficulties—think of that as how hard it is to manage strong feelings; and mental health symptoms captured with standard screens for depression, anxiety, and post-traumatic stress. The plan was to look at simple associations first, then build a multivariable model—logistic regression—where all the plausible factors compete at once. Only variables that showed a hint of association in the initial pass went into that final model, which reports adjusted odds ratios. In human terms, those ratios tell you, for each factor, how the odds of being lonely shift after accounting for everything else. The headline is stark. In those first four weeks of lockdown, twenty-seven percent of respondents met criteria for loneliness. Roughly one in four. That’s not a small ripple; that’s a quarter of adults saying, even with video calls and group chats, they felt cut off. Now, prevalence alone doesn’t tell you where to intervene. The power of this study is in the pattern underneath. Age dominated. Compared with adults sixty-five and older, younger adults had dramatically higher odds of being lonely. The youngest group—eighteen to twenty-four—had about five times the odds. Those twenty-five to thirty-four were close behind, around four point seven times. Even the forty-five to fifty-four group stood out at roughly four point eight times. Middle groups like thirty-five to forty-four and fifty-five to sixty-four trended higher but didn’t clear the usual statistical bar. Pause on that for a beat. In a pandemic that was physically most dangerous for the old, emotional isolation hit the young the hardest. That flips a common stereotype on its head and matches what many younger adults reported anecdotally: work uprooted, social rituals gone, milestones deferred. If your life runs on in-person connection, pressure on that system shows up fast. Social context mattered too, and in both directions. Feeling supported by others—measured as a simple “how strongly do you agree that you have people to rely on?”—was protective. Each notch up on that support scale nudged the odds of loneliness down, about eight percent per step. Living with a partner was especially powerful. Married or cohabiting adults had roughly a third the odds of being lonely compared with people who were single and never married. Add one more heartening detail: every additional adult in the household helped. Each extra housemate—family member, partner, friend—was associated with lower odds of loneliness, a modest effect that makes intuitive sense when daily life compresses to four walls. On the flip side, relationship disruption amplified risk. Being separated or divorced roughly doubled the odds of loneliness, even after taking the rest of the context into account. Widowed adults trended higher but didn’t reach conventional significance in this model, which may reflect the sample’s age skew. The broader message threads through these findings: connection buffers, and ruptures sting. Mental health and emotional skills were the other big levers. Probable depression—using the standard Patient Health Questionnaire-nine screen for clinically significant symptoms—raised the odds of loneliness by about seventy-four percent. Emotional regulation difficulties, measured on a short form that taps things like impulse control and the ability to refocus attention under stress, also nudged risk up. The effect size per point was small, roughly four percent, but across the range, it adds up. Then there’s sleep. Not just “how did you sleep last night?” but “how is COVID affecting your sleep?” Poorer sleep in that specific frame pushed loneliness odds higher, around thirty percent. Anxiety and probable post-traumatic stress, interestingly, didn’t hold in the final model once everything else was in there. That doesn’t mean they don’t matter. It means their relationship with loneliness is tangled up with other pieces like depression and support. What about the policies themselves—self-isolation, quarantine? In the unadjusted look, you might expect a simple line: more isolation equals more loneliness. But when Groarke’s team put all the factors into the model together, the signal from self-isolation faded. In other words, the brute fact of being told to stay home wasn’t the decisive factor once you knew who was in your home, how supported you felt, how you were sleeping, and whether you were wrestling with depression. The context of isolation shaped the experience of it. There’s a pattern here that reaches back before COVID. Long before the virus, studies tied loneliness to social structure—living alone, unemployment, low income, limited community ties—and to mental health, especially depression. Other work found that loneliness can precede depression, suggesting a loop where each feeds the other. During prior outbreaks like SARS, both healthcare workers and the general public reported loneliness, which tells you this isn’t just about one profession or one setting. What’s striking in Groarke’s findings is how much the early lockdown reproduced those pre-existing contours. The pandemic didn’t rewrite the map so much as throw floodlights on the fault lines already there. Younger adults, who rely heavily on face-to-face socializing and were more likely to see work and education disrupted, were exposed. People with fewer or more fragile bonds were exposed. People who struggled to regulate emotion or to sleep under stress were exposed. Let’s talk briefly about the machinery under the hood—because how you analyze matters for what you conclude. The team started with simple comparisons: t-tests for continuous variables like support scores and chi-square tests for categorical ones like relationship status. Anything that showed at least a hint of association with loneliness—a p-value under zero point one—went into a backward stepwise logistic regression. Think of that as a tournament where variables compete for explanatory power; the least helpful get dropped, and the model refits until what remains best explains who is lonely and who is not. The output is an adjusted odds ratio for each surviving factor. Translate that: “given two otherwise similar people, how much does this one difference—age group, support level, depression status—shift the odds of loneliness?” It’s not causal proof, but it’s a sharper lens than a simple one-to-one comparison. Now, the caveats. This was a cross-sectional snapshot. It tells us who was lonely and what their lives looked like at that moment; it cannot tell us whether poor sleep caused loneliness or loneliness wrecked sleep. The sample came from the internet, not a door-to-door random draw, and skewed younger and female, which limits generalizability. All measures were self-reported—what you say about your feelings, your sleep, your symptoms—which can be biased by mood or memory. And it was early—those first intense weeks—so we don’t see how patterns evolved as people adapted. Even with those limits, the practical takeaways are clear. If you want to blunt loneliness under distancing, go where the risk concentrates. Younger adults should be near the front of the line, not because older adults are immune, but because the odds ratios for the young were the highest in this context. Strengthen perceived support in whatever safe ways are available—structured outreach, reliable check-ins, small groups with continuity—because each step up in support mattered. Help people sleep. That might sound soft, but in this model, COVID-related sleep problems were a distinct, sizable risk factor. And teach skills for emotion regulation—brief, evidence-based strategies to downshift arousal and broaden attention when stress narrows it. Those are not abstract ideas. They’re levers that, in the numbers from Groarke’s study, move loneliness. There’s also a quiet message about relationships. Living with a partner was one of the strongest protective factors. That doesn’t mean “get married” is a policy. It does mean that everyday companionship—someone in the next room, a shared meal, another adult presence—has measurable weight when the world shrinks. In settings where living alone is common, that’s a structural vulnerability. Community and workplace designs that build regular, dependable contact—even virtually, with small teams that stay intact—can mimic some of that buffering. If you zoom out, the story clicks into a broader frame. Before the pandemic, loneliness clustered around social and economic edges. During the pandemic, with face-to-face rituals stripped away, those same edges sharpened. As Groarke and colleagues reported in PLOS ONE, the distribution of loneliness under lockdown wasn’t random; it was patterned by age, relationship context, perceived support, mood, sleep, and emotional skills. That alignment is good news in a way. It means we’re not flying blind. The levers we knew about before COVID still work during it. One last thought on what’s next, and I’ll keep it brief. A cross-sectional snapshot can tell you where to look, but not how trajectories unfold. Following people over time—did improving sleep precede reductions in loneliness? Did gains in emotion regulation buffer against later lockdowns?—would sharpen the causal story. And widening the sample to better capture older adults and men would give a fuller view. But the absence of those longitudinal answers doesn’t blunt the present tense of this finding: in a moment of forced distance, about one in four adults in this UK sample felt lonely, and that risk was not random. It was higher for the young, lower for those with a partner or more adults at home, and woven tightly with depression, sleep, and the capacity to steer one’s emotions under stress. If you remember one number, make it twenty-seven percent. If you remember one pattern, make it this: connection protects, even when the doors are closed.