Research Article: Exploring the Seasonality of Reported Treated Malaria Cases in Mpumalanga, South Africa

Date Published: October 29, 2013

Publisher: Public Library of Science

Author(s): Sheetal Prakash Silal, Karen I. Barnes, Gerdalize Kok, Aaron Mabuza, Francesca Little, Clive Shiff.

http://doi.org/10.1371/journal.pone.0076640

Abstract

South Africa, having met the World Health Organisation’s pre-elimination criteria, has set a goal to achieve malaria elimination by 2018. Mpumalanga, one of three provinces where malaria transmission still occurs, has a malaria season subject to unstable transmission that is prone to sporadic outbreaks. As South Africa prepares to intensify efforts towards malaria elimination, there is a need to understand patterns in malaria transmission so that efforts may be targeted appropriately. This paper describes the seasonality of transmission by exploring the relationship between malaria cases and three potential drivers: rainfall, geography (physical location) and the source of infection (local/imported). Seasonal decomposition of the time series by Locally estimated scatterplot smoothing is applied to the case data for the geographical and source of infection sub-groups. The relationship between cases and rainfall is assessed using a cross-correlation analysis. The malaria season was found to have a short period of no/low level of reported cases and a triple peak in reported cases between September and May; the three peaks occurring in October, January and May. The seasonal pattern of locally-sourced infection mimics the triple-peak characteristic of the total series while imported infections contribute mostly to the second and third peak of the season (Christmas and Easter respectively). Geographically, Bushbuckridge municipality, which exhibits a different pattern of cases, contributed mostly to the first and second peaks in cases while Maputo province (Mozambique) experienced a similar pattern in transmission to the imported cases. Though rainfall lagged at 4 weeks was significantly correlated with malaria cases, this effect was dampened due to the growing proportion of imported cases since 2006. These findings may be useful as they enhance the understanding of the current incidence pattern and may inform mathematical models that enable one to predict the impact changes in these drivers will have on malaria transmission.

Partial Text

Despite being a treatable and preventable mosquito-borne disease, malaria is still an immense global health, economic and social burden. In 2010, latest estimates suggest 219 million cases with an uncertainty range of (154 million, 289 million) cases globally. There were 660 000 deaths due to malaria; with 90% of deaths occurring in Africa and most cases and deaths occurring in sub-Saharan Africa [1]. Malaria has been recognized as a disease of poverty with vulnerable groups facing several barriers to access for antimalarial interventions [2], [3]. South Africa, having experienced a sharp decline in malaria cases since the last epidemic in 2000, already meets the pre-elimination phase criteria set out by the World Health Organisation (WHO) () and has been ear-marked to achieve elimination by 2018 [4]. Mpumalanga is one of three provinces in South Africa where malaria transmission still occurs. Malaria in Mpumalanga is seasonal, starting with the first rains in October, peaking in January and remaining high till May; yet transmission is still unstable and prone to sporadic outbreaks. As government begins to intensify efforts and commit scarce resources towards malaria elimination, there is a need to understand patterns in transmission so that efforts may be targeted appropriately. Mathematical modeling is increasingly being used to test policy interventions so as to determine their impact on simulated transmission before implementing the intervention in the field. Understanding the nature of the seasonality of transmission will enable better mathematical modeling and this may lead to better allocation of scarce resources and ultimately a greater impact on malaria. This paper aims to explore the seasonality of malaria cases in the Mpumalanga province as part of a larger project in mathematical modeling of malaria transmission and the impact of antimalarial interventions. We analyse data on reported treated malaria cases from 2002 to 2012. In particular, this paper explores the temporal and geographic behavior patterns as well as potential drivers behind these patterns.

The analysis for the case data is presented first followed by an assessment of the relationship between cases and rainfall, cases and physical geography and thirdly, cases and source of infection.

Malaria transmission in Mpumalanga is characterized by a triple-peak in the season where the first peak occurs in September/October, the second (and also main) peak occurs in January and the third peak occurs in April/May. Assessing the STL seasonal components of the source and geographical location of infection shows that the first peak is driven by mainly locally-sourced infections and the second and third peaks are driven mainly by foreign-sourced infections. This is supported by the Maputo data where the peak of the season in Maputo corresponds to the second and third peak of the Mpumalanga season. Geographically, Bushbuckridge, the northern most municipality, contributes specifically to the first and second peaks, while the other two municipalities studied contribute equally to all three peaks. Monthly rainfall lagged at 1 month was found to be significantly associated with monthly cases, yet the seasonal pattern of rainfall does not appear to characterize the triple peak pattern of the case data. The decline of local cases in Mpumalanga may be a reason for this, where cases from a foreign source are more likely to be affected by rainfall in the source area (e.g. Mozambique) than local rainfall. If the increase in the proportion of foreign-sourced cases persists, rainfall may become even less of a driver for malaria. Having reduced its malaria burden significantly since 2002, South Africa has embraced a malaria elimination target of 2018 [4]. To achieve this, suites of malaria control interventions will be deployed in Mpumalanga and other provinces affected by malaria. The likely impact of these interventions may be measured through mathematical modeling. The MalERA Consultative Group on Modeling has recognized the contribution mathematical modeling can make to the elimination of malaria globally and has developed a framework of priority areas for modeling to assess and inform: optimal resource allocation, strategies to minimize the evolution of drug and pesticide resistance, new tools to interrupt malaria transmission, combinations of tools, coverage targets and expected timelines to achieve goals as well as to assess operational feasibility with respect to costs and human resource capacities [28]. Using mathematical modeling to adequately represent malaria transmission requires knowledge on the drivers of the geographic and temporal trends in malaria transmission and this in turn may lead to finding malaria control strategies that target these drivers directly rather than strategies that may generally control transmission but not interrupt it. The primary interest of mathematical modeling in this setting is to provide practical guidance to malaria programme managers on how to conduct more efficient and effective control and elimination activities. This paper assesses the seasonal trend of cases for different sub-groups of the population, based on the source of infection and geography. If an elimination activity such as the scale-up of larviciding is under consideration, knowing the seasonal pattern of locally sourced cases (larviciding impacts locally sourced cases directly) allows managers to optimally time the larviciding activities with the breeding patterns of the mosquitoes. While malaria elimination requires the reduction to zero of locally sourced cases only, interventions aimed at reducing foreign sourced cases can reduce onward transmission. Knowing the seasonal pattern of foreign sourced cases allows programme managers to optimally time interventions such as mass screen and treat campaigns at border posts with travel/migration patterns. Understanding the seasonal differences between managerial districts like municipalities can also assist with optimal allocation of drugs and staff. These are some of the many practical uses of mathematical modeling in health management. The rising percentage of imported or foreign-sourced cases suggests the need for intensive monitoring. The second and third peaks of the season (which comprise mainly foreign-sourced cases) correspond to holiday seasons (Christmas and Easter). As the majority of cases are sourced from Mozambique, there is a need to intensify cross-border collaborations now and even more so after elimination has been achieved in South Africa. This increase in foreign-sourced cases also highlights the importance of including migration in mathematical models of malaria transmission to obtain realistic estimates of the impact of malaria control and malaria-elimination focused strategies.

Mpumalanga province in South Africa has experienced a 65% decrease in reported treated cases since 2002. In this time, the percentage imported cases of this total has increased from 39% to 87%. Using weekly reported treated cases, the geographic and temporal trends in malaria transmission was explored to reveal an atypical triple-peak pattern in cases with a short period of few cases. This seasonality was explored in relation to rainfall, source of infection and geographic location. A cross correlation analysis revealed that one month lagged rainfall was significantly associated with reported treated cases, but the seasonality of rain did not appear to explain the unusual pattern of malaria cases. Analysis of the source of infection revealed that local cases contribute to all three peaks but foreign cases contribute primarily to the Christmas and Easter peaks. Geographically, Bushbuckridge municipality (northern most municipality) had a greater relative contribution of cases in the first (September/October) peak. Malaria transmission in Mpumalanga may be low, but it is also unstable and a change in climate and the source of infection may lead to a spike in infection and generally a modification of the incidence pattern. These findings may be useful as they enhance the understanding of the current incidence pattern and can be incorporated in mathematical models that enable one to predict the impact changes in these drivers will have on malaria transmission. Further, as mathematical modeling is being used to assess timeframes for malaria elimination and the potential impact of elimination-focused interventions, understanding the seasonal trends of malaria is key to designing a targeted temporal and spatial approach that can be applied in resource-scarce settings.

 

Source:

http://doi.org/10.1371/journal.pone.0076640