Time-stratified case-crossover dataset of multi-pollutant air pollution, meteorology, and mortality in Thailand, 2014–2022

This dataset provides an analysis-ready time-stratified case-crossover table linking short-term exposure to multiple air pollutants, model-predicted PM0.1, and meteorological conditions with individual mortality events in selected subdistricts of Thailand. Study areas are located within 60 km of Thailand Pollution Control Department air quality monitoring stations 13T, 20T, 35T, 36T, 44T, and 59T, and meteorological conditions are aligned using nearby Thai Meteorological Department Automatic Weather Stations, which together were used for exposure assessment. Mortality data (underlying cause of death) were obtained from the Ministry of Public Health and coded using ICD-10.The core data file (clogit_data_2.csv) is a tabular dataset with 2,200,824 rows and 168 columns. Each row represents a single person-day in a matched risk set, including one case day (the death date) and multiple referent (control) days selected using a time-stratified scheme matching on calendar time and day of week. Key variables include an anonymous individual identifier (ID), an indicator of whether the row corresponds to the case day (case, 1 = case, 0 = referent), the calendar date (date), demographic characteristics (sex, age at death, age group), and underlying cause of death coded using ICD-10 (icd_death, ICD10_Chapter, ICD_major, IC_minor). Spatial information is provided through administrative area codes and names at province, district, subdistrict, and ISO 3166 levels (ADM1_EN, ADM2_EN, ADM3_EN, address). Direct personal identifiers are not included.To limit outlier effects, the top and bottom 1% of hourly pollutant concentrations were removed, and daily means were calculated when ≥ 75% of hourly data were available at each air quality monitoring station. Missing daily values were then imputed sequentially using 7-day and 30-day moving averages, requiring at least 3 and 9 valid observations (≥25% of each window), respectively, with all imputations performed separately for each station to preserve site-specific temporal variability. Overall, this procedure resulted in less than 3% of daily exposure values being imputed.