Malaria Risk and Vulnerability of Communities Due to Climate Change in India


 Background: Malaria is an erstwhile public health problem that is further compounded in India due to diverse climatic conditions, ecology, socio-developmental indicators and the imminent threat of climate change. Mapping of malaria risk using climatic and ecological factors has been assessed to some extent but the vulnerability by taking into consideration the human factors i.e. socio-economics and adaptive capacity, in climatically vulnerable areas is not known. The present communication deals with determining the risk and vulnerability at district level by taking into account the hazard due to climate change, population exposed, ecological sensitivity and the community’s adaptive capacity.Methods: Based on the IPCC framework, relevant indicators for the dimensions of risk (Hazard, Exposure, Sensitivity, Adaptive Capacity) were identified from literature and was used to develop indices. Data for each indicator was obtained through government websites. Weights for different indicators were assigned using principle component analysis and the indices were aggregated to develop a district wise risk and vulnerability profile of India.Results: Malaria risk was found to be highest in Bihar (Ri=0.7) due to very low adaptive capacity coupled with high sensitivity. In some states with high climatic hazard and/or sensitivity for malaria, such as Tripura, Mizoram, Kerala and Tamil Nadu, high Adaptive capacity acted as a buffer, reducing the overall risk of malaria. Further, malaria risk is projected to be introduced for the first time in some districts of Himachal Pradesh and Arunachal Pradesh by 2030s due to the effects of climate change. Conclusion: The results of the study highlight the importance of various socio-development indicators in representing a holistic view of malaria risk. Adaptive capacity emerged as an important index as it could help counter the effects of climate change on malaria by enabling the community to cope with the adverse effects.


Introduction
India is the third largest contributor to the global malaria burden after Nigeria and Democratic Republic of Congo, accounting for 6% of the global malaria cases [1]. In recent years, India has successfully reduced its malaria morbidity and mortality signi cantly, and is moving towards elimination. However, in a large country like India, there are diverse climatic conditions, ecology and socio-developmental indicators. In addition to this, climate change is posing a major threat by affecting the distribution of malaria [2][3][4][5].
Mapping of risk using climatic and ecological factors has been assessed to some extent and opening of new transmission windows in some parts of the Himalayan region as well as reduction in malaria intensity in hotter regions by 2030s has been projected [6,7]. However, the vulnerability to malaria by taking into consideration the human factors i.e. socio-economics and adaptive capacity, in climatically vulnerable areas in India is not known. A comprehensive risk assessment should, therefore, incorporate all the socio-economic factors and demographic pressures that in uence local vulnerability to malaria, such as the framework developed by the Intergovernmental Panel on Climate Change (IPCC) which identi es risk as a factor of the hazard, exposure and vulnerability [8][9][10].
The present study was, therefore, undertaken to assess the social impact of malaria taking into account the multitude of factors that shape malaria risk in India. The effect of future climate change (2030) on malaria, at present vulnerabilities was also assessed, to ascertain its impact in the future if present conditions are not alleviated. The malaria risk and vulnerability pro le developed in the study can be used to develop appropriate adaptation measures that are based on local vulnerabilities tailored to each district.

Methodology
An indicator-based approach for risk assessment was used in the study. This approach combines a variety of different indicators of risk into meaningful indices. Such an approach incorporates all three dimensions of risk de ned by the IPCC-hazard, exposure and vulnerability.

Risk Assessment Framework
The IPCC de nes climate change risk as a product of probability of hazard due to climate change, the degree to which populations are exposed to the risk and vulnerability of the local communities.

Selection of Variables
An extensive literature survey of variables used in malaria risk and vulnerability assessment studies was conducted, and 18 indicators were shortlisted for the study. The variables selected represent the climatic and ecological risks to malaria, socio-economic well-being, health and nutritional status, education, infrastructure and demography. Data for each of these indicators was either available in the public domain or procured from various government organizations.
The shortlisted indicators were classi ed based on the de nitions of hazard, exposure and vulnerability as described by the IPCC [12,13]. A list of all the shortlisted indicators and their sources is given in Table 1. Transmission windows for malaria which have been determined in a recent study using the CORDEX model [6], were used as the indicator for hazard. Baseline and projected (2030) transmission windows of malaria at moderate climate change impacts (RCP 4.5) using temperature and relative humidity were used.
The indicators used to determine the degree of exposure to malaria risk were population density, percentage of children below 6 years and adults above 60 years, who are likely to have reduced immunity [23].
The vulnerability to malaria depends on the intrinsic sensitivity of the community and its ability to cope with and adapt to the changes. Sensitivity to malaria was further categorized into biological (API, malnutrition), ecological (forest cover, NDVI, soil moisture, net irrigated area) and demographic (material of wall and roof, livestock per 1000 people). The adaptive capacity of the communities was based on the socio-economic status (Literacy, per capita income), availability of resources (% households (HH) with health insurance, % households lacking essential assets) and robustness of health infrastructure (Health Centre population burden).

Data Processing
The raw data was analysed for missing information and potential outliers. Missing data were completed through interpolation and outliers were treated with 98% winsorization. The indicators were tested for multicollinearity through Pearson's correlation factor, with a value above 0.9 indicating high collinearity [24]. The indicators were then normalized using the standard formula for linear transformation: X s = (x - x min )/ (x max -x min ).

Assigning Weights
Principal component analysis (PCA) was used to assign weights to the indicators which re ect the largest variations. Kaiser-Meyer-Olkin (KMO) value of 0.762 indicated high usability of the PCA test (> 0.6). Weights were calculated by multiplying the square of the rotated factor loadings with the proportion of the explained variance for each component (Table 2) based on the method described by Organization for Economic Co-operation and Development (OECD [24]. where 'x' is the indicator and 'w' is the respective weight. The vulnerability index was calculated using the formula: V i = [ S i + (1 -A i)] / 2, where V i is the vulnerability index, S i is the sensitivity index and A i is the adaptive capacity index. Thereafter, the Risk index was calculated using geometric summation: . Geometric summation was used for the nal risk index in order to account for the fact that if the value any of the three dimensions of risk (Hazard, Exposure and Vulnerability) was 0, the overall risk will also be 0. All indices were transferred to ArcGIS 10.2.1 and district-wise raster surface was created for each index.

Multicollinearities and Validation
Correlation between each of the 18 indicators was less than 0.9 indicating no multicollinearities. Furthermore, the Variance In ation Factor (VIF) was below ve which con rms that there were no errors due to multicollinearity.

Effect of Climate Change Hazard on Malaria
A recent study on the climatic suitability for malaria in India and the impacts of climate change projected an increase in malaria transmission by 2030s in the states of Rajasthan, Maharashtra, Assam, Meghalaya, Tripura and Mizoram and new foci of transmission in some districts of Himachal Pradesh and Arunachal Pradesh [6]. Climate suitability for malaria is projected to reduce in the states of Chhattisgarh, Gujarat, Madhya Pradesh and Bihar.

Degree of Exposure to Malaria
The results show that more population is exposed to malaria risk in the states of Chandigarh (Ei = 0.9), Delhi (0.85) and Bihar (0.61). When the indicators for exposure in these states were compared, it was found that in Delhi and Chandigarh population density was very high. On the other hand, Bihar had the highest proportion of children in the population (18.5%).
In the states of Sikkim, Telangana, Goa, Manipur, Himachal Pradesh and Arunachal Pradesh exposure is lowest (< 0.2). The indicators that contribute the most to low exposure in these states were found to be low population density (< 350 people per sq. km) and proportion of children (~ 10%) in the population. States, such as Bihar (Vi = 0.8), Uttar Pradesh (0.72), Meghalaya (0.61), and Jharkhand (0.66), which are highly sensitive with low adaptive capacity are the most vulnerable to malaria risk. In contrast, states that have high adaptive capacity and are less sensitive to malaria risk are the least vulnerable -Sikkim (0.25), Chandigarh (0.04), Himachal Pradesh (0.29), Goa (0.1), Delhi (0.1), Kerala (0.12). In Jammu & Kashmir and Rajasthan the sensitivity to malaria was relatively low (0.49 & 0.52 respectively), however, due to lack of adaptive capacity the overall vulnerability to malaria risk was high (0.6 and 0.56 respectively). In some states, such as Tripura (0.36), Puducherry (0.26) and Mizoram (0.35), both adaptive capacity and sensitivity were found to be high. Consequently, the vulnerability to malaria risk was found to be only moderate.

Composite Malaria Risk due to Climate Change
The results show that risk for malaria is highest in Bihar (Ri = 0.7), with almost all districts having a malaria risk index higher than 0.6, followed by the states of Meghalaya, Jharkhand, Assam and Odisha with more one district having very high malaria risk (> 0.6). Malaria risk is lowest in the states of Sikkim (0.01), Chandigarh (0.03), Himachal Pradesh (0.04), Goa (0.05), Delhi (0.08) and Kerala (0.09).
Climate change is projected to reduce malaria risk marginally in all the districts of Chhattisgarh, Andhra Pradesh and Haryana, as well as some districts of Gujarat, Rajasthan, Punjab, Jharkhand, Odisha, Madhya Pradesh, Uttar Pradesh and Uttarakhand by the 2030s. On the other hand, malaria risk is projected to increase signi cantly in the states of Meghalaya, Tripura and Manipur, and marginally in Rajasthan, Gujarat, Maharashtra, Himachal Pradesh, Telangana and West Bengal due to the effects of climate change by 2030s. In Odisha, malaria risk is projected to reduce in coastal districts and increase in districts away from the coast. Furthermore, in some districts such as Chamba in Himachal Pradesh and West Kameng in Arunachal Pradesh, that are presently malaria-free, climate change is projected to open new windows for transmission by 2030s. In the southern states such as Kerala and Tamil Nadu, malaria risk is projected to remain constant.

Discussion
Risk and vulnerability assessments have been used widely to determine the consequences of various stressors on the social and economic well-being of the communities exposed to the risk. Several studies have adopted this framework to evaluate the risk of climate change on farming communities [8,[25][26][27], coastal regions [28], disaster prone areas and health sector [29][30][31]. A few studies have also assessed the multi-factorial risk of malaria based on the risk and vulnerability assessment framework [32][33][34]. However, no study has adopted this approach to assess the social risk of malaria with respect to climate change in India.
The results of the study reveal that high malaria risk is found in the north-eastern and eastern parts of India as well as in some parts of the southern states of Maharashtra and Karnataka. High malaria risk in the north-eastern states can be attributed to high forest cover, poor housing materials, low per capita income, high burden of health centres and a large share of households that lack essential assets (such as telephone, computer, television etc.), which is coupled with a high climatic suitability for malaria. A risk and vulnerability assessment study undertaken in Rwanda also found that indicators such as poor housing wall materials, television ownership, poverty rate and clinic density had an excessive in uence on the vulnerability index [32]. In the present study, malaria risk was found highest in the states of Bihar, Uttar Pradesh and Jharkhand, which was mostly a result of low adaptive capacity and high sensitivity to the climatic hazard of malaria in these states. The strong in uence of adaptive capacity was also observed in the states of Rajasthan and Jammu and Kashmir where vulnerability is signi cantly high despite low sensitivity to malaria, as a result of low adaptive capacity. Previous research conducted in Tanzania similarly reported that regions of low adaptive capacity had high vulnerabilities even when susceptibility to malaria was low [34]. The southern states of Kerala and Tamil Nadu were found to have signi cantly low risk of malaria despite very high suitability. High adaptive capacity in this region acts as a buffer to high malaria hazard, thereby reducing the composite risk of malaria. This characteristic of the adaptive capacity dimension has been observed in a similar study undertaken in Rwanda, where high malaria suitability in the eastern lowlands was counter balanced by good socio-economic factors [35]. Climate change is projected to increase malaria risk by 2030s in the north-eastern states where adaptive capacity was low. On the other hand, in districts with high adaptive capacity, climate change had only a marginal or no impact on malaria risk.
Comparison of the malaria risk map with the map of malaria endemicity (Fig. 8) provides an insight into the implications of malaria risk. It can be observed that the study correctly identi es the north eastern and eastern regions of India to be at higher risk for malaria, albeit with a few minor deviations. Malaria risk has been found to be highest in the state of Bihar, even though it has an API less than 0.1. A Joint Monitoring Mission for vector borne diseases conducted by NVBDCP and the World Health Organization (WHO) in 2014 reported signi cant de ciency in malaria surveillance in Bihar due to a large number of vacancies in eld and supervisory staff and an Annual Blood Examination Rate (ABER) less than 1 since 2003 (desired ABER = 10) [36]. This could be the reason for low API despite a very high malaria risk in Bihar. In Odisha, malaria risk appears to be understated in a few districts, when compared to the number of reported cases of malaria (Fig. 8). When observing the effect of individual factors, it is seen that low forest cover, low population density and high availability of essential assets could account for the lower risk in some districts of Odisha.
The results of the study are signi cantly dependent on the selection of suitable indicators due to which it is essential to discuss the relative importance of the indicators used. Most of the indicators used in the present study have also been used in similar risk and vulnerability assessments in the past [32-35, 37, 38]. One such study, conducted in East Africa, used elevation as a proxy indicator for immunity to malaria, and found that this resulted in high vulnerability in regions that are malaria free or have epidemic malaria [33]. However, this was not deemed as an important indicator in the present study. In another malaria risk and vulnerability assessment, undertaken in Tanzania, a notable similarity between the malaria risk and malaria hazard maps was observed [34]. This was attributed to the high weightage of the hazard variable used in the study (Entomological Inoculation Rate). However, such a similarity in the malaria hazard and malaria risk maps is not discernible in the present study even though the weight assigned to the hazard variable (Transmission window) was signi cantly high (0.25). For example, in the southern states of Kerala and Tamil Nadu very high malaria hazard does not lead to a signi cant malaria risk as it is offset by very high adaptive capacity. Therefore, adaptive capacity was found to bear greater in uence on the composite malaria risk as opposed to the hazard dimension, which was more signi cant in previous research [34]. This may be attributed to the fact that most parts of Tanzania were found unsuitable for malaria, due to which malaria hazard served as a limiting factor for malaria risk. On the other hand, in India most of the regions are climatically suitable for malaria and adaptive capacity serves as a limiting factor for malaria risk. This has important implications for malaria elimination as it demonstrates that targeting these socio-economic factors can prove crucial in attaining as well as sustaining malaria elimination.
The study highlights the importance of considering social and economic indicators to present a holistic view of malaria risk. While the hazard and exposure dimensions of risk cannot be changed, vulnerability of the communities to the malaria risk are dependent on several demographic and socio-economic indicators that are controllable. This includes indicators of sensitivity and adaptive capacity such as material of roof and wall of houses, literacy rate, per capita income and Health Centre population burden. Elimination of malaria may pose a larger challenge in regions with lower adaptive capacity as compared to regions with higher adaptive capacity. Therefore, targeting these indicators will be crucial in order to eliminate malaria and prevent any possibility of resurgence

Conclusions
In this study a number of meaningful indicators for malaria were used to construct an index, based on the IPCC approach, to assess the impact of malaria on the social well-being of communities in India and the effects of climate change.
The results highlight the importance of adaptive capacity, which can act as a buffer to reduce malaria risk when climatic hazard and sensitivity is high. The converse is also observed i.e. moderate suitability for malaria could have adverse impacts on the community if the adaptive capacity is poor. Therefore, the social impact of malaria can be different to its epidemiology due to varying degrees of adaptive capacity. Furthermore, the risk and vulnerability assessment captured the high malaria risk in Bihar, which has been neglected due to poor surveillance. In regions where malaria risk is projected to be introduced by 2030s, targeting low performing indicators can help increase resilience and therefore, mitigate the impacts of climate change.

Consent for Publication
Not Applicable.

Availability of data and materials
All datasets used in the study were obtained from the public domain. The vegetation index and soil moisture data that support the ndings of the study are available in the NOAA Climate Data Record (https://doi.org/10.7289/V5ZG6QH9) [14] and the ESA Climate Change Initiative (https://doi.org/10.1016/j.rse.2017.07.001) [15] repositories respectively. Data on transmission windows for malaria was obtained from previous research [6] conducted in the same institute, while the malaria API data was collected from respective State Programme O cers and District Malaria O cers, and can be made available on reasonable request. The data for all other variables used in the study are available either in the 2011 Census of India (https://censusindia.gov.in/digitallibrary/TablesSeries2001.aspx), or in governmental and institutional reports that do not have a doi, the details of which are mentioned in Table  1.

Competing Interests
The authors declare that there are no competing interests.