Headline takeaway
A systematic review and meta‑analysis published in 2020 found that, after accounting for prior health and key demographic factors, leaving work earlier (excluding disability pensions) was not associated with higher all‑cause mortality. Crude or insufficiently adjusted comparisons that do not control for prior health tend to overstate risks associated with retirement — a pattern likely driven by the healthy‑worker effect and reverse causation.
What the study asked and how it was done
The paper pooled evidence from longitudinal, peer‑reviewed studies to examine whether the timing of retirement (early retirement or on‑time retirement) versus continued work is associated with all‑cause mortality. The authors searched multiple databases in March 2018 and included 25 studies that met their criteria (22 from the database search and 3 from reference lists). They excluded exits from work due to disability pensions because health is an intrinsic determinant of that route out of employment. Most included studies were cohorts from Europe and the USA (with a few from Canada, Israel and Japan), and follow‑up in individual studies ranged from 2 to 31 years.
How the analysis was handled
The review combined a qualitative synthesis with meta‑regression to test moderators (for example, gender and prior health) and performed random‑effects meta‑analyses. Importantly, the authors separated studies by whether they adjusted for prior health status and demographic confounders (“fully adjusted”) or did not (“insufficiently adjusted”). Only 12 of the 25 included studies provided effect estimates that could be pooled quantitatively, so some pooled subgroups were small.
Key quantitative findings (selected pooled estimates)
– Early retirement versus working until statutory retirement (fully adjusted for prior health and demographics): pooled hazard ratio (HR) 1.05 (95% CI 0.87 to 1.28). This indicates no statistically significant association between early retirement and mortality once prior health and demographics are controlled for. Seven studies (nine associations) contributed to the early‑retirement meta‑analysis across subgroups.
– Early retirement (insufficiently adjusted): pooled HR 1.32 (95% CI 0.95 to 1.81). Although the point estimate is higher, it is not statistically significant and likely reflects confounding by health and demographic factors.
– On‑time retirement versus continuing work beyond retirement (insufficiently adjusted): pooled HR 1.56 (95% CI 1.41 to 1.73). This suggests higher mortality for on‑time retirees compared with those who keep working, but these estimates do not account for prior health and are prone to selection bias.
– On‑time retirement (adjusted for prior health): pooled HR 1.12 (95% CI 0.98 to 1.28). This adjusted estimate was not statistically significant; however, only a very small number of associations (two) contributed to this subgroup, so precision and confidence are limited.
How the authors interpret these results
When prior health and demographic confounders are taken into account, early retirement (excluding disability pensions) does not seem to increase mortality risk. The larger mortality differences seen in many unadjusted or insufficiently adjusted studies likely reflect the healthy‑worker effect (people who remain working tend to be healthier) and situations where declining health causes people to stop working, producing reverse causation. The authors emphasize that proper adjustment for prior health materially changes the conclusions drawn from study data.
Important caveats and limitations
– Heterogeneity: included studies used different definitions of early and on‑time retirement, sampled populations from different pension systems and countries, and employed varied study designs.
– Small pooled samples: only 12 studies provided quantitative estimates suitable for meta‑analysis; many subgroup pooled estimates were based on very few associations (sometimes only two to four), reducing certainty.
– Observational evidence: most studies were observational, so residual confounding and selection bias cannot be fully excluded and causal claims are not warranted.
– Outcome scope: the review focuses on mortality only and does not synthesise non‑mortality outcomes such as mental health, functional status, chronic disease incidence or wellbeing.
– Search cutoff: the literature search was completed in March 2018, so studies published after that date are not included.
Practical takeaways for readers and policymakers
– Don’t over‑interpret unadjusted comparisons: crude associations that show higher mortality among retirees often reflect underlying health differences rather than a causal effect of retirement.
– Context matters: pension systems, labour markets, social supports and healthcare access likely shape how retirement timing relates to health and mortality across countries and cohorts.
– Evidence is not definitive about causation: while adjusted analyses reduce bias and point away from a major harmful effect of early retirement on mortality, the observational nature of the evidence means we should be cautious about causal statements.
Citation
Sewdas R, de Wind A, Stenholm S, Coenen P, Louwerse I, Boot C, van der Beek A. Association between retirement and mortality: working longer, living longer? A systematic review and meta‑analysis. J Epidemiol Community Health. 2020 May 1. DOI: 10.1136/jech-2019-213023. (Search date cutoff for the review: March 2018.)
Leave a Reply