Applying novel methods to study long-term exposure to air pollution and temperature and annual mortality in India

BACKGROUND

India is particularly affected by environmental exposures such as ambient air pollution and non-optimal temperature. These risk factors could be linked to a major burden on public health. The evidence from India on mortality burden due to long-term exposure to air pollution and ambient temperature is limited. It is crucial to identify areas that are most vulnerable to ambient air pollution and to identify the extent of mortality burden due to ambient air pollution and temperature, especially when the resources are scarce to address these issues.

AIM

The aim of the thesis was to apply novel methods to study the long-term exposure to air pollution and temperature on annual mortality in India.

METHODS

Data sources

All of the three included studies were based on Indian data at national level.

Exposure: Ambient air pollution (PM2.5) and ambient temperature were derived from previously developed nationwide machine-learning models at 1 km x 1 km spatial resolution from 2008 to 2019.

Outcome: For Study II & III, we collected district level annual counts of mortality from the national register for births and deaths between 2009 and 2019. Study I was a descriptive study and did not include any health outcome.

Covariates

We used district level variable like gross domestic product (GDP) per capita and proportion of women literacy, proportion of households using clean cooking fuel, median age, proportion of households with electricity, relative humidity, precipitation and proportion of people aged 60 and older. We also adjusted for quarterly average temperature in Study II and annual average PM2.5 in Study III as confounders.

Statistical analyses

In Study I, we conducted a geospatial analysis on gridded annual PM2.5 averages using Getis-Ord Gi* statistics for every year to classify hot and cold spots from 2008 to 2019. Then we analyzed the temporal trends of the identified hotspots to categorize them as consistent, emerging and declining hotspots and the population exposed. We also explored if the identified hotspots were included in the National Clean Air Programme (NCAP), a national effort focused on reducing the ambient air pollution levels.

In Study II & III, we conducted quasi-experimental studies using a modified version of the Difference-In-Differences (DID) approach. For the main exposure of interest, in Study II, we used ambient air pollution (district level annual averages), and in Study III we used ambient temperature (district level quarterly averages and annual counts of episodes of extremes of heat and cold). We used generalized additive models with quasi-Poisson distribution with interactive fixed effects for the spatial unit and time trends in both Study II & III. The analysis was conducted on a national level for Study II and climate zone-specific for Study III. The spatial unit of analysis was administrative divisions (collection of districts) in Study II and districts in Study III.

RESULTS

Our results (Study I) highlighted areas of the country with consistently higher levels of air pollution compared to the spatial neighbors and emerging hotspots with a concerning trend and the significant levels of population exposure. The results have alerted us to areas of declining hotspots which could be serve as examples for the country for any policies or interventions. From our analysis we found that additional cities could be included in the NCAP. Further, in Study II, we observed an increased mortality risk of 8.6% for with every additional 10 ug/m3 increment of annual average PM2.5 corresponding to an attributable burden of 3.8 million deaths (in comparison to Indian National Ambient Air Quality Standards) and 16.6 million deaths (in comparison to revised WHO guidelines) throughout the study period. We also observed a near linear exposure-response function for PM2.5 and mortality. In Study III we studied the relationship between long-term ambient temperature exposure and mortality and found that some climatic zones demonstrated increased risk for mortality while others did not display a clear pattern of increased risk for quarterly temperatures. The increased annual mortality risk for annual counts of extreme heat and cold episodes was also specific to some climate zones and not uniform across the country. We also observed an effect modification by socio-economic status, with higher risk for mortality by increasing quarterly temperature among poorer districts.

CONCLUSIONS

To conclude, the three constituent papers of this thesis quantify on the extent of population exposure to air pollution as well as the mortality risk and burden from long-term exposure to air pollution and temperature in India using novel methods. The methods used in thesis allowed us to fully utilize the mortality data and other covariates available at the national level, albeit lack of individual-level data and differing spatial and temporal availability of the covariates, to elucidate the complex relationship of long -term exposure to air pollution and temperature on mortality. The results clearly indicate higher mortality risk for increasing levels of air pollution and a significant burden of mortality above the Indian National Ambient Air Quality Guidelines. Finally, the results indicate varying levels of increased risk for mortality across different climate zones for increasing quarterly temperatures and episodes of heat and cold extremes. Our findings indicate a need for stricter regulatory standards and targeted interventions to improve public health.

List of scientific papers

I. Jaganathan S, Rajiva A, Amini H, de Bont J, Dixit S, Dutta A, Kloog I, Lane KJ, Menon JS, Nori-Sarma A, Prabhakaran D, Schwartz J, Sharma P, Stafoggia M, Walia GK, Wellenius GA, Prabhakaran P, Ljungman P, Mandal S. Nationwide analysis of air pollution hotspots across India: A spatiotemporal PM2.5 trend analysis (2008-2019). Environ Res. 2025 Jan 1;264(Pt 1):120276. Epub 2024 Nov 5. PMID: 39510231; PMCID: PMC11790316.
https://doi.org/10.1016/j.envres.2024.120276

II. Jaganathan S, Stafoggia M, Rajiva A, Mandal S, Dixit S, de Bont J, Wellenius GA, Lane KJ, Nori-Sarma A, Kloog I, Prabhakaran D, Prabhakaran P, Schwartz J, Ljungman P. Estimating the effect of annual PM25 exposure on mortality in India: a difference-in-differences approach. Lancet Planet Health. 2024 Dec;8(12):e987-e996. PMID: 39674205; PMCID: PMC11790315.
https://doi.org/10.1016/S2542-5196(24)00248-1

III. Jaganathan S, Stafoggia M, Rajiva A, Mandal S, Dixit S, de Bont J, Wellenius GA, Lane KJ, Nori-Sarma A, Kloog I, Prabhakaran D, Prabhakaran P, Schwartz J, Ljungman P. Long-term exposure to ambient temperature and mortality risk in six climatic zones of India: a difference-in-differences approach. [Manuscript]