Effects of sources of social support and resilience on the mental health of different age groups during the COVID-19 pandemic

Picture a 70-year-old man, three weeks into lockdown. His adult children are calling every day. His doctor is checking in. On paper, he has support. Now put a 25-year-old woman next to him — same pandemic, same city, same restrictions, and same official guidance to stay home. She has a wider social network, more digital fluency, and more years ahead of her. Ask yourself: who is doing better mentally? The answer that Li and colleagues found, in a survey of 23,192 people, runs against what most of us would assume. They had to build an entirely new way of grouping people before they could see why. The COVID-19 pandemic created an unusually potent mix of biological risk and social disruption. Contact tracing, isolation, and quarantine slowed transmission, but they also dismantled the routines that hold people together — work rhythms, casual encounters, and physical proximity to the people who matter most. Li and colleagues describe the result as "heavy psychological and emotional burdens for the general population." Uncertainty, financial pressure, enforced separation, and fear of infection all arrived at once. Social support and psychological resilience became the two variables that would determine, in large part, how people coped. Why focus on age? The mortality data gave researchers good reason to. Li et al. cite figures from Nature Medicine showing that compared to people aged 30 to 59, those over 59 were five point one times as likely to die after developing symptoms. That gradient pushed clinicians and researchers to treat older adults as the most vulnerable group. But there were reasons to complicate that assumption. Lifespan research had long documented an "aging paradox" — the observation that emotional well-being often stays stable or even improves as people get older. Socioemotional selectivity theory, or SST, offers one explanation: as people age and become more aware of time's limits, they narrow their social networks and concentrate on emotionally meaningful close relationships rather than broad social circles. So older adults might face higher objective danger from the virus while simultaneously drawing on a more curated, high-quality set of social ties. The pandemic offered a rare natural experiment for testing which force won out. The researchers' first methodological decision was the one that made everything else possible. Instead of averaging each person's social support into a single number, Li et al. asked participants to rate support from five distinct sources — family, friends or small groups, communities, organizations and institutions, and society as a whole — each on a scale of one to five. Then they used latent profile analysis, or LPA, a person-centered technique that groups people by their pattern of responses across all five sources rather than by a single mean. The idea is that two people can have the same average support score while having completely different support configurations: one drawing heavily on family, the other on institutions. LPA surfaces those configurations. Fitting models with two through six profiles, the team selected a five-profile solution based on statistical fit criteria and whether each additional profile added a qualitatively new pattern. The five groups that emerged across the total sample are vivid in their differences. Class one, sixteen percent of the sample, was low across every source — family support averaging around one point nine, friends around one point seven, and society around one point eight. Class two, about seventeen percent, showed high family support but low scores elsewhere — family near four point zero, communities and organizations below two point zero. Class three, the largest group at nearly thirty-six percent, had lower family support but higher scores on non-family sources — a profile Li et al. call "predominantly remote support." Class four, around twenty-one percent, was moderate across the board. And class five, just under ten percent, scored near the ceiling on everything — family at four point eighty-eight, communities and society both near four point ninety-five. High support everywhere, simultaneously. The distribution of people across these five profiles was not random with respect to age. Chi-square testing confirmed a significant difference across the three cohorts — emerging adults aged 18 to 25, adults aged 26 to 59, and older adults aged 60 and above. Older adults were most likely to fall into the family-focused class two: twenty-two point one percent of them, compared to sixteen point four percent of emerging adults. Emerging adults, meanwhile, were heavily concentrated in the remote-support class three — forty point two percent of them versus thirty-five point three percent of older adults. The smallest share of older adults landed in the high-support class five, just seven point six percent, compared to eleven point five percent of emerging adults. This is SST in the data. Older adults cluster in a profile built around close family ties, while younger adults spread across a wider, less emotionally concentrated network of sources. The theory predicts exactly this narrowing. What makes it striking is what happens to mental health as a result. Li et al. measured mental health using the Mental Health Inventory, a five-item scale scored from one to six, with the total sample averaging five point zero one. An analysis of variance comparing the three age groups on mental health was significant, and post-hoc tests showed older adults reporting higher mental health than both emerging adults and working-age adults. The mortality risk was higher. The social network was narrower. And yet the mental health scores were better. That is the aging paradox, confirmed quantitatively in a sample of over 23,000 people during the worst public health crisis in a generation. Support profiles also predicted mental health directly. Mean scores rose from class one, where the average was four point ninety-one, to class five, where it reached five point forty-four. For working-age adults specifically, the family-focused profile yielded higher mental health than the remote-support profile — suggesting that for adults in midlife, proximal family ties carry more protective weight than institutional or community sources. For older adults, the bottom three profiles did not differ significantly from each other in mental health outcomes, which hints that for this group, even modest support concentrates in ways that provide a floor. Now bring in the second resource: resilience. Li et al. measured it with an abbreviated two-item Connor-Davidson Resilience Scale — "Able to adapt to change" and "Tend to bounce back after illness or hardship" — rated from one to five, with a sample mean of three point seven six. Resilience correlated with mental health at zero point forty. Higher resilience, better mental health. That part is intuitive. The more important finding is what happens when resilience is low. Using Hayes' PROCESS to test interactions, the team found that both the moderate-support and high-support profiles buffered the negative impact of low resilience on mental health. For the moderate-support profile, that interaction term was small but significant. For the high-support profile, it was substantially larger — a coefficient of negative zero point twenty-four. In plain terms: being in the high-support profile compensated meaningfully for having low internal resilience resources. Social connection stood in for an inner capacity when that capacity was depleted. The researchers also tested whether age changed this buffering pattern — a three-way interaction of age, resilience, and social support. The interaction was statistically significant, but when they ran simple-slope tests within each age group, none of the individual slopes reached significance. The buffering effect of social support on low resilience appears to operate similarly across age groups, even if it does not operate identically. Li et al. are careful about the study's boundaries. The data are cross-sectional — a snapshot, not a trajectory. Social support was measured with single-item ratings for each source rather than validated multi-item scales. The data were collected between March twenty-fifth and April first, 2020, when China's outbreak was largely under control, which may have shaped the emotional baseline. These are real constraints on what the findings can claim causally. Still, the architecture of the results points somewhere useful. The five support profiles are reproducible across age groups and predict mental health in a consistent direction. When inner resilience is low, broad, high-quality social support compensates. And older adults — the group we worried about most — have quietly organized their social lives in a way that provides exactly the kind of support that matters: close, emotionally meaningful, family-centered. Go back to the 70-year-old and the 25-year-old. The older man's narrower network turns out to be the right shape for weathering a crisis. The younger woman's wider, more diffuse connections offer less of the specific protection that counts. Vulnerability is not just a function of exposure to danger. It is a function of what your connections look like when danger arrives. That is what twenty-three thousand surveys, one pandemic, and a careful statistical model had to reveal before we could see it clearly. 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.