Abstract
<title>Abstract</title> <p>Background The Job Demand-Control-Support (JDC-S) model predicts that low decision latitude, high demand, and poor support raise musculoskeletal disorder (MSD) risk. Most evidence comes from single-timepoint surveys of wage-employed Western populations. Where self-employment is common, decision latitude may be bound up with occupation and employment type. We examined whether JDC-S holds in South Korea, where about a quarter of workers are self-employed. Methods We pooled six repeated cross-sectional waves of the Korean Working Conditions Survey (2010–2023; analytic sample 119,912 respondents from a pooled raw sample of 260,996; 83,288 eligible for regression). With directed-acyclic-graph-informed adjustment for occupation, employment type, and ergonomic exposures, we estimated MSD prevalence ratios (PRs) by modified Poisson regression, assessed autonomy-by-exposure interaction, and ran industry-stratified and sensitivity analyses. Results Between 2010 and 2023 the survey-weighted prevalence of low decision latitude nearly doubled (34% to 64%), while long working hours fell from 46% to 15% (unweighted). Contrary to the model, lower decision latitude was associated with lower MSD prevalence (any MSD PR 0.84, 95% CI 0.83–0.86); high demand showed the expected direction, whereas the social-support association was weak and unstable across sensitivity analyses. Quadrant and industry-stratified analyses showed a 30-percentage-point gradient: high-autonomy, high-exposure workers (predominantly self-employed) had the highest MSD prevalence (51.9%), low-autonomy, low-exposure workers the lowest (21.9%). The paradox held across body sites, sexes, industries, and sensitivity analyses. Conclusions In a labour market with substantial self-employment, this autonomy paradox is not explained by occupational composition: the association is unchanged when industry and employment type are removed and persists within industries. We interpret it as a non-causal signal, not a protective effect of low control: it reflects the decision-latitude measure, healthy-worker selection, and reverse causation, mechanisms a cross-sectional design cannot disentangle. JDC-S exposures should be reported separately for self-employed and wage-employed workers.</p>