NHS Shielding Letters Reveal the True Mental Health Cost of Lockdown Isolation for Older Adults

When the UK government told its most clinically vulnerable citizens to stay home in the spring of 2020, it also handed social scientists something rare: a natural experiment in forced isolation. A new study published in SSM – Population Health exploits that experiment to answer a question that has frustrated researchers for decades—does social isolation actually cause depression and anxiety in older adults, or do people with poor mental health simply withdraw from the world? The answer, according to economists Cecilia Rodríguez Ingold and Daniel Cho, is that isolation itself measurably worsens the mental health of people over fifty, and the damage scales with how strictly and how often they isolate.

The difficulty with nearly all previous research on this topic is a statistical trap known as reverse causality. Isolation is associated with worse mental health, but people experiencing depression or anxiety also tend to pull away from their social networks, making it impossible to tell which direction the arrow points. Most studies, as a 2017 literature review by Courtin and Knapp highlighted, document correlations rather than causal effects. To break this deadlock, the authors turned to an instrumental variable approach, a technique borrowed from econometrics that uses a source of variation unrelated to the outcome itself to isolate a genuine causal effect.

The instrument they chose was the UK’s shielding policy. During the pandemic, the National Health Service and general practitioners contacted people identified as clinically vulnerable—everyone aged seventy or older, plus younger adults with conditions on the annual flu vaccine list—advising them to stay at home and avoid all face-to-face contact. The government backed this advice with practical support, including weekly food parcel deliveries direct to the doorstep and pharmacy-arranged medication delivery, designed specifically to make isolation logistically feasible. Because this contact strongly predicted whether a person self-isolated but was not itself intended to influence mental health, it provides what statisticians call a plausible exogenous shock: a nudge toward isolation that operates independently of a person’s psychological state.

The data come from the English Longitudinal Study of Ageing, a nationally representative panel study of people aged fifty and older that has run roughly every two years since 2002. The research team combined demographic and health information from Wave 9, collected in 2018 and 2019 before the pandemic, with two special COVID-19 substudy waves fielded in June and July 2020 and again in November and December 2020. After restricting the sample to people who responded in both pandemic waves, 4,725 individuals remained, with an average age of about seventy. Roughly 56 percent were women, three-quarters had a partner, and about 5 percent had lost a relative or close friend to COVID-19.

Mental health was measured with clinically validated instruments rather than ad hoc questions. Anxiety was assessed with the seven-item General Anxiety Disorder scale, or GAD-7, which ranges from zero to twenty-one and showed strong internal consistency in the sample. Depression was measured with an adapted eight-item version of the Center for Epidemiologic Studies Depression scale, scored from zero to seven, with one item inadvertently omitted from the pandemic survey handled by computing a comparable seven-item score for the pre-pandemic wave. Social isolation was defined strictly: self-isolation meant not leaving the home for any reason, not going out even for food, and not seeing anyone outside the household during the week before the interview.

The headline results are striking in their magnitude. Applying two-stage least squares estimation to the pooled sample of 9,450 person-wave observations, the authors found that self-isolation raised depression scores by 0.438 points, equivalent to 0.36 standard deviations, and anxiety scores by 2.294 points, equivalent to 0.56 standard deviations. Both effects were statistically significant, and both were substantially larger than the corresponding ordinary least squares estimates, which the authors attribute to a combination of measurement error, reverse causality, and the fact that the instrumental variable identifies a Local Average Treatment Effect for a specific group: compliers, meaning older adults whose isolation behavior actually changed in response to NHS contact. Because these individuals were targeted by policy guidance, they represent precisely the population that public health interventions can reach, making the estimate directly relevant for policy design.

The authors went further, constructing an isolation intensity measure that summed each person’s behavior across three time points—April 2020, recalled retrospectively; June 2020; and November 2020—scoring zero points for no isolation, one point for staying at home, and two points for full self-isolation in each period, yielding a scale from zero to six. Using the aggregated NHS contact across both lockdowns as an instrument, they found that each one-category increase in isolation intensity raised depression scores by 0.106 points and anxiety scores by 0.580 points. Moving from never isolating to self-isolating in all three periods would worsen depression by more than half a standard deviation and anxiety by roughly 0.82 of a standard deviation. Notably, the relationship did not differ significantly between the two pandemic waves, suggesting the mental health cost of isolation persisted across both the harsher first lockdown and the more permissive second one.

Robustness checks strengthened the case. Specifications without controls produced larger estimates, indicating that failing to account for demographics, health conditions, and pandemic-related stressors overstates the raw association. Alternative definitions of isolation intensity, including one proposed by Di Gessa and Price, preserved the significance of the results. When the authors relaxed the strict exclusion restriction using the imperfect-instrument bounds of Nevo and Rosen, the lower bounds remained above zero, meaning the detrimental effect of isolation on both depression and anxiety survives even if the instrument has some direct influence on mental health. The Durbin-Wu-Hausman test rejected the null of exogeneity in all two-stage models, confirming that the instrumental approach was warranted in the first place.

Perhaps the most revealing part of the study is its exploration of the mechanisms through which isolation inflicts its psychological toll. The authors tested four candidate pathways: loneliness, health behaviors, access to resources, and use of social networks. Social isolation increased loneliness substantially, consistent with prior evidence that loneliness predicts later depression in older adults. It also reduced the likelihood of drinking alcohol, plausibly because the social gatherings where drinking typically occurs disappeared. Worrying about access to essential items and food rose among those isolating, though actual food access did not deteriorate, likely reflecting the effectiveness of the government’s delivery support. Meanwhile, isolated individuals increased text-based contact with friends and family, suggesting a partial digital substitution for in-person connection whose mental health value remains an open question for future research.

The authors are careful about the limits of their findings. NHS contact was targeted by age and clinical vulnerability, so even after controlling for a broad set of physical health conditions, lagged mental health, financial circumstances, and employment status, the causal interpretation rests on the assumption that no unobserved factor correlated with both the instrument and mental health outcomes. Only two pandemic waves were available, limiting the ability to assess longer-term consequences, and the anxiety measure was not fully consistent across waves. Still, the study delivers a clear message for aging societies: as populations grow older and technology reduces the need for in-person interaction, policies that isolate vulnerable people for legitimate public health reasons carry a measurable psychological price. The authors argue that interventions targeting the mental health of older adults are long overdue, and that future isolation policies should be designed from the outset with strategies to mitigate the loneliness and anxiety they predictably produce.

Subject of Research: Causal effect of social isolation on depression and anxiety in older adults during the UK COVID-19 lockdown

Article Title: The Effect of Social Isolation on the Mental Health of Older Adults: Evidence from the UK COVID-19 Lockdown

Article References: Ingold, C. R., & Cho, D. (2026). The Effect of Social Isolation on the Mental Health of Older Adults: Evidence from the UK COVID-19 Lockdown. SSM – Population Health, Article 101987. https://doi.org/10.1016/j.ssmph.2026.101987

Image Credits: AI Generated

DOI: 10.1016/j.ssmph.2026.101987

Keywords: social isolation, older adults, mental health, COVID-19 lockdown, depression, anxiety, instrumental variables, ELSA, shielding policy, loneliness, UK, public health