Short-term PM2.5 exposure and emergency admissions for sleep disorders: lag patterns in a time-series study at a psychiatric hospital in Shenzhen, China

Objective To evaluate the association between short-term ambient fine particulate matter (PM 2 . 5 ) exposure and sleep-disorder-related emergency admissions at a psychiatric hospital, characterize the short-term lag pattern, and explore potential differences by diagnostic category and sex. Methods We conducted a 120-day time-series study at a tertiary psychiatric hospital in Shenzhen, China, from February 1 to May 31, 2026. Sleep-disorder-related emergency admission records were identified using International Classification of Diseases, 10th Revision (ICD-10) codes G47 and F51. Quasi-Poisson regression combined with a distributed lag non-linear model (DLNM) framework was used to assess the short-term association between daily PM 2 . 5 concentrations and sleep-disorder-related emergency admissions over lags 0–7 days, adjusting for mean temperature, relative humidity, sunshine duration, day of the week, and calendar month. Cumulative associations over lags 0–2 days were additionally examined. Sensitivity and supplementary analyses included extending the maximum lag to 14 days, fitting negative binomial models, and estimating independent single-lag associations. Results A total of 538 eligible sleep-disorder-related diagnostic records associated with emergency admissions were included, with a mean of 4.5 ± 2.6 records per day. The mean age associated with these records was 37.5 ± 16.9 years, and 64.3% were female. The mean daily PM 2 . 5 concentration was 16.8 ± 6.7 μg/m³. In independent single-lag models, positive associations were observed primarily at lag 0 and lag 1. For each 10 μg/m³ increase in PM 2 . 5 , the relative risk (RR) was 1.340 (95% confidence interval [CI], 1.130–1.588) at lag 0 and 1.222 (95% CI, 1.026–1.456) at lag 1, whereas associations at lags 2–7 were not statistically significant. In the DLNM, the cumulative RR over lags 0–2 was 1.329 (95% CI, 1.068–1.652), whereas the cumulative RR over lags 0–7 was 1.472 (95% CI, 0.977–2.217). No statistically significant PM 2 . 5 -by-sex interaction was observed. Exploratory analyses suggested that G47 and F51 diagnoses may exhibit different lag patterns. Conclusions Short-term PM 2 . 5 exposure was associated with increased sleep-disorder-related emergency admissions, with the observed associations concentrated mainly in the early post-exposure period. Diagnostic subgroups may exhibit different lag patterns, although these findings require confirmation in larger, multicenter studies.