Travel Is a Key Risk Factor for Malaria Transmission in Pre-Elimination Settings in Sub-Saharan Africa: A Review of the Literature and Meta-Analysis

Abstract. By sustaining transmission or causing malaria outbreaks, imported malaria undermines malaria elimination efforts. Few studies have examined the impact of travel on malaria epidemiology. We conducted a literature review and meta-analysis of studies investigating travel as a risk factor for malaria infection in sub-Saharan Africa using PubMed. We identified 22 studies and calculated a random-effects meta-analysis pooled odds ratio (OR) of 3.77 (95% CI: 2.49–5.70), indicating that travel is a significant risk factor for malaria infection. Odds ratios were particularly high in urban locations when travel was to rural areas, to more endemic/high transmission areas, and in young children. Although there was substantial heterogeneity in the magnitude of association across the studies, the pooled estimate and directional consistency support travel as an important risk factor for malaria infection.


INTRODUCTION
It is widely recognized that human mobility influences the spread of infectious diseases, 1 and evidence suggests that human movement was partly responsible for the failure in the previous global malaria eradication campaign. 2 Over the last century, following widespread malaria control/elimination strategies, transmission risk varies markedly between and within countries; thus, travel across different transmission settings markedly influences malaria vulnerability, especially in areas with high malaria receptivity. 3 In the past, malaria and travel largely focused on international travelers to countries or areas at risk of transmission arriving from countries of no risk, who were more susceptible to severe malaria because they lacked immunity. 4 Tremendous progress in the reduction in malaria morbidity and mortality has been achieved in the last decade, with 21 countries projected to attain elimination by the year 2020. 5,6 Therefore, imported malaria has become an important risk factor in formerly malaria-endemic countries that had attained elimination and in low transmission countries that are targeted for elimination. These settings continue to experience thousands of malaria cases every year through importation, resulting in increased morbidity and mortality, a substantial burden to the health system, and occasionally secondary transmission. 7 Findings from analyses of international population movements using census-based migration and reported malaria data suggested that certain groups of countries were much more strongly connected by relatively high levels of population and infection movement than others. 2,7 Therefore, in malaria post-elimination and preelimination settings, imported malaria remains an important threat to the gains that have been achieved. 8,9 Notwithstanding, the literature on travel in the context of malaria transmission is limited, and few studies have investigated this relationship, particularly in sub-Saharan Africa. To examine the role of travel as a potential risk factor for malaria infection and imported malaria at large, we conducted a literature review and meta-analysis of studies investigating the association between history of travel and malaria infection status in sub-Saharan Africa.

METHODS
Literature review. The selection of studies was conducted using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. A literature search was conducted using PubMed in September 2019 to identify studies that investigated the relationship between travel and malaria using combinations of the Medical Subject Heading (MeSH) terms "travel," "risk factor," and "malaria." Articles were screened first by title, then by abstract, and then by full text for inclusion in the review. References of key articles were also reviewed to identify additional relevant studies. Inclusion criteria included 1) assessment of the relationship between travel and malaria infection in an endemic setting using odds ratio (OR) as the measure of association; 2) history of travel, defined as an exposure with any geographic or temporal variation used (i.e., travel outside the village/country to an area of higher transmission and travel in the past 1 month/2 months); and 3) study location in sub-Saharan Africa. Exclusion criteria included 1) no assessment of a relationship with travel (i.e., studies looking at clinical management, treatment, interventions, giving travel advice, and using a measure of association other than OR); 2) history of travel not defined as a specific exposure or risk factor (i.e., studies looking only at travel/ migration patterns and importation burdens); and 3) study location in endemic country outside of sub-Saharan Africa. Two authors (S. A. and S. K. K.) independently screened the titles and abstracts of the articles using the inclusion/exclusion criteria. Any discrepancies were discussed with the third author (J. M. N.), and decision on inclusion was made jointly.
Statistical analysis. Measures of association (ORs with 95% CIs) and key study or participant characteristics (including geographic location, study design, sample population, sampling/recruitment methods, sample size, malaria case definition, definition of travel history, and confounders adjusted for) were extracted for the meta-analysis. For a few studies that only reported separate OR estimates for predefined subgroups (e.g., by gender), a single estimate was first calculated for the study by inverse-variance weighted fixed-effect meta-analysis and further used in the overall meta-analysis. A random-effects meta-analysis was conducted using the most adjusted OR from each study using Stata version 13 (Stata Corporation, College Station, TX). Between-study heterogeneity was assessed using the I 2 statistic, 10 and association with study-level characteristics was assessed using meta-regression. Sensitivity analysis was undertaken by stratifying the meta-analysis by study characteristics, including geographical region type of malaria tests, and level of adjustment of confounders.

RESULTS
Studies included in meta-analysis. The literature search identified 645 publications. Following title, abstract, and fulltext review, only 22 studies from the PubMed search and one study published as a conference abstract met the inclusion and exclusion criteria, and thus were retained for the metaanalysis ( Figure 1, Table 1).
A summary of the study locations, design, sample selection, definition of travel, reported ORs, and confounding factors adjusted is shown in Table 1. The 22 studies included in this review were carried out in Botswana, Burkina Faso (2), Ethiopia (6), Ghana (2), Ivory Coast, Kenya, Malawi (2), Mozambique, Namibia, Uganda, Swaziland, and Tanzania (3). They primarily reported results from study sites with low transmission intensity, and all used logistic regression modeling to test the association of history of travel (and other covariates) with malaria infection status. All studies used either a case-control or cross-sectional design, but some differed in their case ascertainment methods and geographical/ temporal criteria for defining the travel variable. Most of the studies recruited patients diagnosed with malaria at particular health facilities for inclusion; seven studies surveyed randomly selected households across the study area; one study in Botswana 11 looked at all inhabitants of the study site, and one study in Ethiopia 19 looked at all current residents of the seven villages selected. Two studies used data collected from active surveillance systems (Tanzania 31 and Swaziland 29 ). Most studies defined history of travel as having travelled overnight outside of the home village or area in the past month. A handful of studies restricted participants to a particular age criteria: both of the Malawi and Ghana studies looked only at children aged < 5 years 20,21,24,25 ; one of the Burkina Faso studies looked only at children aged < 12 years, 13 and one of the Ethiopia studies looked only at adults aged 18 years and older. 14 Some of the studies further specified if travel was to areas that were rural, 12,20-22,24,30 malaria endemic, 16,18,19 or had an equal/lower or higher transmission intensity, 28 but most of the studies did not.
Meta-analysis results. A forest plot of the meta-analysis, with studies grouped by geographical location, is presented in Figure 2. Overall, all but three of the studies 12,13,27 found that history of travel was significantly associated with malaria infection. The pooled OR was 3.77 (95% CI: 2.49-5.70)indicating the odds of having malaria infection were almost 4-fold higher among individuals with a travel history than to those with none. There was substantial heterogeneity across studies (I 2 = 97% [95% CI: 97%, 98%]) that was only partly explained by study-level characteristics such as geographic location, method of malaria testing, extent of adjustment for confounders, and duration of travel (meta-regression residual I 2 = 70% [44%, 84%]). There was no evidence of small study effects, including potential publication bias, as visually judged from symmetry of funnel plot ( Figure 3) and confirmed with formal statistical test (P = 0.833 for Egger's test of small-study effects).
The ORs from the individual studies ranged from 0.7 in Rukunguri, Uganda (for people of the highland fringe site travelling to areas of higher transmission intensity), 28 29 Travel to an area of higher malaria risk from the highland site in Uganda also had a greater OR (AOR = 6.9, 95% CI: 1.4-33.1, P-value = 0.02) as opposed to an area of lower or equal risk (AOR = 1.5, 95% CI: 0.3-7.9, P-value = 0.7). 28 Studies in Mozambique and Malawi found travel history outside city limits and malaria infection to have a strong association, specifically for urban populations (OR = 3.93, 95% CI: 1.56-9.89 and AOR = 2.36, 95% CI: 1.31-4.2, respectively). 25,26 The two Burkina Faso studies calculated slightly increased odds for those travelling contracting malaria, but these odds were not statistically significant. 12,13 In Ghana, a significant association between travel and malaria was only observed for people in the city of Kumasi but not in Accra. 20

DISCUSSION
The results of the meta-analysis strongly suggest that travel is a key risk factor for acquiring malaria infection, particularly in FIGURE 3. Funnel plot assessing potential for small-study effects. urban areas and when travel is to rural/malaria-endemic areas or areas of higher transmission intensity than the origin location. The Swaziland and Zanzibar surveillance studies hold the second greatest weight in our meta-analysis (based on the number of participants and cases) and found people with a history of travel to be almost 14 and 10 times more likely to have malaria infection than those with no history of travel, respectively. 29,31 As described earlier, only three studies in the current analysis specified that the destination of travel was to an endemic area and only one looked at the transmission intensity at the destination. Six studies in the current analysis looked specifically at travel to rural areas as the cause of malaria importation into urban and peri-urban areas, of which all but one found a significant association. [20][21][22]24,30 This is in agreement with a previous analysis that suggested that travel to rural areas is a major source of malaria transmission in urban areas. 33 Some of the studies attempted to elucidate further details regarding the population of travelers, such as major reasons for travel, age, gender, time, and seasonality of travel. In the context of Ethiopia, for example, it is believed that most of the imported infections are due to routine human travel to areas of higher transmission intensity, with nonimmune people from high-altitude areas 15 travelling to low-altitude areas during periods of high mosquito-biting (e.g., movement for harvesting crops). 14 Similarly, the study in Kenya found the increased risk of malaria infection with travel to be concentrated in children aged < 5 years. 33 Identifying characteristics of groups at high risk for travel and possibly importing infection should be a key area of focus for future studies to elucidate further knowledge gaps and enable the design of appropriate strategies for control. 2 The relationship between malaria and travel is complex; nonetheless, imported malaria has potential for establishing transmission in the most receptive and vulnerable locations 3 in addition to burdening the health system with imported malaria cases. 7 It is likely that residents from non-malariaendemic settings who travel to malaria-endemic settings have less access to information and interventions on malaria prevention, which further makes them more susceptible to infection during travel. On the other hand, although infected visitors from malaria-endemic areas to non-endemic areas may be better informed about prevention, it is likely that they have less access to malaria prevention interventions during travel. Therefore, given the optimal receptivity and vulnerability conditions, imported malaria has potential to reestablish and sustain transmission in low transmission settings. Moreover, evidence from Zanzibar suggests that imported malaria among returning residents was more important than infected visitors from Tanzania mainland. 34 Therefore, given these complexities, tracking the at-risk populations and designing combinations of approaches to address malaria importation are needed to enable targeting residents travelling to malariaendemic areas, residents returning from malaria-endemic areas, and visitors from malaria-endemic areas with appropriate interventions. 8 This review has several potential limitations. First, travel history in all studies was based on self-reports, and there is, to our knowledge, no literature on establishing the validity and accuracy of this exposure in studied populations. In addition, self-reported travel history was likely to be influenced by recall bias, especially in studies that looked at travel history dating back to several months. Second, most of the studies were case-control studies, and they are inherently susceptible to selection bias. Third, our review only included studies undertaken in sub-Saharan Africa and where there was a priori expectation that travel would put a person at risk rather than one in which travel might reduce risk. Finally, although the analysis does not suggest any evidence of publication bias, it showed substantial heterogeneity in the magnitude of association across studies.
As more and more malaria-endemic countries aim to attain elimination of malaria, imported malaria has remained an important risk factor for reestablished transmission. The strength of the association implied by the random-effects metaanalysis pooled estimate and directional consistency of the study-specific findings support travel as an important risk factor for infection and, possibly, sustaining malaria transmission in malaria elimination settings. Therefore, national malaria control or elimination programs need to identify at-risk populations and travel patterns that drive malaria transmission and implement appropriate interventions to limit malaria importation and minimize its potential impact on autochthonous transmission.