The relationship between climate change and malaria in South-East Asia: A systematic review of the evidence

Search strategy

The conduct of this study was guided by the PRISMA 2020 checklist for review studies^13,14 and registered in PROSPERO: CRD42022301826 on 5^th February 2022. One Indonesian database (Garuda) and three international databases (PubMed, SpringerLink, ProQuest and Scopus) were searched to gather peer-reviewed articles for this review of evidence. The Indonesian database was included to capture additional articles catered in the authors’ local language as previously done in Babaie et al. (2018)¹⁵ with the inclusion of Persian databases for a review of Iran and Fischer et al. (2020)¹⁶ with the inclusion of German and French articles for a review of Europe. The search strategies applied were deployed in multiple sequences to mitigate any biases arising from missed articles from each database. The articles were then immediately disbursed amongst authors following selection. The keywords for our search included the terms ‘climate’, ‘iklim’, ‘malaria’ and ‘Southeast Asia’ or the 10 Southeast Asian countries searched as ‘Indonesia’, ‘Malaysia’, ‘Singapore’, ‘Philippines’, ‘Thailand’, ‘Myanmar’, ‘Laos’, ‘Cambodia’, ‘Vietnam’ and ‘Brunei’. The search terms and methods along with the appropriate incorporation of truncations and operators specific to each database was discussed and consulted between DS and AR. Our search strategy did not include any limitations on publication periods. The search methods used and respective retrieved results from each database used are detailed as follows:

Inclusion and exclusion criteria

Original articles written in Indonesian and English which were analytical ecological studies and utilized longitudinal or time-series data of climatic variables with malaria incidence and/or prevalence in regions across Southeast Asia were included. Additionally, the articles included were studies which quantitatively analyzed the data through correlation, regression and/or mathematical models to infer relationships between multiple meteorological measures to reflect climate change and malaria incidence and/or prevalence. The studies that did not include data from a multi-year period and quantitative models with only one climatic variable were excluded.

Study selection

Three reviewers were involved in this systematic review of evidence with discussions and decisions conducted online. The selected papers were systematically reviewed thematically, and their methodologies assessed by AR with independent verifications by TF who also ensured no relevant articles were missing in the systematic review. Then, eligibility of the full-text records following screening of abstracts was conducted by AR and further corroborated by DS and TF. The three authors were familiar with reviewing and presenting descriptive assessments of quantitative results which should minimize the potential bias arising from reporting conflicting results. Any differing judgements on the selection of articles and extraction of results were resolved with the expert verdict of DS who also gave the final confirmation on credibility of the synthesis with regards to the climatic variables’ influence on malaria dynamics. AR compiled the retrieved and summarized data of studies selected from the eligibility assessment stage into a separate review table in a shared Microsoft Word document (Version 2207 Build 16.0.15427.20182). DS and TF worked independently in corroborating the extracted data and narrative outlined in the shared review table. The systematic review flowchart to document our research process according to PRISMA conventions is shown in Figure 1.

Figure 1. PRISMA review flowchart.⁴⁵

Data extraction

Data was extracted from the studies that met the eligibility conditions. The studies were imported from their respective online databases and managed with the reference management software Zotero (Version 6.0.13, RRID: SCR_013784). The information included in the extractions were title, author and year of publication, location of study, period of study, independent and dependent variables, the scale and measure of the variables, quantative analysis methods, and summary of results. These extracted data were retrieved and summarized for qualitative synthesis within a table in a shared Microsoft Word document that was independently corroborated by the reviewing authors.

Data synthesis

The study utilizes a narrative synthesis method to describe the overarching influence of climatic variables indicated by meteorological measures such as temperature, humidity, wind speed and precipitation on malaria prevalence and/or incidence in countries across Southeast Asia. The strength and direction of the influence between variables is derived from the outcomes and summaries of the quantitative correlation, regression or mathematical models presented in the reviewed studies. Every climatic variable found throughout the review is described in the narrative synthesis irrespective of the number of studies that include them for the quantitative models. The reported outcomes of studies included in the final review are tabulated to group them according the the temperature, precipitation, humidity, and windspeed indicators as a meteorological proxy variables of climate change against their respective effects on malaria. Any incomplete reported outcomes in statistical tables found within the reviewed articles were thoroughly searched in the textual section of the articles with corroborations from the three authors. Categorized tabulations of the summarized articles were made using Microsoft Word’s table features and shared amongst the three authors concurrently.

Temperature and malaria (19 studies)

Temperatures across different locations in Southeast Asia proved to be a worthy inclusion for analyzing the relationship with malaria cases based on the reviewed studies. Both Anopheles mosquitoes as the vector for malaria and the parasitic agent, Plasmodium group, rely on optimally warm temperatures.^21–27 The optimal temperature for Anopheline reproduction according to Mau et al. (2020) is between 25–27°C whilst the Plasmodium group’s extrinsic cycle is optimum within the 20–30°C range.²² This implies that as temperatures become warmer within the optimum range, the duration of incubation is shortened, and mosquitoes become infective much sooner.^22,24,26

That said, the quantitative analyses conducted in the reviewed studies exhibited varying results with temperature variables being averaged over a year, a month, split into maximum and minimum ranges as well as added lagged variations. For instance, in Kiang et al.'s (2006) neural network analysis, the composite use of mean monthly temperature with other climatic variables and vegetation index resulted in a model configuration to assess malaria cases with a training accuracy of 73% and testing accuracy of 53%. The model was developed based on variable data from 19 provinces across Thailand throughout a seven-year period (1994–2001).

Meanwhile, a study of the Koh Chang district in Thailand throughout 2001–2011 indicates that maximum temperature and mean temperature are positively correlated with malaria cases (r = 0.150 and r = 0.190 at α ≤ 0.01) by one year according to the count regression models.²⁹ Similarly, Rejeki et al. (2018) also utilized count models in which maximum and minimum temperatures were included with the addition of lagging by one, two, three and 12 months. The results of their baseline Poisson model suggest that minimum and maximum temperatures have a significant influence on monthly malaria cases at the negative and positive directions respectively. However, after the inclusion of a dispersion parameter and testing for fit, a negative binomial model was selected in Rejeki et al. (2018) in which maximum and minimum temperatures exhibited no significance at α ≤ 0.05.

In Mau et al. (2020), temperature was also shown to have a significant influence on malaria cases in their linear regression model. The difference being those temperatures were averaged over a year and negatively influenced malaria cases, which were measured as annual parasite incidence (API). The results in Rejeki et al. (2018) and Mau et al. (2020) suggest that further increases of temperatures beyond certain levels will be associated with reductions in malaria cases – in line with the optimum range for both Anopheline reproduction and Plasmodia incubation. In the Rejeki et al. (2018) Purworejo study, maximum temperatures recorded in the study period ranged between 28–30°C,²⁴ while the average temperatures between 2013–2019 at the study site in Mau et al. (2020) ranged between 25.13–27.58°C (which is already the optimum levels for Anopheline reproduction). Interestingly, the studies of Ninphanomchai et al. (2014) and Servadio et al. (2018) have offered their explanatory evidence through a non-linear thresholding effect with spline-fitted models that establish a peak positive effect of temperature at around 30°C.^29,30

Moreover, another alternative take on the temperature-malaria case relation is found in studies by Bui et al. (2011), Mercado et al. (2019) and Noppradit et al. (2021) who similarly argue that other factors confound the role temperature plays in variations of malaria cases. In their study of 670 districts across Vietnam between 2007–2008, Bui et al. (2011) found that temperature variables have district specific effects with many regions exhibiting opposite interactions with malaria cases, i.e., in some regions malaria cases soar as temperatures rise while others reduce with increased heat. The same is stated in Kotepui and Kotepui (2018) and Noppradit et al. (2021), where the latter argues that the lack of statistical significance between malaria and temperature variables in their study was due to topographic factors of the study site which were unfavorable to malaria from the commencement of the study.

Precipitation and malaria (19 studies)

The next climatic variable analyzed for its association with malaria in the reviewed studies was precipitation. Although the amount of rain does not directly affect both vector and parasite proliferation, low to medium intensity rain creates reservoirs in the form of pools and puddles for Anopheles to breed.^21–27 Aside from forming breeding sites, precipitation also leads to increased relative humidity which prolongs the age of infective adult Anopheles mosquitoes as explained in the following section.^22,24–26

Across the 19 reviewed studies, the measures of precipitation as an independent variable varies. A mean yearly precipitation is used in Sandy & Wike (2019) and Mau et al. (2020) for their study in Indonesia, Gallalee Sarah et al. (2021) in Myanmar, and Noppradit et al. (2021) in Thailand while mean monthly precipitation is used in Kiang et al. (2006), Jeefoo et al. (2009), Bui et al. (2011), Ninphanomchai et al. (2014), Nurmala (2017), Rejeki et al. (2018), Rejeki et al. (2019), Mercado et al. (2019), and Wangdi et al. (2020). Similar to the latter articles, Jubaidi (2015) also includes mean monthly precipitation in addition to the number of rainy days per month. Suwito et al. (2010) on the other hand uses a precipitation index. They defined the index as the product of the sum of precipitation and number of rainy days within a given month that is then divided by the total number of days in said month.²⁷

The relationship between precipitation between malaria cases across the reviewed studies was inconclusive as the resultsof some studies did not exhibit statistical significance for the variable.^{30,32,33,35,38} The study by Suwito et al. (2010) did not statistically associate precipitation index with malaria cases, but their resulting correlation between the former with Anopheles density measured by man-biting rate (MBR) exhibited a positive relationship that was statistically significant. As for direct associations with statistical significance, the correlation and linear regression results in Mau et al. (2020) showed that yearly precipitation has a positively linear relationship with API where their linear regression model implies that for every 1 mm increase in precipitation, API cases increase by 0.028. The negative binomial regression results in Rejeki et al. (2018) also indicate that a 1 mm increase in precipitation has a positive influence on malaria cases by 0.08% and 0.09% with the use of three-month and 12-month lags respectively. However, in contrast to the three previous studies, results in Jubaidi (2015) showcases negative correlations that are statistically significant between monthly precipitation and number of rainy days with monthly malaria incidence (r = -0.431, α ≤ 0.01 and r = -0.349, α ≤ 0.05 respectively). Taking into account the previously mentioned studies, the results in Jubaidi (2015) indicate the need to take precipitation as a lagging indicator for direct associations with malaria cases as suggested in Kim et al. (2012), Krefis et al. (2011), and Wu et al. (2017).^39–41

Humidity and malaria (14 studies)

As previously mentioned, humidity as a meteorological measure of climate is an important indicator for malaria cases as it enables the lifespan of an infective adult Anopheles mosquito – where relative humidities of at least 60% and above optimize the mosquitoes’ activity to bite and infect.^23,24,26,27

Only seven of the reviewed studies had results where humidity was a statistically significant independent variable. That said, the results were inconclusive. In the negative binomial model of Rejekti et al. (2018), results suggest that a 1% increase in maximum relative humidity is associated with a 10.47% decrease in malaria cases after two months. However, the results of Suwito et al. (2010) instead further validate the influence of humidity on Anopheles activity to bite and infect as average humidity is shown to have a statistically significant positive correlation with MBR. Narrative background for humidity posed by the other reviewed studies, where previous prevailing studies are also referred to, would suggest the latter study as being the sounder evidence for humidity’s relationship with malaria cases (albeit indirectly). The authors in the former study unfortunately did not provide any theoretical explanations as to why their two-month lagged maximum humidity was associated with a decrease in malaria cases. However, an argument could be made regarding the range of the maximum humidity throughout the period and location of study. In Purworejo, the maximum humidity between 2005–2014 varied between 83–99%, which is well beyond the 60% necessary optimum for Anopheles mosquito activity. Alternatively, the results in Rejeki et al. (2018) could suggest that the use of humidity as a lagged indicator associated with malaria cases is unwarranted.

Windspeed and malaria (Two studies)

Only two of the reviewed articles included windspeed as a climatic variable that was assessed for its relationship with malaria cases.^21,26 Based on references included in their article, Sandy and Wike (2019) state that windspeed has an influence on Anopheles mosquitoes’ flight range, hence enabling an expanded scope of humans to bite. That said, the evidence in both Jubaidi (2015) and Sandy and Wike (2019) indicate that the resulting correlation between windspeed with monthly malaria incidence and API respectively did not exhibit statistical significance. The authors in both attribute the narrow windspeed range across the periods and their respective locations of study as the potential reason for the variable not resulting in statistically significant correlations.

Underlying data

Figshare: Underlying data for ‘The relationship between climate change and malaria in South-East Asia: A systematic review of the evidence’. https://doi.org/10.6084/m9.figshare.20697298.v1.²⁰

Reporting guidelines

Figshare: PRISMA checklist for ‘The relationship between climate change and malaria in South-East Asia: A systematic review of the evidence’. https://doi.org/10.6084/m9.figshare.20489235.v1.¹³

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).