Travel patterns and demographic characteristics of malaria cases in Swaziland, 2010–2014

Global malaria eradication is back on the international agenda [1, 2], with many countries making malaria elimination a national goal [3]. The past decade has seen significant declines in malaria prevalence [4, 5], with the risk of acquiring malaria reduced by 37% since 2000 and the risk of dying from the disease decreasing by 60% [6].

Achieving elimination requires a re-orientation from the universal prevention and treatment measures that typically define a control or surveillance programme towards targeted operations, such as identifying residual transmission foci, identifying and curing both asymptomatic and symptomatic infections, focusing vector control or parasite-based attack measures to high-risk areas and managing importation risk. Understanding human movement, which can provide connections between disparate high transmission or receptive areas [7], is important for designing appropriate elimination strategies and avoiding resurgence in post-elimination settings [8, 9]. Additionally, the identification of the sources of imported infections play a crucial role in malaria programme pre-elimination and elimination phases, through the assessment of patient travel history records [10], which makes possible the identification of imported cases and rates, importation routes and the delimitation of areas with high-risk travellers and sources [8, 9, 11].

Swaziland is a small, landlocked country in southern Africa, and due to elevation and climatic conditions the majority of Swaziland has historically had low levels of malaria transmission [12]. The scale-up of interventions and the expansion of surveillance in recent years, incorporating active, reactive and proactive case detection, have brought the country close to elimination [12‐15]. Between 2000 and 2014 Swaziland reported a decrease of more than 97% in malaria cases and a reduction of over 93% in malaria deaths between 2001 and 2014 [6]. However, population movements, both within and outside the country, especially to malaria-endemic areas, are a hindrance to elimination success [16, 17]. Understanding the levels of malaria importation, the sources and routes taken by infected travellers, and the characteristics of infected travellers compared to those who are uninfected is valuable for guiding elimination strategies and designing interventions to achieve and sustain malaria elimination [7, 18].

This article describes the characteristics of all individuals passively detected by health facilities and all those surveyed in follow-on reactive case detection (RACD) carried out between January 2010 and June 2014 and map the reported domestic and international travel histories. It also explores the differences in travel patterns between confirmed positive cases, with those found to be negative, and examined demographic, socio-economic and geographic factors associated with any differences.

In 2009, the Swaziland National Malaria Control Programme (NMCP) introduced active surveillance, consisting of: (1) active case investigation at the household of the index case, triggered by parasitological confirmation of a malaria case at a health facility; (2) RACD, triggered by the location of a confirmed malaria case in Swaziland’s receptive area; and, (3) pro-active case detection, triggered by strong suspicion of malaria transmission within a defined detection area and on high-risk populations, such as Mozambican farm workers [19] (Fig. 1). Every case that is parasitologically confirmed at health facilities (‘index case’), whether it is identified by rapid diagnostic test (RDT) and/or by microscopy, is investigated at the patient’s home within 7 days (January 2010 to June 2013) or 48 h (July 2013 to June 2014) of the patient’s presentation date, subject to consent by the patient or guardian. If the confirmed malaria case lives in Swaziland’s receptive area (approximately the eastern half of the country), every person residing within a radius of either 1 km (January 2010 to June 2013) or 500 m (July 2013 to June 2014) from the residence of the index case is tested for malaria. A RACD event remains open for up to 5 weeks, where the NMCP additionally conducts fever screening and where individuals near the index case report a recent fever, enabling identification of any additional malaria cases (‘secondary cases’). Any identified positive case is referred (often driven) to the nearest health facility for treatment and followed up [19]. Figure 1 outlines the surveillance structure.

During both index case investigation and RACD, the surveillance agent is required to question each participant on their age, gender, occupation, and nationality. Additionally, travel history is recorded, including places visited during the 8 weeks prior to the investigation, duration of stay, reason for travel, means of transport, border crossing sites, and access to/utilization of vector control and personal protection measures. The latitude and longitude of the residence of each individual is also recorded. Once the case investigation is undertaken, a case review is carried out to identify the origin of the infection and classify cases into the following groups: (1) imported: a case for which the origin can be traced to a known malarious area outside the country in which the case was diagnosed; (2) local: a case for which the origin is likely from local transmission; (3) intraported: a case which has its origins in another part of Swaziland; and, (4) unknown: where the information available does not enable a case classification [19].

The data used in this study are from the Swaziland Malaria Surveillance Database System (MSDS) and include data recorded between January 2010 and June 2014 for all index cases and between August 2012 and June 2014 for all individuals tested for malaria as part of RACD in pre-defined receptive areas. The data covers four full malaria seasons defined by the Swaziland NMCP: July 2010–June 2011, July 2011–June 2012, July 2012–June 2013, and July 2013–June 2014, with the addition of half a season: January 2010–June 2010. The data were anonymized and the place of residence of each individual and travel destination were georeferenced using ArcGIS v10.2.2 [20]. The geolocation of residences was aggregated by localities, which is the relevant operational unit at which public health decisions are made in Swaziland. The travel destinations outside Swaziland were identified to include city/town and country. For those locations in South Africa and Mozambique where the majority of international travel took place, the district level (administrative level 1) was also included. The international administrative units were obtained from the Global Administrative Areas Database (GADM) [21].

The surveillance data were cleaned by removing any potential duplicate or miscoded records, and any reactive case entries without a corresponding index case were excluded from the analyses. The data were then analysed using R [22], where descriptive statistics were used to create sociodemographic and travel profiles for each index case, and both the positive and negative cases identified though reactive case investigation. Logistic univariate and multivariate regressions were carried out to identify the relative risks factors associated to the travel patterns among all positive (index and secondary cases) and negative cases identified through RACD, not only in terms of sociodemographic characteristics but also protection used during travel and at home. The data from all index cases and individuals tested in RACD (positive and negative malaria confirmation) were split into three sets for further analyses in understanding travel patterns: (1) all those who travelled abroad; (2) those who travelled within Swaziland; and, (3) those who did not travel. The Plasmodium falciparum malaria prevalence map produced by the Malaria Atlas Project [5] was used to provide estimates of malaria transmission intensity at the travel destinations outside of Swaziland. Population distribution data were extracted from the WorldPop project [23] to quantify population densities at origin and destination locations. These datasets were in raster format, with a resolution of 0.0083 decimal degrees (approximately 1 km at the Equator) for 2010. Additionally, urban and rural extents were used to identify if the origin and/or travel destinations were urban and to estimate the distance to the nearest urban area, using a global urban map [24].

The total number of confirmed malaria cases passively detected at health facilities between January 2010 and June 2014 was 2123. Active case investigation was successfully carried out on 1517 (71.5%) of these malaria positive cases (index cases). January had the largest number of index cases registered at the health facilities, with 371 (24.5%) cases, and August had the lowest number (36(2.3%)). Some 60% of cases were reported between January and April, and January 2014 had the largest number of cases reported overall, with 150 cases. Figure 2 depicts the number of cases by season and month.

In RACD events carried out between August 2012 and June 2014, a total of 9859 people were tested, leading to the identification of 105 new positive malaria cases via RDTs (secondary cases). Including both index and secondary cases, overall there were 1622 positive cases and 9754 negative cases. Positive cases identified through RACD were linked to 64 index cases. Forty-one (64%) of these index cases led to the identification of one secondary case, 15 (23%) to two secondary cases and five (8%) to three cases. Of the remainder, there were three (5%) index cases that led to the identification of five, six and eight additional cases, respectively. Some 44% of positive reactively detected cases were family members of index cases living in the same property or in the vicinity, and the remaining 56% of positive cases were detected in the households of neighbours within the surveyed radius of the index case.

Comparing the demographics of all negative and confirmed positive cases from RACD events, the latter were predominantly male (67%), while the former were mostly female (55%). Overall, the 24- to 44.9-year old age group had the highest number of cases, including both positive and negative cases (33% for positive cases and 25% for negative cases). Although the occupation for a large proportion of positive cases (42%) was not determined, the most commonly reported occupation for both positive and negative cases was ‘unemployed’ (16 and 34%, respectively), followed by ‘other’ (12 and 29%, respectively) and ‘student’ (12 and 26%, respectively). Furthermore, a large proportion of positive cases reported domestic or international travel history (65%) compared to negative cases (10%), (further information can be found in Additional file 1: Table S1). A total of 15 nationalities were reported among the positive cases, with the top nationalities being Swazi nationals (72%), Mozambican (26%) and South African (0.6%), while most of the negative cases were Swazi nationals (99%) with seven other nationalities being present.

Figure 3 shows the classification of the origin of malaria infection for all positive cases (index and secondary) by season. In total, there were 866 (53%) cases reported as imported, 548 (34%) as local, 42(3%) as intraported (from different areas within the country) and 166 (10%) as unknown. The first season of this analysis only had data from January to June 2010, and it showed the lowest percentage of cases classified as imported with 25% (21/84). From this period, the number of imported cases increased with July 2013 to June 2014 having the largest percentage of 56% (320/575). In contrast, the number of cases reported as local declined from 74% (62/84) of cases in between January 2010 to June 2010 to 29% (166/575) between July 2013 and June 2014.

Countries that are transitioning towards malaria elimination are improving their surveillance systems to capture remaining cases and gather more information about them. Robust analysis of these data is key to understanding the changing epidemiology of malaria and designing appropriate strategies to accelerate towards elimination. For countries that border neighbours with higher levels of transmission and have significant rates of cross-border travel, there is a need to understand rates of importation and risk factors for travellers to inform strategies. This research presents approaches to analysing the socio-demographic and travel profiles for each passively detected case recorded in Swaziland, and both the positive and negative cases identified during RACD. The results further the understanding of factors that contribute to the likelihood of acquiring malaria in Swaziland, and provide a detailed account of the most popular travel origin and destinations, highlighting the increasing trend in case importation as a result of travel to provide evidence for designing targeted strategies.

The results presented in this analysis highlight the burden that international travel to areas of higher transmission continues to have on the importation of malaria into Swaziland, being major challenge in sustaining malaria control and elimination [25]. The high number of imported and locally transmitted cases recorded consistently during the first few months of the year is likely caused by workers from Mozambique returning to Swaziland in January following the Christmas and New Year holidays [17]. Additionally, high importation rates have been attributed to sugar plantation workers whose travel patterns are well known between these two countries. Unsurprisingly, international travellers tend to spend longer away than domestic travellers, and given the higher risks of acquiring malaria in Mozambique, where the majority of travellers go, this length of stay increases the risks of acquiring and returning with parasites [26]. Differences were seen in travel patterns between genders, age groups and occupations. The reasons behind these are many and varied, but likely include adolescent age groups travelling to and from boarding schools and employed males travelling for work at plantations [16].

Additionally, this research examined the travel patterns of those who tested negative for malaria in RACD, enabling a comparison with the patterns presented by those who tested positive. The primary risk factor for malaria infection in Swaziland was shown to be travel (Additional file 1: Table S1), more specifically international travel to Mozambique by 25- to 44-year old males, who spent on average 28 nights away. Males that were malaria positive travelled to a total of 12 different countries, double the number visited by positive females; 67% of the reported international travel by positive males were carried out between December and March, which was higher than the 60% of international travel reported by females who tested positive during the same period; 96% of RDT-positive international travellers were either Swazi (52%) or Mozambican (45%) nationals, however Swazis were more likely to test negative.

Focused screening and treatment at borders has shown that malaria cases can be identified before reaching receptive areas [27], and this could be a viable option at those crossing points with the largest number of malaria positive cases such as those at the border with Mozambique at Mhlumeni and Lomahasha, targeting specific demographic groups such as males aged 25–44 that have travelled to areas of high malaria transmission and at specific times of year when the largest influx of people occurs. It is however important that robust public-acceptability and cost-effectiveness studies on such interventions be conducted first, as considering the use of RDTs for this purpose for example, may not be as efficient at identifying cases when the levels of parasites are low. Focussed screening could also be strengthened by network sampling, where tested individuals can help recruit others with similar travel characteristics that can potentially be at high risk of importing parasites [13, 28].

The current surveillance system of Swaziland’s NMCP, that incorporates RACD, shows the ability to detect cases that would otherwise be missed by passive surveillance [29]; however, additional efforts could be targeted to those high-risk groups, such as international travellers to high transmission areas, through education of the risks and provision of chemoprophylaxis, ensuring that that they are protected prior to travel, and adhering to the drug regimen properly. The availability of individual-level travel history data here allows for the identification of areas that have an elevated risk of malaria importation. Combining this information with high resolution predicted malaria incidence and health access maps [30] can provide guidance on targeting effective intervention packages by space, time and within populations.

There are limitations to this study, in particular the classifications of imported versus local cases, where a travel history prior to case confirmation does not necessarily mean that the case was imported. Additionally, assigning a likely location of case acquisition is uncertain when a patient travelled to multiple locations. The surveillance data and the changes within it during the time period covered also resulted in some uncertainties. The RACD was carried out in the latter half of the study period and was mainly focused on receptive areas, therefore there is a lack of information on negative cases and any potential secondary cases from other areas of the country. The length of time since recorded travel changed through time, as follows: January 2010–June 2010: travel within last 2 weeks; July 2010–June 2013: travel within last 4 weeks; and July 2013–June 2014: travel within last 8 weeks. Therefore, an increase in number of trips may have inflated the trend towards the latter part of the study. These changes occurred throughout the study in order to improve the quality of data. The use of RDTs may not have been the most sensitive method to identify malaria cases, as recent studies determined that RDTs have shown low sensitive owing to an unexpectedly high proportion of low-density infection among symptomatic subjects [31], but RDTs have some advantages over other methods, not only because they give a relatively rapid results, but because they require minimal training to use, there is little or no manipulation of the sample required and most RDTs do not require refrigeration, making it easier to use in areas where there is no power supply [32]. Interpretation of results should also consider classifications such as age classes, employment and reason of travel, as these were grouped here for comparative purposes. Furthermore, the index case classification was a subjective classification made by the NMCP’s Chief Surveillance Officer based on the likelihood of the origin of the case with reference to travel history. Throughout the study the number of unknown cases increased and these were mostly (72%) people that did not travel. The logistic regressions carried out in this analysis used specific reference groups to compare the likelihood of acquiring malaria based on known groups that were the least likely to test positive or had the majority of negative cases, so it is possible that different results could have been obtained if these reference groups varied, additionally, some variables were too small to develop meaningful multivariate models to describe risk factors. Future work will focus on a number of directions. The analyses will be updated as new surveillance data continue to be collected to provide contemporary evidence based upon which to guide programmatic decisions. The data will also provide a valuable input to modelling efforts to understand the effects of individual-level characteristics and spatial importation risk, e.g., where imported cases most often appear and their likelihood of prompting onward transmission to refine existing estimates and risk maps [18, 33]. The data could also prove useful in parameterizing models of human and parasite mobility that can be extended beyond Swaziland, as well as compared to other forms of mobility data, such as from mobile phones [7], census microdata [33, 34] or parasite genotyping [35] and, to strengthen assessments of the impact of climatic variations and incidence on malaria transmission [36] or genetic studies that further the understanding of transmission dynamic in low transmission settings [18].

The information provided in this analysis is useful to direct national and regional targeting of interventions, over space, time and by population groups. Results presented here highlight how Swaziland can continue to benefit from cross-border collaboration to manage importation of malaria at the regional level. Previous regional schemes, such as the Lubombo spatial development initiative (LSDI), made significant contributions to the reduction of malaria prevalence, with the introduction of effective treatment and vector control in the region [13]. The establishment of the elimination eight (E8) initiative, which aims to eliminate malaria from eight southern African countries, including Swaziland, South Africa and Mozambique by 2030 [2], will continue such collaborations as well as the revival of the LSDI as the Mozambique, South Africa and Swaziland (MOSASWA) Malaria cross-border initiative.