The first two years of the COVID-19 pandemic claimed roughly 13,120 more lives than expected among older people in the Netherlands, according to a nationwide analysis that for the first time weaves infection, vaccination and mortality into a single statistical model. The study, published in the European Journal of Epidemiology by a team at Leiden University Medical Center, followed all 3,826,770 residents of the Netherlands born before 1957 — everyone aged 63 or older on 1 January 2020 — using anonymized individual-level data from Statistics Netherlands. Between 2020 and 2021, absolute excess mortality in this population was 0.34 percent, meaning that 4.41 percent of all deaths observed during those two years were above what would have been expected under normal circumstances. The figure sounds small at the population level, but the researchers stress that it conceals enormous disparities: excess mortality ranged from just 0.06 percent in men aged 63 to 64 to a staggering 2.82 percent in men aged 90 and over, and every sex and age group in the cohort was affected.
What distinguishes the new work from earlier excess-mortality studies is its methodological machinery. Rather than simply comparing death counts before and during the pandemic, the team split observed mortality into a background component — the deaths that would have occurred anyway — and an excess component attributable, directly or indirectly, to the pandemic. To do this, they used relative survival modelling, a framework originally developed in cancer epidemiology, in which each person’s hazard of dying is decomposed into a population hazard derived from historical mortality rates and an excess hazard that can be interpreted as the counterfactual impact of COVID-19. The background hazard was estimated from mortality in 2015–2019, stratified not only by sex and age but also by month of the year, a refinement that captures seasonal fluctuations in death rates driven by influenza seasons and temperature extremes. The choice of the five pre-pandemic reference years was deliberate: it averages out unusually mild and severe influenza seasons and decades-long trends of improving life expectancy, giving a fair counterfactual baseline against which the pandemic years could be measured.
The second pillar of the analysis was a multi-state model, a statistical architecture in which individuals move through a series of states over time — here, being alive and event-free, recently infected, no-longer-recently infected, vaccinated, various combinations of infection and vaccination, and, ultimately, death. Because the Netherlands only declared COVID-19 a notifiable disease with widespread testing capacity from June 2020, the model begins on 1 June 2020, just after the first wave. Every transition into a death state was split into background and excess death using the relative survival framework, following a recently developed non-parametric approach by Manevski and colleagues. Because the risk of dying after infection depends strongly on how much time has passed since the positive test, the classical Markov assumption of the model — that transitions depend only on the current state, not on how long one has been in it — was violated for infection states. The researchers therefore employed landmark Aalen-Johansen analyses, repeatedly re-fitting the model from defined starting points, such as 1 January 2021, 1 June 2021, 1 September 2021 and 1 December 2021, and from each state occupied at those moments. Infection was defined as a positive test after at least 90 days without one, and the “acute” phase was set at the first 28 days after a positive test, based on evidence that the fatality risk declines with time since infection.
The results map the pandemic’s deadly rhythm with unusual precision. Before mid-March 2020, fewer people died than expected — a negative excess mortality hazard, partly reflecting the remarkably mild influenza season of early 2020. During the first wave, from March to June 2020, excess deaths surged; after the wave subsided, the hazard turned negative again before spiking once more with the second wave in late 2020 and early 2021 and again with the third wave at the end of 2021. For the full study period, absolute excess mortality was 0.34 percent overall: 0.19 percent in 2020 and 0.16 percent in 2021. Men fared worse than women on the absolute scale in every age group except the youngest, and risk rose steeply with age. Yet on the relative scale — excess deaths as a share of observed deaths — the age gradient became much flatter, particularly among women. Women aged 63 to 64 experienced excess mortality amounting to 4.25 percent of their observed deaths, the highest relative figure in the entire cohort, whereas women aged 90 and over, despite suffering the greatest absolute toll, had the lowest relative excess mortality at 2.25 percent. This dissociation highlights how an older group with high baseline mortality can absorb a large absolute burden while a younger group experiences a proportionally larger disruption.
The multi-state model revealed where and when those excess deaths occurred. By the end of 2021, the cumulative incidence of documented infection in the cohort was 9.42 percent and of vaccination 89.13 percent. The vast majority of excess deaths occurred during the acute phase — the first four weeks after a positive test. For the whole cohort, the probability of dying an excess death during a recent documented infection reached 0.40 percent, accounting for 96.07 percent of all deaths from that state. Critically, excess mortality did not stop at 28 days: people who had passed the acute phase of an unvaccinated infection still died more often than expected, with a residual excess death probability of 0.06 percent, or 55.23 percent of deaths following non-recent infection without prior vaccination. This aligns with international findings, including a study of US veterans showing elevated mortality up to 180 days after infection and cohort data from Hong Kong and the UK demonstrating higher all-cause death risk among COVID-19 survivors at 21, 30 and 90 days post-infection. The most extreme burden fell on men aged 90 and older with a recent unvaccinated infection, whose excess mortality reached 2.85 percent, equivalent to 95.91 percent of all deaths observed during that phase.
Vaccination emerged as a decisive protective factor across every subgroup. The Dutch campaign, launched in January 2021 and rolled out from the oldest to the youngest, had reached all members of the cohort by May 2021. By 1 June 2021, the model estimated that 81.52 percent of the cohort had been vaccinated. Landmark analyses quantified the benefit starkly: from 1 January 2021, during the second wave, people with a recent documented infection faced a 4.61 percent probability of excess death by the end of the year, while those with a non-recent infection faced 0.82 percent. By contrast, people in the vaccination state showed negative excess mortality — fewer deaths than expected. When the third wave hit in late 2021, the contrast remained: those with a recent unvaccinated infection at 1 September or 1 December 2021 had excess mortality probabilities of 5.14 percent and 5.20 percent respectively, while those with a recent infection after vaccination faced markedly lower figures of 2.46 percent and 1.82 percent. Even when vaccinated people did become infected, their excess mortality was far lower than that of unvaccinated infected individuals — 0.16 percent during recent infection and 0.01 percent beyond it. A sensitivity analysis delaying each person’s vaccination status by two weeks, to allow for immune build-up, produced virtually identical results.
The findings also illuminate subtler dynamics. Vaccination was associated with a lower probability of documented infection itself, though this effect was weaker than the effect on death, consistent with evidence that SARS-CoV-2 vaccines perform best against severe outcomes. Notably, the team observed excess mortality even among people with no documented infection or vaccination — a state that likely mixes undiagnosed infections, indirect pandemic effects on other diseases, and the so-called healthy vaccinee bias, whereby those who get vaccinated tend to be healthier than those who do not. Because vaccination was voluntary and required informed consent for data sharing, an estimated 8 to 25 percent of people classified as unvaccinated had actually received a shot but withheld consent, a misclassification that would attenuate rather than inflate the observed vaccine benefit. The authors argue that the persistence of the gap between vaccinated and unvaccinated infected individuals at every landmark time — without shrinking over the course of the pandemic — strongly indicates a genuine protective effect rather than a selection artefact. Seroprevalence studies suggest many mild infections went undocumented: cumulative documented infection in the cohort was only about 9 percent by the end of 2021, while serosurveys pointed to infection rates near 20 percent in those aged 60 and over, meaning the true population-wide impact of infection was likely larger than the documented figures suggest.
Beyond its findings on COVID-19, the study is a proof of principle for a new analytical toolkit. The authors are, to their knowledge, the first to apply the combination of relative survival and multi-state modelling to compare the same population across different time periods — crisis versus normal — rather than comparing two populations simultaneously, as in cancer research. Because the framework does not require cause-of-death information, it captures both the direct toll of the virus and indirect effects of pandemic measures, and because it runs on a continuous timescale, it can track the ebb and flow of risk in fine detail. Its reliance on background mortality rates also makes absolute excess mortality directly comparable across countries. All analysis code, built in R using the mstate and relsurv packages, is publicly available on GitHub, and the authors explicitly frame the method as a template for assessing mortality during future crises, incorporating crisis-specific intermediate events such as vaccination. The work was funded by ZonMw, the Dutch health research funding agency.
The researchers conclude that infection clearly increased the risk of excess death while vaccination clearly decreased it, with absolute risk differences growing with age — a pattern they say vindicates the decision to vaccinate the oldest first. Yet their central message transcends the Dutch context: all groups experienced excess mortality, concentrated in the acute phase of infection but lingering beyond it, and every subgroup derived a clear survival benefit from vaccination. As societies debate pandemic preparedness, the study offers both a sobering quantification of COVID-19’s true toll among the elderly — deaths that would never appear in official COVID-19 counts, since it deliberately avoided cause-of-death data — and a rigorous demonstration that vaccination remains one of the most powerful tools for reducing mortality in a pandemic.
Subject of Research: Dynamics of SARS-CoV-2 infection, COVID-19 vaccination and all-cause excess mortality among people aged over 63 in the Netherlands during 2020–2021, analyzed with a nationwide multi-state relative survival model
Subject of Research: Medicine
Article Title: Dynamics of infection, vaccination and excess mortality during the COVID-19 pandemic among older individuals—a nationwide analysis
Article References: Koster, E. A. S., Sluiskes, M. H., Putter, H., Rosendaal, F. R., van Hylckama Vlieg, A., de Boer, M. G. J., & de Wreede, L. C. (2026). Dynamics of infection, vaccination and excess mortality during the COVID-19 pandemic among older individuals—a nationwide analysis. European Journal of Epidemiology. https://doi.org/10.1007/s10654-026-01414-1
Image Credits: AI Generated
DOI: 10.1007/s10654-026-01414-1
Keywords: COVID-19 pandemic, excess mortality, vaccination, relative survival, multi-state modelling, SARS-CoV-2 infection, older adults, Netherlands, nationwide cohort, background mortality, landmark analysis, epidemiology
Cite Scienmag News
APA
MLA
Chicago
Kristina Jarvis. (September 5, 2026). Nationwide study tracks infection, vaccination and excess deaths in older adults during COVID-19. Scienmag. https://scienmag.com/nationwide-study-tracks-infection-vaccination-and-excess-deaths-in-older-adults-during-covid-19/
Kristina Jarvis. “Nationwide study tracks infection, vaccination and excess deaths in older adults during COVID-19.” Scienmag, 5 September 2026, https://scienmag.com/nationwide-study-tracks-infection-vaccination-and-excess-deaths-in-older-adults-during-covid-19/. Accessed 5 September 2026.
Kristina Jarvis. “Nationwide study tracks infection, vaccination and excess deaths in older adults during COVID-19.” Scienmag. September 5, 2026. https://scienmag.com/nationwide-study-tracks-infection-vaccination-and-excess-deaths-in-older-adults-during-covid-19/
Copy citation
Download RIS
Tags: age-specific COVID-19 death disparitiesage-specific COVID-19 mortality disparitiesanalysis of COVID-19 mortality trends in older adultsCOVID-19 excess mortality in older adults in the NetherlandsCOVID-19 infection rates among seniorsCOVID-19 pandemic impact on older adultsdemographic differences in COVID-19 outcomes among seniorsexcess death rates among elderly during COVID-19infection and vaccination impact on elderly populationslong-term effects of COVID-19 on elderly mortalitylong-term effects of COVID-19 on seniorsmethodological advancements in COVID-19mortality analysis using individual-level datanationwide epidemiological study on COVID-19 mortalitypublic health insights into COVID-19 impactsex and age disparities in COVID-19 mortalitystatistical modeling of COVID-19 deaths and vaccinationstatistical modeling of pandemic-related mortalityvaccination effectiveness in reducing excess deaths
