Early Mortality Spiked Sharply After SARS-CoV-2 Infection in Long-Term Care Residents
In Swedish long-term care facilities, SARS-CoV-2 infection carried a steep and immediate mortality penalty among very old residents, but that excess did not stay flat over time. In this matched cohort study based on the Swedish Senior Alert register, the death rate surged in the first days after documented infection, then fell progressively as follow-up continued. The pattern fits a familiar biological and clinical sequence in frail nursing home populations: the acute viral insult exposes severe physiologic reserve limitations, while survivors beyond the first weeks increasingly represent a healthier subset of the original cohort.
The study followed 3,604 residents with documented infection through 15 September 2020 and matched them to 3,604 uninfected controls using time-dependent propensity scores. Matching incorporated age, sex, health status, comorbidities, prescription medications, geographical region, and the timing of Senior Alert registration. A secondary analysis widened the control pool to 3,731 in each group by leaving out geographical region and registration time, which allowed longer follow-up and an estimate of median survival.
Median age in the cohort was 87 years, and 65% were women. That age profile alone places the population far from the middle-aged or community-dwelling cohorts often used to estimate COVID-19 burden. Residents in long-term care bring a different baseline: advanced frailty, multiple comorbidities, high medication burden, and a limited remaining life expectancy even without infection. Those features shape both the short-term lethality of SARS-CoV-2 and the interpretation of any “years of life lost” calculation.
The First Five Days Carried the Greatest Risk
The mortality curve rose quickly and then turned downward. Excess mortality peaked 5 days after documented SARS-CoV-2 infection, when the hazard ratio for death reached 21.5 (95% CI 15.9 to 29.2). In practical terms, the first days after diagnosis were the most dangerous window by far, consistent with severe viral pneumonia, abrupt decompensation of chronic illness, aspiration, delirium, or a terminal acceleration in residents with little physiologic reserve.
The danger was not distributed evenly across the follow-up period. After the initial peak, excess mortality diminished, and from the second month onward the mortality rate was lower among infected residents than among controls. During days 61 to 210 of follow-up, the hazard ratio for death was 0.76 (95% CI 0.62 to 0.93). This reversal does not imply protection from infection. Instead, it most likely reflects selection effects in a frail cohort: the sickest residents died during the acute phase, leaving behind a group with greater survival potential than the average control resident.
A sharp early hazard can coexist with a lower later death rate in the same cohort.
That pattern is central to understanding long-term outcomes in nursing home medicine. If the most vulnerable residents die early after infection, subsequent mortality comparisons can make the surviving infected group appear healthier than controls, even though the infection caused major acute harm. The biology is less mysterious than the statistics: a severe infection can act as a stress test that removes those with the least reserve, while the remaining survivors are a selected subgroup with better baseline resilience.
Median Survival Was Far Shorter Than National Life Expectancy
The secondary analysis estimated the median survival of uninfected controls at 1.6 years. That figure sits far below the national life expectancy at age 87 in Sweden, which was 5.05 years for men and 6.07 years for women. The gap matters because it places long-term care residence itself at the center of prognosis, independent of COVID-19 status. Residents in these facilities are not a generic elderly population; they are a highly selected group with substantial frailty and competing mortality risks.
This matters for how excess deaths are translated into public health metrics. If a resident already has a median survival measured in months to a few years, then a COVID-related death count cannot be interpreted in the same way as death in a healthier community-dwelling older adult. Years of life lost should be estimated with explicit recognition of baseline LTC status, otherwise the burden of infection may be overstated or mischaracterized when conventional life tables are used without adjustment.
The finding also carries a moral and operational implication for infection control in care homes. Preventing transmission in this setting protects a population that is already living near the edge of survival curves. Even when the later hazard declines, the early spike captures the clinical reality most relevant to bedside practice: these residents can deteriorate rapidly, and many will not have the reserve needed to survive the first week of infection.
Why the Survival Curve Turned Down After the Acute Phase
The most plausible explanation for the time-varying risk is depletion of the frailest residents early in follow-up. In geriatric epidemiology, this is a classic form of survivor selection. Those who die first are often the people with the highest burden of multimorbidity, the least robust cardiopulmonary reserve, or the most advanced neurocognitive impairment. Once they are removed from the risk set, the remaining infected residents may look comparatively stable when measured against matched controls who continue to include a broader mix of frailty levels.
The design strengthens that interpretation. Time-dependent propensity matching reduced confounding by aligning infected and uninfected residents on many baseline characteristics, and the use of the National Cause of Death Register gave a complete mortality endpoint. Still, observational follow-up cannot fully separate infection effects from the complex trajectories of very old residents in long-term care. Changes in staffing, testing practices, care intensity, transfer thresholds, and pandemic-era mortality patterns can all shape outcomes in ways that are difficult to fully capture in registry data.
Even with those caveats, the clinical message is direct. The acute phase after SARS-CoV-2 infection in a Swedish LTC resident was the period of greatest lethality, and the first five days stood out as the most dangerous. After two months, the mortality pattern shifted, not because the infection became harmless, but because the cohort itself had changed. The frailest residents were no longer in the follow-up pool.
Estimating Years of Life Lost Requires a Different Baseline
This study forces a more careful way of talking about COVID-19 burden in long-term care. Using population life expectancy for an 87-year-old Swede would substantially overestimate the expected survival time of many residents living in LTC facilities. The median survival of 1.6 years in the uninfected control group shows how different this setting is from the general older-adult population. Prognosis in long-term care is already limited before infection occurs.
For clinicians and policymakers, that distinction affects both communication and planning. Infection prevention in LTC remains essential because the early mortality hazard is enormous. At the same time, mortality metrics should reflect the frailty and baseline survival of the resident population, or the burden of COVID-19 will be misread at the level of years rather than deaths.
The study leaves one practical issue unresolved: how best to incorporate LTC residency into standardized models of excess mortality and life-years lost across future pandemics. That question matters because the risk curve in this cohort was not uniform, and any public health estimate that ignores that time-varying pattern will miss the way severe infection actually unfolds in frail institutionalized older adults.