Mortality associations with overlapping PM2.5 pollution waves and cold events in Hebei Province, China: a county-level two-stage time-series study

@article{zhang2026mortality,
  title = {{Mortality associations with overlapping PM2.5 pollution waves and cold events in Hebei Province, China: a county-level two-stage time-series study}},
  author = {Zechen Zhang and Lina Yan and Yuqing Jin and Haoyang Chu and Ting Qi and Yuming Guo and Xiaoli Liu and Lei Yang},
  journal = {BMC Public Health},
  year = {2026},
  month = oct,
  doi = {10.1186/s12889-026-29816-6},
  url = {https://link.springer.com/10.1186/s12889-026-29816-6},
  note = {CC BY 4.0},
}

TY  - JOUR
TI  - Mortality associations with overlapping PM2.5 pollution waves and cold events in Hebei Province, China: a county-level two-stage time-series study
AU  - Zechen Zhang
AU  - Lina Yan
AU  - Yuqing Jin
AU  - Haoyang Chu
AU  - Ting Qi
AU  - Yuming Guo
AU  - Xiaoli Liu
AU  - Lei Yang
PY  - 2026
DA  - 2026/10/12
JO  - BMC Public Health
DO  - 10.1186/s12889-026-29816-6
UR  - https://link.springer.com/10.1186/s12889-026-29816-6
AB  - Background: PM 2.5 pollution waves and cold events may overlap during the cold season, but evidence on mortality associations during their actual temporal overlap remains limited.  Methods: We conducted a county-level two-stage time-series study using daily mortality, PM 2.5 , and meteorological data from 172 counties in Hebei Province during the cold seasons of 2014–2019. Pollution waves and cold events were defined as periods of at least two consecutive days exceeding county-specific PM 2.5 percentile thresholds or falling below county-specific temperature percentile thresholds, respectively. Joint exposure was restricted to days when the two event periods overlapped, beginning when the later event started and ending when either event ended. County-specific cumulative relative risks (RRs) over lag days 0–10 following a one-day change in the binary exposure indicator were estimated using quasi-Poisson distributed lag models and pooled using multivariate random-effects meta-analysis. A 95% prediction interval (PI) was additionally calculated for the primary pooled Joint estimate. The primary analysis evaluated non-accidental mortality under the P75T10 definition, which retained the greatest event support and county coverage after exposure redefinition.  Results: A total of 975,471 non-accidental deaths were recorded. Under the primary P75T10 definition, 113 counties contributed to the pooled analysis, of which 109 were located in central-southern Hebei. A one-day Joint exposure indicator was associated with an average pooled cumulative RR of 2.918 for non-accidental mortality over lag days 0–10 (95% CI: 2.151–3.958; 95% PI: 0.190–44.734; k = 113); the corresponding RRs were 1.144 (95% CI: 1.094–1.196) for pollution-wave-only days and 1.098 (95% CI: 1.053–1.144) for cold-event-only days. Joint RRs under P80T10 and P90T10 were also elevated. Estimates under the T05 definitions were less precise because only 8–16 counties contributed, and P90T05 was not estimable. The primary estimate was generally consistent across alternative lag-spline, temporal-trend, humidity, continuous-exposure, and spatial-covariance specifications, although it was attenuated when the maximum lag was extended to 27 days.  Conclusions: A one-day indicator of actual temporal overlap between PM₂.₅ pollution waves and cold events was associated with an elevated cumulative mortality risk over lag days 0–10. However, the primary pooled estimate was driven predominantly by counties in the central-southern plains and showed substantial between-county heterogeneity. These findings support further evaluation of geographically targeted, coordinated preparedness for overlapping air-pollution and cold events.
SN  - 1471-2458
ER  -