To summarize, in this border region with persistent malaria transmission participants consenting to carry a GPS data logger were highly mobile year-round within the study area, throughout Zimbabwe, and into neighboring countries. Although overall mobility was high, movement patterns were highly heterogeneous among individuals. A high proportion of participant time was spent near the home on average, and half of participants remained within 5 km of their household at all times; however, 30% of participants traveled out of Zimbabwe, 23% reported sleeping away from home, and nearly 10% traveled more than 200 km away. Movement indices were slightly higher during the dry season and were significantly higher among male participants and adults. Adolescents, in particular, tended to stay within the study area and remained near the home a higher proportion of the time. Thirty percent of participants travelled into Mozambique at least once, and 8% were recorded in Mozambique overnight with trips lasting up to 35 nights. Overall, these results agree with previous investigations of movement patterns in this region (42).
While 1-month malaria incidence in this population was low during the period of observation at only 1.9%, participant movement patterns have clear implications for malaria control. Over 40% of participants spent time within 1 km of the Mozambique border, which has been shown to have a higher risk of malaria transmission (28, 29). During peak mosquito biting hours, 43% of participants spent at least 2 hours away from home on average, during which time they may be more likely to be exposed to biting vectors. Furthermore, Manicaland Province, where Mutasa District is located, consistently has the highest number of malaria cases per year in Zimbabwe (43), and the high rate of participant travel throughout the country could further contribute to introductions in regions of lower relative transmission.
Perhaps most importantly, the high proportion of participants who traveled across the border into Mozambique may impact malaria control efforts in Zimbabwe. At the time of the study, Mozambique reported lower rates of key malaria control interventions and higher indices of transmission than Zimbabwe (42). In particular, the Manica Province of Mozambique directly across the border from the study area has among the highest prevalence of malaria in the country, particularly in rural and border areas (44, 45). The frequency of cross-border travel, including overnight stays, thereby presents a clear risk of malaria importation from travelers into Mutasa District. This risk is further compounded by reported high rates of incoming cross-border movement from residents of Mozambique into Mutasa District to visit family, for healthcare seeking, for commerce, and other reasons (42). Identifying individuals or groups at higher risk for malaria importation can aid in designing surveillance systems and malaria control interventions for these populations (6).
This investigation also further highlights the need for collaboration among neighboring countries for malaria control. Zimbabwe is one of the countries participating in the Elimination 8 Regional Initiative, aimed to accelerate malaria elimination in key countries in Southern Africa, and promotion of regional coordination among member countries and reduction of cross-border malaria are two of the key objectives (46). Existing cross-border initiatives to reduce malaria transmission could serve as models for malaria control in this region (9–11, 47), and activities such as integrated surveillance systems, screening travelers at border crossings, and targeting economic migrants for interventions may reduce transmission across countries (5, 48, 49). However, these interventions require further knowledge about formal and informal migration patterns, and efforts to understand the extent of formal and informal cross-border movement in particular settings are key to planning effective interventions in border populations.
To this end, the high heterogeneity of movement patterns among residents of Mutasa District in this study underscores the continuing need for multiple sources of individual-level data to fully capture the range of mobility behavior across different settings, accounting for local context and demographic factors. Of note, the population observed in this study traveled higher total distances, distances per day, and maximum distances from home compared to populations observed in Nchelenge District and Choma District in Zambia in similar studies (Fig. 2) (25, 26). These individual and regional differences in small- and large-scale movement can be expected to influence malaria transmission risk (50–53). Therefore, the suite of interventions needed for cross-border malaria control must vary according to local population movement patterns in addition to epidemiologic and environmental factors. This is particularly important as national malaria control programs move toward targeted interventions in areas of higher transmission (54–56).
This investigation was subject to several limitations. Participants were selected as a convenience sample within a larger surveillance study and are therefore not representative of the underlying population of Mutasa District. The exclusion of children from the sample in particular limits the ability to make inferences about the population most at risk of mortality and morbidity due to malaria. Furthermore, there were potential inaccuracies caused by limitations of the GPS logger devices. Although a 5-minute interval provides a high degree of precision compared to most population movement studies, all movement between recorded points was not captured, and therefore the total distance traveled may be underestimated. Due to limits in the precision of individual points, participants could not be classified as indoors or immediately outside their home, so calculated time away from home may underestimate time spent outdoors. Compliance with study protocol was also difficult to assess with these devices since no biometric data was collected. Participants could have forgotten the logger at home, intentionally not worn it, or allowed it to be worn by another person. Since these scenarios were difficult to verify, logger data was analyzed as collected unless a specific issue was reported to the field team. A further potential discrepancy was the presence of two persistent positive RDT results at logger collection, which could indicate either new incident infection, persistent infection, or persistent HRP-2 antigen positivity following treatment. Several studies have reported false positive RDTs after a prior malaria infection due to HRP-2 antigen persistence, and therefore a one-month interval may not be long enough to accurately detect incident infections with RDTs (57–60).
Despite these limitations, this study was able to collect a large quantity of high-quality movement data and captured significant migration patterns among participants in a remote rural area. The investigation built off previous population movement studies conducted in Peru and Zambia and was the first to capture significant cross-border movement, which has clear implications for malaria control in this population. Furthermore, the study had high acceptability in this population and data quality was high overall; 95% of loggers distributed yielded usable movement tracks, corresponding to 96% of participants who provided at least one usable data file. These results indicate that this technology continues to be a feasible and useful method for detailed population movement studies, including in remote areas.