Daily exposure to PM2.5 and 1.5 million deaths: A time-stratified case-crossover analysis in the Mexico City Metropolitan Area

Background. Satellite-based PM2.5 predictions are being used to advance exposure science and air-pollution epidemiology in developed countries; including emerging evidence about the impacts of PM2.5 on acute health outcomes beyond the cardiovascular and respiratory systems, and the potential modifying effects from individual-level factors in these associations. Research on these topics is lacking in Latin America. Methods. We used a time-stratified case-crossover study design with 1,479,950 non-accidental deaths from Mexico City Metropolitan Area for the period of 2004–2019. Daily 1×1 km PM2.5 (median=23.4 μg/m3; IQR=13.6 μg/m3) estimates from our satellite-based regional model were employed for exposure assessment at the sub-municipality level. Associations between PM2.5 with broad-category (organ-system) and cause-specific mortality outcomes were estimated with distributed lag conditional logistic models. We also fit models stratifying by potential individual-level effect modifiers including; age, sex, and individual SES-related characteristics namely: education, health insurance coverage, and job categories. Results. PM2.5 exposure was associated with higher total non-accidental, cardiovascular, cerebrovascular, respiratory, and digestive mortality. A 10-μg/m3 PM2.5 higher cumulative exposure over one week (lag06) was associated with higher cause-specific mortality outcomes including hypertensive disease [2.28% (95%CI: 0.26%–4.33%)], acute ischemic heart disease [1.61% (95%CI: 0.59%–2.64%)], other forms of heart disease [2.39% (95%CI: −0.35%–5.20%)], hemorrhagic stroke [3.63% (95%CI: 0.79%–6.55%)], influenza and pneumonia [4.91% (95%CI: 2.84%–7.02%)], chronic respiratory disease [2.49% (95%CI: 0.71%–4.31%)], diseases of the liver [1.85% (95%CI: 0.31%–3.41%)], and renal failure [3.48% (95%CI: 0.79%–6.24%)]. No differences in effect size of associations were observed between SES strata. Conclusions. Exposure to PM2.5 was associated with mortality outcomes beyond the cardiovascular and respiratory systems, including specific death-causes from the digestive and genitourinary systems, with no indications of effect modification by individual SES-related characteristics.

In this study, we analyzed all ~1.5 million adult mortality records from the Mexico City Metropolitan Area for the period from 2004 to 2019, and estimated the daily percent increase in non-accidental mortality for broad-category and cause-specific mortality outcomes; as well as sex and age-group (adults and the elderly) categories, and SES-related indicators, associated with a 10-μg/m 3 higher PM 2.5 concentration.

Mortality data
We obtained mortality records from the National Institute of Statistics and Geography of Mexico and temperature, i.e. up to one week (lag 6) before the date of event. All mortality records included information on date of death, geographic identifiers for place of residence, underlying cause of death classified according to the International Classification of Diseases Tenth Revision (ICD-10), sex, age, education, healthcare affiliation, and job category.
We restricted our analysis to non-accidental causes of death excluding most ICD-10 codes from the blocks V-Y (i.e. various accidents and side effects of treatments), and deaths in people ≥ 18 years-old.
ICD-10 codes X6-Y0 were retained based on the evidence that PM 2.5 can trigger intentional self-harm and aggressive behavior (Berman et al., 2019;Davoudi et al., 2021). We organized individual ICD-10 coded death records into mutually exclusive broad-category (organ-system) and cause-specific mortality outcomes. The selected ICD-10 codes included in our research were chosen to facilitate comparison with recently-published studies focused on under-studied associations between short-term exposure to PM 2.5 with mortality and morbidity outcomes Yu et al., 2019).
To address the potential modifying effects of age, sex and SES on the PM 2.5 -mortality relationship, we performed stratified analyses by age-group (adults: 18-64 years-old, and elderly: +65 years-old) and sex (males and females) for all broad-category mortality outcomes. For all non-accidental mortality, we also performed stratified analyses by age-group (adults and the elderly), degree of education (basic and more than basic education level), type of healthcare affiliation (having health social security affiliation or not), and employment status based on job categories. We used health social security affiliation rather than health insurance alone, since it is related to formal employment and other individual social benefits that differ from specific healthcare programs aimed to provide health services to the most vulnerable . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 17, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 populations in Mexico. Employment status was defined from the original job category variable in the mortality records that included the categories "Does not work" and "Looking for a job", which were combined into one single category "Not employed". However, this categorization does not rule out the possibility that part of the category "Does not work" can also include retired people, with a higher SES compared with those who are actually unemployed. The rest of the job categories were combined into the category "Employed". We also explored education-specific and job-specific associations. For the latter, given that the Mexican classification of occupations was updated in 2012, and specific occupations coded before 2012 were redistributed into new job categories in the most recent version, it was not possible to homogenize all job categories for the whole period of study. Thus, analyses of specific job categories were restricted to the period from 2013 to 2019. We compared if the effect size of PM 2.5 exposure differed by age group, and also by education, healthcare affiliation, and employment status, in the broadcategory mortality outcomes, and total non-external mortality by following the methods described by Altman and Bland, 2003.

Environmental data
We utilized daily PM 2.5 predictions with spatial resolution of 1×1 km from our recently developed model based on extreme gradient boosting (XGBoost), and inverse-distance weighting (IDW) that uses aerosol optical depth data, meteorology, and land-use variables to predict daily mean PM 2.5 for the Mexico City Metropolitan Area (Gutiérrez-Avila et al., 2022). Daily mean air temperature with the same 1x1 km resolution was obtained from our satellite-based land surface temperature model for Central Mexico (Gutiérrez-Avila et al., 2021). We restricted our analyses to the spatial domain with available predictions for both PM 2.5 and temperature exposures over the Mexico City Metropolitan Area. The study area was composed of 667 sub-municipal geographic areas defined as "localities" by the Instituto Nacional de Estadistica y Geografia. Localities correspond to the third level of subnational division in Mexico, after states and municipalities (Instituto Nacional de Geografía y Estadística, 2010). The total population in the Mexico City Metropolitan Area in 2010 was 20,116,842 (Secretaría de Desarrollo Agrario, Territorial y Urbano, 2018). For the same year, the population living in the localities included in our study region was 19,711,516 inhabitants, ~96% of the total Mexico City Metropolitan Area. Urban localities had a population of 19,346,527 (median of 8,523 inhabitants; range of 910 to 1,815,786) inhabitants, and rural localities had a total population of 364,989 (median of 582 inhabitants; range 1 to 5,135) inhabitants. We assigned exposures to PM 2.5 and temperature for all mortality records at the locality level. For 255 urban localities with census-provided polygons (median land area of 2.9 km 2 ; range 0.2 to 129.4 km 2 ), we estimated daily exposures using population-weighted aggregation with population density from the Gridded Population of the World (GPWv4) ~1-km raster cells (Center for International Earth Science . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 17, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 Information Network-CIESIN-Columbia University, 2016), which utilizes data from the 2010 Mexican Census, and the R package exactextractr (Baston, 2020). For 412 rural localities, the census only assigned points (rather than enclosing polygons), and for this subset of rural localities we assigned exposure to PM 2.5 and temperature using the 1x1 km grid cell containing the corresponding censusassigned points.

Statistical Analysis
We estimated the association between short-term exposure to PM 2.5 with mortality using a time-stratified case-crossover design (Levy et al., 2001). Case days were defined as the date of death, and control days (3 to 4 days per month) were chosen on the same day of week as the case day within the same month and year. This approach controls for potential confounding effects by day of week, seasonal patterns and long-term time trends (Janes et al., 2005). Time-invariant covariates are controlled by design (withinperson matching), and are not considered to be confounders; thus, daily variations in exposure to PM 2.5 and temperature on the case day are compared to exposures on control days. All models were adjusted for non-linear effects of temperature with quadratic b-splines (4 degrees of freedom), and equally-spaced knots in the space of this predictor (Gasparrini, 2021). The odds ratios for all dependent variables associated with short-term exposure to PM 2.5 were estimated with linear terms in stratified Cox proportional hazards models, equivalent to conditional logistic regression. We explored non-linearities in the concentration-response relationships between PM 2.5 with all broad-category mortality outcomes by comparing the fit of generalized additive models (GAMs) including parsimonious non-linear (penalized thin-plate splines) PM 2.5 terms (lag 0 to lag 6 ) with the fit of linear PM 2.5 terms using likelihood ratio tests.
To address the delayed effects of exposure to PM 2.5 on all mortality outcomes we included distributed lag terms up to 6 days before the case day (seven terms for lags 0 to 6) to estimate mortality risks (Gasparrini et al., 2010). For ease of interpretation, all odds ratios were converted to percentage increase per 10 μ g/m 3 higher PM 2.5 .
All analyses were performed in R version 4.2.1 (R Core Team, 2020) with packages: data.table (Dowle et al., 2019), survival (Therneau and Lumley, 2015), dlnm (Gasparrini, 2011), andmgcv (Wood SN, 2015). Because the records used in all analyses were publicly available, the PI previously received a determination of exempt human research: 45 CFR 46. 101(b) (Category 4) from the Institutional Review Board of the Icahn School of Medicine at Mount Sinai.

Results
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 17, 2023. ; https://doi.org/10. 1101/2023 The total number of non-accidental deaths (≥18 years-old) analyzed from 2004 to 2019 in the Mexico City Metropolitan Area was 1,479,950. Table 1 describes the demographic characteristics of the study population. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 17, 2023. ; https://doi.org/10.1101/2023.01.15.23284576 doi: medRxiv preprint There were slightly more deaths in men than women, and deaths in the elderly (≥65 years-old) accounted for ~63% of all deaths. Around 50% of the decedents attended only elementary school, 63% had health social security affiliation, and 32% were not working at the time of death. The five most common job categories in the mortality records for the period 2013-2019 were: not-employed or not working at the time of death, sales, unspecified occupations, craft and related trades, and professionals and technicians.
Over the study period, the mean population-weighted exposure to PM 2.5 and temperature in the study region were 24.6µg/m 3 , and 16.6º C, respectively. Table 2 shows the total number of organ-system and cause-specific deaths (≥18 years-old) analyzed over the study period. Circulatory system diseases were the most prevalent death causes, accounting for almost 30% of the total, followed by digestive (12%) and respiratory (10%) diseases. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 17, 2023. ; https://doi.org/10.1101/2023.01.15.23284576 doi: medRxiv preprint Among the specific death causes, acute ischemic heart disease and hypertensive disease were the leading causes from the circulatory system. Diseases of the liver were the leading death causes from the digestive system, and chronic respiratory disease, influenza and pneumonia were the most common causes of death from the respiratory system (Table 2).
Results from our assessment of linearity in the concentration-response functions between short-term exposure to PM 2.5 with all broad-category mortality outcomes showed no significant differences in model's fit when including non-linear PM 2.5 exposures; therefore all our further results are based on linear concentration-response functions. Figure 1 shows associations between PM 2.5 exposure with the broad-category death causes analyzed over the study period. For consistency with previous studies, twodays (lag 0 + lag1) and one-week, (sum from lag 0 to lag 6 ) cumulative associations with PM 2.5 are shown.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 17, 2023. ; Associations between PM 2.5 exposure and cause-specific mortality for lag 01  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 17, 2023. ; Cumulative associations from distributed lags can capture the impact of multiple days after initial exposure to PM 2.5 on mortality. In general, consistent associations between short-term exposure to PM 2 with all-cause (i.e. total non-accidental) or broad-category death causes (mostly from cardiovascular and respiratory diseases) have been observed during the first week of exposure across study sites. Therefore communicating results for lag 01 and lag 06 , has become the standard, as the evidence shows immediate (lag 01 ) and more prolonged (lag 06 ) effects on cardiovascular, and respiratory mortality, respectively. Tha said, the onset and duration of the mechanisms of damage leading to death can vary for other health outcomes, with mortality from PM 2.5 exposure occurring with a delay of a few days (World Health Organization. Regional Office for Europe, 2021). Little has been documented about the timing of when the largest single lag effects of PM 2.5 on cause-specific mortality occur, and these "peaks" in mortality can be missed when reporting associations only for lag 01 , or lag 06 , and this information can provide meaningful information for health professionals. Figure 3 shows the largest single lag associations for two cause-specific mortality outcomes (hemorrhagic stroke and influenza and pneumonia); and to complement this information, results for all single lag and cumulative associations for all cause-specific mortality outcomes are reported in Table S2. Same-day exposure to PM 2.5 (lag 0 ) showed the largest single lag associations with diseases of the arteries [4.62% (95%CI: 0.73%-8.67%)], hemorrhagic strok CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 17, 2023. ; https://doi.org/10.1101/2023.01.15.23284576 doi: medRxiv preprint elderly; compared to those who were working with mortality risks of 0.90% (95%CI: -0.48%-2.29%) and 1.70% (95%CI: 0.23%-3.20%), for adults and the elderly, respectively. Our test of heterogeneity in effect size between SES related strata, does not suggest effect modification between SES strata. Figure   S1, shows job-specific associations with PM 2.5 exposure for the same period of time.

Discussion
This study presents a detailed analysis on the associations between short-term exposure to PM 2.5 with total non-external, broad-catergory (organ-system) and cause-specific mortality outcomes; and the role that age, sex, and potential SES-related effect modifiers play in such associations in the Mexico City Metropolitan Area. Our research involved the use of state-of-the-art methods to estimate daily exposure to PM 2.5 , using highly spatially-resolved satellite-based PM 2.5 predictions. With our PM 2.5 predictions, w were able to assign daily exposure to PM 2.5 to all case and control days at the sub-municipality level; leveraging the smallest geographic identifier for place of residence included in the Mexican mortality records. To our knowledge, this is the most comprehensive analysis in Central Mexico in terms of geographic extent, amount of mortality outcomes analyzed, and assessment of potential effect modifiers in these associations using individual mortality records. in re tion, er 10 Area le ure s, we ers . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
However, it is possible that the selection of heterogeneous effect estimates from different lags structures that were combined in the meta-analyzes of Atkinson et al. (2014), and Orellano et al. (2020) could have induced some biases. Differential toxicity related to PM 2.5 composition, variation in the underlying health status of the population, and distribution of potential effect modifiers between study regions can also play a role in the differences reported in the literature. Also, it has been suggested that at higher concentrations of PM 2.5 like those observed in Asian cities, the slopes in the concentration-response function between PM 2.5 and mortality flattens out (supralinear exposure-response curve) for cardiorespiratory outcomes, which might also explain the differences in effect estimates between study regions (Liu et al., 2019).
Future research efforts exploring the shape of the concentration response functions for mortality outcomes other than cardiorespiratory outcomes could aid to understand the differences in mortality risks across study regions.
Among the cause-specific mortality associations with PM 2.5 , those related to the circulatory system (e.g. diseases of the arteries, hypertensive disease, acute ischemic heart disease, and stroke) were the most clear, with the largest associations on the same day of exposure (lag 0 in Table S2). Li et al. (2018) also reported consistent associations between PM 2.5 with mortality outcomes from the circulatory and respiratory systems in Beijing. However, we found larger associations compared to those reported by Li et al. (2018) for ischemic heart disease (0.46%; 95% CI: 0.19%-0.72%), influenza and pneumonia . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 17, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 (0.35%; 95% CI: -0.32%-1.03%), and chronic respiratory disease (0.62%; 95% CI: 0.08%-1.16%).  also reported associations between PM 2.5 and sub-specific causes of ischemic heart disease such as: acute ischemic heart disease (ICD-10 codes: I20-I22, I24), acute myocardial infarction (ICD-10 codes: I21-I22), and myocardial infarction (ICD-10 codes: I21-I23). We did not explore such subtypes of ischemic heart disease, as Li et al. (2018) did, because the grouping of such health outcomes involves the overlap of several ICD-10 codes; while we only focused on reporting associations for mutually exclusive health outcomes. We found larger associations than those reported by Xu et al. (2020) for ischemic heart disease (-0.06%; 95% CI: -0.33%-0.22%), and chronic respiratory disease (0.96%; 95% CI: 0.35%-1.57%) also in Beijing. Our results for specific cardiovascular and cerebrovascular outcomes were larger than those reported by Chen et al. (2018) for 30 Chinese counties (Chen et al., 2018). Intentional-self harm mortality was not associated with PM 2.5 exposure in our study, which is consistent with results of Contrary to the results from Li et al. (2018), we did not find associations between short-term exposure to PM 2.5 with acute mortality outcomes from the nervous system (e.g. extrapyramidal and movement disorders), however we observed positive associations between cumulative exposure to PM 2.5 with digestive (diseases of the liver) and genitourinary (renal failure) mortality. Exposure to PM 2.5 has been associated with increased serum levels of hepatic enzymes, such as γ -glutamyltranspeptidase (GTP), and alanine transaminase (ALT), which are biomarkers of liver damage. PM 2.5 can induce systemic oxidative stress and inflammation when deposited in the airways, and when swallowed PM 2.5 are removed from the airways by mucociliary clearance leading to gastrointestinal exposure with the largest internal dose and adverse effects observed in the liver (Kim et al., 2014;Tavera Busso et al., 2020). On the other hand, recent evidence has shown that PM 2.5 is related to renal injury and increases the risk of nephropathy, as well as increased risk of first hospital admission from kidney and total urinary system diseases (Lee et al., 2022;Rasking et al., 2022). PM 2.5 can unbalance the kidney function by accumulation in the kidney tissue, endothelial dysfunction, abnormal renin-angiotensin system, and immune complex deposition.
The mechanisms of damage from PM 2.5 to the kidney involve inflammation, oxidative stress, apoptosis, DNA damage, and autophagy (Xu et al., 2022).
The most recent version of the EPA's Integrated Science Assessment (ISA) for Particulate Matter (2019) has defined the associations between short-term exposure to PM 2.5 with cardiovascular and respiratory mortality as causal and likely causal, respectively; but less evidence is available for making conclusions about other causes of death (EPA, 2019). Although our results suggest that short-term exposure to PM 2.5 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 17, 2023. ; https://doi.org/10.1101/2023.01.15.23284576 doi: medRxiv preprint may trigger cause-specific mortality beyond the circulatory and respiratory systems, more epidemiological evidence is needed to understand the links between PM 2.5 exposure with the different specific death causes evaluated in our study. Emerging evidence about the adverse effects of PM 2.5 exposure on multiple prevalent but rarely studied causes of hospital admissions (Wei et al., 2019) adds to the body of epidemiologic evidence that supports our findings, and it opens the possibility to replicate such results in different locations with ongoing or future health studies.
Information on the timing when health effects occur after initial exposure to PM 2.5 , or any other air pollutant, is of importance for public health administrators and pollution control managers. For instance, identifying the lag days over which health effects are observed can inform the development of air quality standards, risk communication tools (e.g. air quality indices), quantification of health impacts, mitigation strategies (e.g. planning and allocation of clinical resources), and the design of emission control measures. Our results are consistent with the epidemiologic evidence showing that short-term exposure to PM 2.5 has an immediate effect on non-accidental and cardiovascular mortality, and a more prolonged effect on respiratory mortality (EPA, 2019). However, when considering specific death causes (Table S2) within the broad-category mortality outcomes, there seems to be more variability about the timing when the largest (single and cumulative) associations are observed. For instance, chronic ischemic heart disease and ischemic stroke showed the largest associations with PM 2.5 on lag 4 , while the rest of the death causes from the circulatory system showed the largest association on lag 0 and lag 1 . For respiratory mortality, our results are within the range of observed associations, with chronic respiratory disease showing the largest association with PM 2.5 on lag 1 , while the association with influenza and pneumonia remained high at lag 6 . On the other hand, it is known that the strength of the relationship between exposure to PM 2.5 and health effects varies depending on the exposure duration i.e., short-or long-term exposure, but less is known about the strength of this relationship for sub-daily exposures e.g. hourly peak exposures. Few epidemiologic studies have focused on the associations between sub-daily exposures to PM 2.5 with morbidity and mortality outcomes (Lin et al., 2017;Madsen et al., 2012). Results from such investigations are inconclusive on whether sub-daily exposures pose a greater risk of mortality, or not, compared to daily averaged exposures; being possible that the sub-daily metrics of PM 2.5 employed for exposure assessment in previous studies had a limited spatiotemporal variability (EPA, 2019). In our research, we only focused on exposures to daily mean PM 2.5 concentrations assuming that the same averaging time could have the same relevance for all mortality outcomes.
However, we can not rule out the possibility that there might be averaging times (sub-daily, or subchronic exposures) that could be of greater relevance for the different death causes analyzed in our research.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 17, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 It is known that subgroups of the population with lower SES indicators may have higher prevalence of health disorders and comorbidities, lower living standards, and higher exposure to air pollutants (e.g. living closer to a highway or lack of green spaces) (Shavers, 2007;Xu et al., 2020). We used three indicators of SES, namely education, health social security affiliation, and employment status to assess their roles as potential effect modifiers in the PM 2.5 -mortality relationship. Although in general point estimates for the lower SES related strata (lower education level, not having health social security affiliation, and not being working at the time of death) seemed larger than those in the higher SES indicators; we did not find evidence of effect modification in the PM 2.5 -mortality associations (analyzed on the log scale) between SES strata (Altman and Bland, 2003). In Mexico there is limited evidence about the role of education in the association between short-term exposure to PM 2.5 and mortality, however O'Neill et al. (2008) did not observe evidence of heterogeneity in the association between shortterm exposure to PM 10 and mortality across education levels (O'Neill et al., 2008). Health insurance access can modify the susceptibility to detrimental effects of environmental stressors, including PM 2.5 , by providing access to medications, supplements and overall preventive services (Woolhandler and Himmelstein, 2017). Our results did not show significant effect modification between strata of health social security affiliation. Also, our results for employment status did not suggest greater risk of mortality for those in the not-employed strata compared with people in the employed category. The evidence on the effects of working status on health suggests that unemployment is related to poor health (e.g. greater risk of mental illness, physical complaints, and an increased risk for coronary heart diseases) and early mortality (Kim and von dem Knesebeck, 2015). Although our models were further stratified by age group to reduce its potential confounding effects, our results should be carefully interpreted, since there was not a well-defined measure of unemployment excluding all the economically inactive (Clemens et al., 2015). Also, we did not have information on the health status of those in the "non-working" group before death, so it is not possible to determine if a deteriorating health status led to a "non-working" condition (healthy worker effect). When evaluating the role of specific job types in the association between short-term exposure to PM 2.5 and mortality, the lack of consistency in the codification of specific job categories throughout the period of study led us to evaluate these associations for a shorter time period (2013-2019), likely reducing power to identify associations by individual job categories ( Figure S1). Two major problems with the use of death certificates for classifying employment include ambiguity about whether unemployed persons are actually retired or if a prior occupation is reported for retired decedents and thus may not represent their recent exposures.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 17, 2023. ; https://doi.org/10. 1101/2023 Other limitations in our study could be related to the inaccurate classification of death causes for some diseases. In general, inaccurate classification of death causes in Mexico is considered low (Naghavi et al., 2010). However, errors might exist in the classification of some mortality outcomes, including those from the circulatory system, which have been recognized among the most affected by the use of "garbage codes", i.e. those that are not underlying causes of death such as heart failure, but codes that define poorly-specified diagnoses not clearly identifying a death cause. For Mexico City, we have previously reported that the use of garbage codes tends to bias effect estimates of cardiovascular outcomes to the null, and that proportional redistribution of garbage codes into actual death causes may help to reduce such bias (Gutiérrez-Avila et al., 2018). Also, there are contributions to exposure measurement error that accumulate from the combination of prediction model error and the use of locality of residence to represent all relevant time-activity, which may bias effect estimates (albeit less than the use of a single city-wide exposure time series). Finally, the study of effect modification by cause-specific mortality, age group, sex, and SES indicators can be sensitive to analytic methods. Although our stratified analyses have been the standard in case-crossover and time-series studies, the sparse numbers in some of the strata could have reduced statistical power to detect associations.

Conclusion
Our findings support the growing evidence that short-term exposure to PM 2.5 may trigger cause-specific mortality beyond cardiovascular and respiratory outcomes, including mortality from diseases of the digestive and genitourinary systems. This evidence can be used to inform and update local environmental health policies, and to more comprehensively assess the benefits of interventions aimed to reduce PM 2.5 pollution in the Mexico City Metropolitan Area. Our findings suggest that short-term exposure to PM 2.5 increases the risk of acute mortality in the Mexico City Metropolitan Area, regardless of age groups, and SES characteristics.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 17, 2023. ; https://doi.org/10.1101/2023.01.15.23284576 doi: medRxiv preprint

Funding information
This work was supported by grants from the National Institutes of Health (NIH): R01ES031295 and P30ES023515.

Ethical Approval
Because the records used in all analyses were already publicly available, the senior author (ACJ) received a determination of exempt human research [45 CFR 46. 101(b) (Category 4)] from the Institutional Review Board of the Icahn School of Medicine at Mount Sinai.

Competing interests
The authors declare that they have no conflicts of interest regarding this study.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 17, 2023. ; https://doi.org/10. 1101/2023