Policy dosing in school physical education and adolescent fitness: a threshold-type association in a two-wave panel...

The minimal effective "dosage" of school physical education (PE) and its nonlinear relationship to adolescent fitness are not well defined. We developed a policy-based Education Policy Dosing Index (EPDI, 0-100) from four exam-oriented domains-compulsory test intensity, elective test diversity, scoring/enforcement strictness, and competition & incentive provision-aligned with Yunnan's 2019 education-exam reform and the junior-middle-school PE examination scheme. Multiple weighting checks (equal, ±20%, PCA/entropy) were used to confirm its formative nature.

We analyzed a two-wave school panel from a single public middle school in Kunming, China, yielding 2,673 student-occasion records, comprising the same 891 students measured at three time points (Grade 7 baseline, followed by two follow-ups in Grades 8-9; 2021-2023). Segmented regressions with two-way fixed effects (per 10 EPDI; reference threshold = 60) used CR2 inference. We profiled the breakpoint (τ), compared restricted cubic splines versus linear models, and conducted robustness checks (alternative weights, 10-fold delete-group jackknife, placebo threshold, negative-control outcome). Outcomes were sex-by-grade z-scores; component-specific Ns vary slightly due to test-day absence/invalid measures.

EPDI showed a threshold-like association with fitness; τ ≈ 66, with diagnostic breakpoints mostly 60-75 across outcomes. For every 10 EPDI units, composite PFI showed a pre-threshold slope of -0.074 (95% CI -0.116 to -0.032), slope change of +0.274 (0.151-0.397), and post-threshold slope of +0.201 (0.110-0.291). Component effects were stronger and directionally consistent: vital capacity +0.588 (0.501-0.675), 50-m sprint (higher z = slower) -0.207 (-0.272 to -0.143); standing long jump and sit-and-reach showed moderate gains; body mass index (BMI) was negligible. Girls benefited at lower EPDI levels, whereas boys and Grade 8 students gained more after the breakpoint. For the composite PFI, evidence of non-linearity was modest; signals were stronger at the component level.

An EPDI around 60 is a practical, exam-aligned planning benchmark, with diminishing marginal returns beyond the breakpoint and equity-relevant heterogeneity by sex and grade. Segmented and spline models outperformed linear models for most outcomes, and findings were robust across weighting/sensitivity checks. Estimates reflect one school over a short time period and should be replicated across schools/districts, accounting for maturation and socioeconomic markers. Associations are interpreted as non-causal.