Health SDGs are at risk from climate change: Evidence from India

PLOS One | https://doi.org/10.1371/journal.pone.0335529 November 26, 2025 5 / 17 the general form of the specification used. βi(i = 1,2….k) denotes the coefficient of the corresponding independent variable, β0 is the intercept term and Xi represents the independent variables in the model. p is the probability of the outcome (for instance: if the child is stunted, then p = 1). The term (p/1-p) denotes the odds of the outcome occurring, while log(p/1-p) represents the log-odds (logit) of the outcome. log(p/1 – p) = β0 + β1 X1 + β2 X2 + · · · + βk Xk (1) The model is applied to a combined dataset, obtained by merging two datasets, namely the National Family Health Survey-Round 5 (referred to hereafter as NFHS) [82] and the Central Research Institute for Dryland Agriculture (CRIDA) [83] datasets. The former provides information on the progress in the SDG targets, the individual and household level information on health status, socio-economic variables of interest, and on access to health care. The CRIDA dataset provides information on the climate vulnerability across districts. District boundaries for the NFHS and the CRIDA datasets were matched and aligned to ensure consistency in merging the datasets. The data for stunting, wasting, being under- weight and non-institutional deliveries is extracted from the Kids Recode (KR) file of the NFHS, while the data for the ‘problems with access to healthcare’ variable is extracted from the Individual Recode (IR) file for women. The IR and KR

data files were merged with the climate vulnerability dataset at the district level, across 575 districts. There are 154,547 observations for children, and 447348 observations on women available in the merged datasets. The statistical and econometric estimations on the datasets were conducted using Stata software, version 16. Ethics statement This manuscript and its contents are based on publicly available, fully anonymized, secondary data. No human partic- ipants were involved and informed consent requirements do not apply. This study is one component of a larger study approved by the Institutional Ethics Committee of the Institute of Economic Growth, certification dated Aug 27, 2024, [226740/Z/22/Z]. Data Climate vulnerability data. District-wise scores on climatic vulnerability were sourced from the dataset on risk and vulnerability assessment of Indian agriculture to climate change published by the. This dataset used fifteen indicators capturing five dimensions of capital endowment- natural, human, social, physical and financial, to develop a ranking on climatic vulnerability across rural districts in India. Districts are classified into five levels of vulnerability as having very high, high, medium, low or very low levels of vulnerability. Fig 1 shows the distribution of climatic vulnerability across districts in India. Rajasthan has the highest number of districts which are either very highly or highly vulnerable, followed by Jharkhand, Chhattisgarh and Uttarakhand among the larger states. In absolute terms, the area, climatic conditions, geographic

features and number of districts vary widely across states. It is revealing that the extent of vulnerability to climate change is not clustered uniquely by geographical location. Within a state, the vulnerability to climate change can vary substan- tially across constituent districts. For instance, among the larger states, districts in Gujarat, West Bengal, Uttar Pradesh and Uttarakhand have high within state variability in the distribution of districts by levels of climatic vulnerability. Several districts in the North Eastern region of the country have high levels of climatic vulnerability, with all districts in Arunachal Pradesh being highly vulnerable. (S1 Fig. provides the distribution of very high and highly vulnerable districts across states in India). The percentage distribution of observations in the merged dataset is such that there are relatively very few observa- tions falling in the very highly vulnerable districts or the very low vulnerability districts. (S1 Table 1 provides the distribu- tion of observations by climatic vulnerability levels). It is to be noted that for the econometric estimations, the climatic