The study, published in Scientific Reports, drew on five biennial waves of the Korean Longitudinal Study of Aging (KLoSA) between 2014 and 2022, a nationally representative panel survey of adults aged 45 and older. The research team, led by Gyeong-min Lee and Jae-hyun Kim of Dankook University together with colleagues at Gachon University and Seoul National University, set out to answer two questions with direct policy consequences. First, does the multidimensional Active Ageing Index (AAI) follow different trajectories across the retirement transition in metropolitan versus non-metropolitan areas? Second, do the individual characteristics that push retirees into ‘low active ageing’ differ by region?

The Active Ageing Index, built on standards proposed by the United Nations Economic Commission for Europe, is not a simple health score. In this Korean adaptation it combined three domains — health, participation and security — measured through eleven indicators. Health captured self-rated health, psychological distress on the CES-D10 scale, functional limitations in daily and instrumental activities, and chronic disease burden. Participation included labor force involvement, caregiving for family or grandchildren, volunteering and membership in social, religious or civic groups. Security rested on equivalized household income, pension status and homeownership. Each domain was normalized to a scale of zero to one using the United Nations Development Programme formula familiar from the Human Development Index, and the total AAI was their mean.

Because retirees are not a random cross-section of society, the team first had to make the regional comparison fair. From an initial pool of 8,238 respondents, they identified 4,801 people aged 55 or older who experienced retirement during the study period — 58.5 percent of those with complete key variables. Event-aligned analysis required an identifiable retirement wave and at least one AAI observation at the retirement wave and an adjacent time point, leaving 1,307 eligible retirees. Propensity score matching, using 1:1 nearest-neighbor matching with a caliper of 0.25, then balanced metropolitan residents of Seoul, Gyeonggi-do and Incheon against everyone else, producing a matched repeated-measures sample of 872 participants split evenly between the two groups. Metropolitan and non-metropolitan retirees, who differed in age, smoking and physical activity before matching, were statistically indistinguishable afterward.

The statistical machinery then turned to trajectories. Using piecewise generalized estimating equation models with region-by-time interactions, the researchers aligned each person’s data around their own retirement date: one wave before, the retirement wave itself, and up to two waves after. The results showed a clear dip at the transition itself. The pre-retirement time coefficient was negative and highly significant (β = −0.0925, p < 0.001), meaning total AAI fell from the wave before retirement to the retirement wave. After that, the index was stable: the post-retirement slope was essentially flat and not statistically significant (β = 0.0032, p = 0.315). Crucially, neither the main effect of metropolitan residence nor any region-by-time interaction approached significance, indicating that the shape of the trajectory was the same in the capital region and elsewhere.

But here the study delivers its most technically instructive twist. Labor force participation is one of the indicators inside the AAI’s participation domain — which means the index and the event of retirement overlap conceptually. Drop out of work, and your AAI drops partly by definition. In a sensitivity analysis that removed labor force participation from the participation domain, the retirement-wave decline in the revised total AAI shrank to statistical insignificance (β = −0.0076, p = 0.218). The apparent cliff, in other words, was largely a measurement artifact of the index’s construction. Yet something genuinely important survived: the revised participation index showed a significant negative post-retirement slope (β = −0.0196, p = 0.012), suggesting that non-work engagement — caregiving, volunteering, social groups — continues to erode after people leave employment. Retirement may not damage overall active ageing so much as slowly hollow out its participatory core.

The second analysis shifted from trajectories to risk. Among matched retirees whose AAI was at or above 0.5 at the retirement wave, the team used Cox proportional hazards models to examine who crossed below that midpoint threshold first. During follow-up, 170 of 553 participants transitioned to low active ageing while 383 were censored. Kaplan–Meier curves with log-rank tests showed no difference between regions (χ² = 0.443, p = 0.506), and metropolitan residence itself carried no elevated hazard (HR = 1.122, 95% CI 0.827–1.521, p = 0.460). Interactions between region and individual characteristics were likewise non-significant across the board.

Two individual factors did stand out with striking consistency. Retirees aged 75 and older faced a 70 percent higher hazard of transitioning to low active ageing than those aged 55 to 64 (HR = 1.698, 95% CI 1.080–2.669, p = 0.022). Regular physical activity, meanwhile, cut the hazard by roughly 30 percent (HR = 0.695, 95% CI 0.509–0.949, p = 0.022). The proportional hazards assumption held, verified with Schoenfeld residuals. In plain terms: getting older is dangerous for active ageing, and moving your body is protective — regardless of your postal code.

The authors are careful about what the null regional findings do and do not mean. The absence of significant interactions does not prove that place never matters for ageing well; it shows that the crude KLoSA metropolitan/non-metropolitan distinction failed to capture measurable heterogeneity in aligned AAI change. International research, including studies from China showing sharp east-versus-west differences and European sub-national AAI comparisons, suggests that finer-grained contextual features — local services, transport, community resources and social capital — may drive place-based differences that broad geographic labels obscure. The study’s event-aligned design also carries limits: KLoSA’s biennial rhythm means retirement timing is known only to the survey wave, only one pre-retirement observation exists, and the authors explicitly caution that changes around retirement should not be read as causal effects of retirement itself.

For a nation that crossed into super-aged status in 2025, with adults 65 and older now 20.3 percent of the population, the implications are pointed. Korea’s institutional peculiarities — mandatory retirement allowed at 60 while pension eligibility arrives later — create a vulnerable gap in which work exits, income security and daily routines must be renegotiated simultaneously. This study suggests that blanket regional targeting is the wrong lens for that gap. Instead, the risk of sliding into low active ageing concentrates among the oldest retirees and those who do not exercise regularly, and the quiet decay of social participation after work ends may be the signal worth watching. If retirement is a window of opportunity, as earlier reviews have proposed, these results indicate the window is best used not by moving to the city or fleeing it, but by staying active and staying connected.

Subject of Research: Active ageing trajectories across the retirement transition among Korean older adults

Article Title: Active ageing across the retirement transition in Korea: regional trajectories and risk of low active ageing

Article References: Lee, G.-M., Cho, H.-K., Hong, J.-W., Moon, J. Y., & Kim, J.-H. (2026). Active ageing across the retirement transition in Korea: regional trajectories and risk of low active ageing. Scientific Reports, 16(1), Article 32847. https://doi.org/10.1038/s41598-026-61165-7

Image Credits: AI Generated

DOI: 10.1038/s41598-026-61165-7

Keywords: active ageing, retirement, Active Ageing Index, KLoSA, South Korea, regional disparities, older adults, physical activity, longitudinal study, Cox proportional hazards, propensity score matching, social participation

The methodological ambition of the study is what sets it apart from earlier work. Most previous research on retirement and movement relied on questionnaires, which are notoriously vulnerable to recall bias and tend to miss the small, incidental movements that make up much of daily life. Device-based studies, by contrast, have almost all been conducted in Finland and Sweden, raising questions about whether their findings generalize to other cultures and environments. The Belgian team addressed both problems at once. Participants wore triaxial ActiGraph wGT3X-BT accelerometers on their non-dominant wrist for seven consecutive 24-hour cycles at four time points: one to six months before retirement, and again at three, six, and twelve months after. Wrist placement was a deliberate choice, since it yields the highest wear-time compliance and, crucially, allows sleep to be measured alongside waking behaviors, giving researchers a complete picture of the full day rather than fragments of it.

Raw acceleration data were processed with the open-source R package GGIR, which estimates time spent sleeping, sitting still, and moving at different intensities. Sedentary behavior was defined as acceleration below 44.8 milligravities, light physical activity between 44.8 and 100.6 milligravities, and moderate-to-vigorous activity above 100.6 milligravities. But the researchers went a step further, adding two cut-point-free metrics developed by Rowlands and colleagues: the intensity gradient, which describes how activity intensity is distributed across the day, and average acceleration, which captures total activity volume independent of any thresholds. Together, these measures allowed the team to detect shifts not just in how long people moved, but in how intensely they moved, a distinction that matters enormously for health outcomes in older adults.

The deepest technical innovation, however, lies in how the team handled a stubborn statistical problem. Sleep, sedentary behavior, and physical activity are not independent variables; every minute of the day belongs to exactly one of them. When sleep lengthens, something else must shrink. This perfect interdependence, known as multicollinearity, breaks traditional multivariate models. The solution was compositional data analysis, or CoDA, a framework borrowed from geology that treats the 24-hour day as a composition whose parts carry meaning only in relation to one another. The researchers converted their time-use data into isometric log-ratio coordinates using a sequential binary partition: the first coordinate separated active behaviors from passive ones, the second contrasted light activity with moderate-to-vigorous activity, and the third opposed sedentary time to sleep. These coordinates could then be fed into linear mixed models without violating any statistical assumptions.

Across the whole sample, the results painted a broadly optimistic picture. Before retirement, participants averaged 74 minutes of moderate-to-vigorous activity, 157 minutes of light activity, about 13 hours of sedentary time, and just over 7 hours of sleep per day. Twelve months after leaving work, moderate-to-vigorous activity had barely changed, sedentary time had fallen by roughly 16 minutes, and sleep had grown by about 15 minutes. The statistical models confirmed a significant shift in the balance between sedentary time and sleep, with moderate effect sizes, while the balance between active and passive behaviors remained stable. Notably, the timing of these changes was concentrated in the first three months after retirement; after that, the new patterns settled in and remained remarkably stable through the one-year mark. The intensity gradient and average acceleration showed no significant change, confirming that the reallocation happened on the inactive side of the day rather than in the structure of physical activity itself.

The sleep findings align neatly with earlier device-based studies from the Nordic countries, which also documented longer sleep after retirement. Evidence from both quantitative and qualitative work suggests the mechanism is simple: retirees postpone both their bedtime and, especially, their wake-up time, and the later awakening is what extends total sleep. In this study, average sleep rose from 7 hours and 13 minutes to 7 hours and 28 minutes, still comfortably within the 7-to-9-hour range recommended by major sleep societies. The sedentary result is more surprising. Previous device-based studies generally found only small reductions in sitting time, and one Finnish analysis even reported an increase of 24 minutes per day. The Belgian team suggests their sample, which included proportionally more men in non-manual occupations, may explain the difference, since office workers lose large blocks of occupational sitting when they retire.

The most consequential findings, though, emerged when the researchers stratified the data by socio-economic position, measured through three separate indicators: education, occupation, and income. Education and occupation showed only descriptive, non-significant trends toward more favorable changes among higher-status groups. Income told a different story. Retirees in the higher income bracket increased both moderate-to-vigorous activity by 7 minutes and light activity by 6 minutes per day while cutting sedentary time by 28 minutes in the first three months after retirement. The lower income group did the opposite, losing 7 minutes of moderate-to-vigorous activity and 10 minutes of light activity while their sedentary time stayed flat. By six months, the higher income group had shifted toward more intense activity while the lower income group had shifted toward less. By twelve months, the higher income group had cut sitting time by 31 minutes and added 20 minutes of sleep, while the lower income group managed only a 1-minute reduction in sitting and a 6-minute gain in sleep.

Why would income, rather than education or occupation, be the decisive factor? The researchers point out that income is recognized in the gerontological literature as the socio-economic indicator most strongly associated with late-life health, because it reflects accumulated resources across the entire life course. Retirement typically reduces financial resources, and rising awareness of the cost of activities has been documented as a barrier to physical activity in qualitative studies of new retirees. There is also the well-known physical activity paradox to consider: occupational activity, which lower socio-economic workers perform in abundance, carries different and often less favorable health implications than recreational activity. What happens to health when a physically demanding job is replaced by discretionary movement is, remarkably, a question that has not yet been directly investigated, and this study adds urgency to it.

The authors are candid about their limitations. The sample of 96 participants is small, limiting statistical power and generalizability, and manual workers were underrepresented, likely reflecting recruitment through channels that attract health-conscious volunteers. Wrist-worn accelerometers estimate inactivity rather than true sedentary behavior, which by definition requires a sitting or reclining posture that the device cannot verify, and the applied intensity cut-points were validated in younger adults. Still, the study’s strengths, multiple socio-economic indicators, objective 24-hour measurement, four time points, and compositional statistics, make its message hard to ignore. Retirement is a natural window of opportunity when daily routines dissolve and new habits can take root, and for wealthier retirees that window opens onto healthier lives. For lower-income retirees, it may instead mark the beginning of a downward drift. The authors argue that targeted lifestyle interventions during the retirement transition are needed precisely for these groups, who already face elevated risks of chronic disease. As populations across Europe and beyond age, ensuring that the health dividend of retirement is shared equally may become one of the defining public health challenges of the coming decades.

Subject of Research: Socio-economic differences in changes in device-measured physical activity, sedentary behavior and sleep during the retirement transition

Article Title: Socio-economic position and changes in 24-h movement behaviors during the retirement transition

Article References: Vansweevelt, N., van Uffelen, J., Boen, F., Suorsa, K., & Seghers, J. (2025). Socio-economic position and changes in 24-h movement behaviors during the retirement transition. Journal of Activity, Sedentary and Sleep Behaviors, 4(1), Article 17. https://doi.org/10.1186/s44167-025-00087-7

Image Credits: AI Generated

DOI: 10.1186/s44167-025-00087-7

Keywords: retirement, physical activity, sedentary behavior, sleep, socio-economic position, accelerometry, compositional data analysis, 24-h movement behaviors, aging, health inequalities, longitudinal study, Belgium