Acute Air Pollution Exposure and Risk of Suicide Completion
Acute Air Pollution Exposure and Risk of Suicide Completion
The Utah Department of Health's Office of the Medical Examiner provided data on all completed suicides occurring in Salt Lake County, Utah, between January 1, 2000, and December 31, 2010 (n = 1,546; Table 1). Data acquired on each suicide decedent included date of death, age, sex, and method of death. Age was categorized on the basis of similar published studies. Method of death was classified as violent or nonviolent. Nonviolent methods of death included drug overdose, drowning, poisoning, and gas; all other methods of death were categorized as violent.
Data on 4 air pollutants—including nitrogen dioxide (2 monitors), particulate matter with an aerodynamic diameter ≤2.5 μm (PM2.5; 4 monitors), particulate matter with an aerodynamic diameter ≤10 μm (PM10; 4 monitors), and sulfur dioxide (3 monitors)—were acquired for the study period (December 29, 1999–December 31, 2010) from the Environmental Protection Agency's AirData website (http://www.epa.gov/airquality/airdata/). The daily exposure to air pollutants assigned to Salt Lake County residents from January 1, 2000, to December 31, 2010, was estimated by averaging hourly air pollution measures across all monitors for a particular pollutant. Data were 99.9% complete for PM2.5, 71.5% complete for PM10, 94.9% complete for nitrogen dioxide, and 95.9% complete for sulfur dioxide during the study period. Multiple imputation of missing air pollutant data was conducted in SAS, version 9.3 (SAS Institute, Inc., Cary, North Carolina) using PROC MI. Five copies of the original data set were made and 5 estimates of missing data were generated, producing 5 complete data sets available for the analysis.
Meteorological data on daily mean temperature (°F), daily mean dew point temperature (°F), and daily mean air pressure (mb) for Salt Lake City International Airport were downloaded from the National Oceanic and Atmospheric Administration's National Climatic Data Center (Global Surface Summary of the Day data (https://data.noaa.gov/dataset/global-surface-summary-of-the-day-gsod)). Information on mean daily sunlight (KJ/m) for Salt Lake County was downloaded from the Centers for Disease Control and Prevention's WONDER website (http://wonder.cdc.gov/).
A time-stratified case-crossover design was used to examine the relationship between suicide risk and air pollutant exposure among suicide decedents who had resided in Salt Lake County, Utah. The day of the suicide was selected as the case period, and the control period consisted of all days falling on the same day of the week within the same month as the case period. Each control period per case period was comprised of 3 or 4 control days. The time-stratified design was selected because it is not subject to time-trend bias and because it controls for seasonality, day of the week, and cross-subject variability.
Conditional logistic regression models were formulated to measure the association of individual air pollutants (nitrogen dioxide, PM2.5, PM10, and sulfur dioxide) with suicide risk, as represented by the odds ratios and 95% confidence intervals related to an interquartile-range (IQR) increase in mean air pollutant concentration. The strength of the relationship was also measured as the percent change in suicide odds associated with an IQR increase in pollutant concentration. A series of models were fitted considering exposure to single-lag and cumulative-lag air pollution values. The single-lag models examined the relationship between suicide risk and air pollutant concentration on the day of the suicide (lag day 0) and on each of the 3 days preceding the suicide (lag day 1, lag day 2, and lag day 3). For the cumulative-lag models, 3 cumulative pollutant concentrations were estimated separately for each pollutant, where cumulative lag 1 was defined as the average concentration for the day of the suicide and the day preceding the suicide, cumulative lag 2 was the average concentration for the day of the suicide and the 2 days preceding the suicide, and cumulative lag 3 was the average for the day of the suicide and the 3 days preceding the suicide.
The models included adjustment for the confounding effects of meteorological variables, including average daily sunlight during the previous 3 days, daily mean temperature, daily mean temperature for the previous 3 days, mean dew point temperature, mean dew point temperature for the previous 3 days, daily mean air pressure, and daily mean air pressure for the previous 3 days. All meteorological variables except average daily sunlight were cubic spline–transformed prior to being added to the models.
Further analyses were performed to evaluate variation in the strength of the association between acute exposure to nitrogen dioxide and PM2.5 and suicide risk on lag days 2 and 3 and during cumulative lags 2 and 3 by suicide characteristic or population subgroup. In this secondary analysis, data in the models were stratified by age group (≤35 years, 36–64 years, or ≥65 years), sex, season of death (spring, summer, fall, or winter), and method of death (violent vs. nonviolent). To allow for comparison of this study's findings with those of the Kim et al. study by season, we also categorized season of death into 2 periods: winter/summer (the "stable period") and spring/fall (the "transition period").
Separate models were formulated for each of the 5 data sets in which missing data were imputed. The MIANALYZE procedure in SAS 9.3 was used to combine model output and to generate overall parameter estimates and standard errors. Institutional review board–exempted approval was obtained for this study from the University of Utah.
Methods
Data on Suicide Decedents
The Utah Department of Health's Office of the Medical Examiner provided data on all completed suicides occurring in Salt Lake County, Utah, between January 1, 2000, and December 31, 2010 (n = 1,546; Table 1). Data acquired on each suicide decedent included date of death, age, sex, and method of death. Age was categorized on the basis of similar published studies. Method of death was classified as violent or nonviolent. Nonviolent methods of death included drug overdose, drowning, poisoning, and gas; all other methods of death were categorized as violent.
Air Pollution and Meteorological Data
Data on 4 air pollutants—including nitrogen dioxide (2 monitors), particulate matter with an aerodynamic diameter ≤2.5 μm (PM2.5; 4 monitors), particulate matter with an aerodynamic diameter ≤10 μm (PM10; 4 monitors), and sulfur dioxide (3 monitors)—were acquired for the study period (December 29, 1999–December 31, 2010) from the Environmental Protection Agency's AirData website (http://www.epa.gov/airquality/airdata/). The daily exposure to air pollutants assigned to Salt Lake County residents from January 1, 2000, to December 31, 2010, was estimated by averaging hourly air pollution measures across all monitors for a particular pollutant. Data were 99.9% complete for PM2.5, 71.5% complete for PM10, 94.9% complete for nitrogen dioxide, and 95.9% complete for sulfur dioxide during the study period. Multiple imputation of missing air pollutant data was conducted in SAS, version 9.3 (SAS Institute, Inc., Cary, North Carolina) using PROC MI. Five copies of the original data set were made and 5 estimates of missing data were generated, producing 5 complete data sets available for the analysis.
Meteorological data on daily mean temperature (°F), daily mean dew point temperature (°F), and daily mean air pressure (mb) for Salt Lake City International Airport were downloaded from the National Oceanic and Atmospheric Administration's National Climatic Data Center (Global Surface Summary of the Day data (https://data.noaa.gov/dataset/global-surface-summary-of-the-day-gsod)). Information on mean daily sunlight (KJ/m) for Salt Lake County was downloaded from the Centers for Disease Control and Prevention's WONDER website (http://wonder.cdc.gov/).
Statistical Analyses
A time-stratified case-crossover design was used to examine the relationship between suicide risk and air pollutant exposure among suicide decedents who had resided in Salt Lake County, Utah. The day of the suicide was selected as the case period, and the control period consisted of all days falling on the same day of the week within the same month as the case period. Each control period per case period was comprised of 3 or 4 control days. The time-stratified design was selected because it is not subject to time-trend bias and because it controls for seasonality, day of the week, and cross-subject variability.
Conditional logistic regression models were formulated to measure the association of individual air pollutants (nitrogen dioxide, PM2.5, PM10, and sulfur dioxide) with suicide risk, as represented by the odds ratios and 95% confidence intervals related to an interquartile-range (IQR) increase in mean air pollutant concentration. The strength of the relationship was also measured as the percent change in suicide odds associated with an IQR increase in pollutant concentration. A series of models were fitted considering exposure to single-lag and cumulative-lag air pollution values. The single-lag models examined the relationship between suicide risk and air pollutant concentration on the day of the suicide (lag day 0) and on each of the 3 days preceding the suicide (lag day 1, lag day 2, and lag day 3). For the cumulative-lag models, 3 cumulative pollutant concentrations were estimated separately for each pollutant, where cumulative lag 1 was defined as the average concentration for the day of the suicide and the day preceding the suicide, cumulative lag 2 was the average concentration for the day of the suicide and the 2 days preceding the suicide, and cumulative lag 3 was the average for the day of the suicide and the 3 days preceding the suicide.
The models included adjustment for the confounding effects of meteorological variables, including average daily sunlight during the previous 3 days, daily mean temperature, daily mean temperature for the previous 3 days, mean dew point temperature, mean dew point temperature for the previous 3 days, daily mean air pressure, and daily mean air pressure for the previous 3 days. All meteorological variables except average daily sunlight were cubic spline–transformed prior to being added to the models.
Further analyses were performed to evaluate variation in the strength of the association between acute exposure to nitrogen dioxide and PM2.5 and suicide risk on lag days 2 and 3 and during cumulative lags 2 and 3 by suicide characteristic or population subgroup. In this secondary analysis, data in the models were stratified by age group (≤35 years, 36–64 years, or ≥65 years), sex, season of death (spring, summer, fall, or winter), and method of death (violent vs. nonviolent). To allow for comparison of this study's findings with those of the Kim et al. study by season, we also categorized season of death into 2 periods: winter/summer (the "stable period") and spring/fall (the "transition period").
Separate models were formulated for each of the 5 data sets in which missing data were imputed. The MIANALYZE procedure in SAS 9.3 was used to combine model output and to generate overall parameter estimates and standard errors. Institutional review board–exempted approval was obtained for this study from the University of Utah.
Source...