While travel to a malaria-endemic area in Ethiopia is recognized as a risk factor for malaria infection, much of the country is endemic for malaria transmission and little attention has been paid to whether or not routine human movement patterns can lead to higher risks of malaria infection, even on a small geographic scale within endemic areas. This study shows that travel may be a risk factor for symptomatic P. falciparum infection even when the travel occurs within endemic areas. Such endemic areas, including the area where this study was conducted, may vary in risk of infection even within relatively small geographic zones, for reasons which may related to environmental factors as well as a myriad of other influences [17, 18]. Thus the relationship of travel to malaria risk is likely due to a combination of factors including movement into areas of higher transmission during risk periods for mos-quito biting, such as movement for harvesting of crops which correlates with the transmission season in Ethiopia, relaxed use of preventative measures when individuals are away from home, and possible behavioural differences by sex or during travel. Ethiopia is home to a large pastoralist population which frequently moves in search of pasturage for animals, and further other short-term travel to visit nearby market areas, friends and relatives or for school attendance is also common. No distinction between travel to areas with differing malaria risk was made in this analysis, a fact which could lead to bias in the estimate of travel history as a risk factor compared with assessing travel history to a higher malaria risk area. Such a bias would be expected to be in the direction of the null hypothesis because including areas with low or no malaria risk in the analysis should reduce the overall association of malaria infection and travel history.
In this study blood slides and RDTs were performed only once, as such misclassification bias could arise due to the potential for inaccurate microscopy readings by facility technicians. However, the facility was part of an on-going quality assurance and control programme (QA/QC) at the time of the study. The contemporaneous QA/QC results showed 96% agreement of facility readings with expert readings over the five month period surrounding the study with regards to positivity and negativity, and 90% agreement on species identification. This suggests that any misclassification bias associated with microscopy conducted at the facility level is likely to be small, furthermore, such a bias would likely be in the direction of the null hypothesis.
Within low transmission areas, importation of infections could lead to maintenance of reservoirs of infection in the absence of local vectorial capacity to maintain transmission [25, 26]. Understanding where and how infections originate in the Ethiopian context will be important for assessing prospects for local transmission interruption. The data in this study show that it may be important to consider importation as a potential source of P. falciparum infections, even within moderately endemic areas of the country (Bulbulla health centre’s estimated prevalence (PfPR2-10 based on the Malaria Atlas Project is approximately 7%) .
As P. vivax possesses a dormant hypnozoite stage and causes relapses of malaria, many of the P. vivax infections in this study may have originated not from initial sporozoite inoculations, but rather as relapses. The epidemiology of the disease appears different among P. vivax infections in this study, travel, at least recent travel, was not a marked risk factor, while age and sex were important. In this study male sex was the greatest risk factor for all categories of malaria infection. It is possible that behavioural factors associated with being male (e.g, staying out late in the evening) could be predisposing towards malaria infection. While men in this study were more likely to travel than women, and spend the night away from their home village, even after adjusting for this travel, sex remained a significant risk factor for all types of malaria infection.
In this study, ownership or use of an ITN showed no statistically significant protective effect against P. falciparum, and was associated with increased infection with P. vivax. While this finding is out of line with the literature and large-scale evaluation research, lack of protection from net use has been reported in observational studies in Ethiopia before [28–31]. In this context, these results could derive from a combination of factors including the prevalence of old and ineffective nets, the epidemiology of P. vivax infection, endogeneity between the use of nets and malaria risk or simply the limited sample size.
In order to assess effect-modification through the relaxed use of protective measures during travel the interaction between net use and travel was tested. Though the interaction term was statistically insignificant in this context, this was possibly hampered by small sample size. Regardless, the low level of use of LLIN use indicates the need to focus on developing a strong distribution system to maintain LLIN coverage over time, and to increase use among those travelling and staying away from their home overnight.
While this study indicates travel is a risk factor for P. falciparum infection, the study is a small case–control study of limited geographic scope. Because this study focused only on patients presenting at one health centre, the study has limited external validity. Data collection from national parasitemia surveys or at a minimum from multiple health centres would help to make these results more generalizable. Further, the size of the adjusted odds ratios found indicate that travel is not a strong risk factor. However, given that the study area has higher transmission than some other areas of Ethiopia, notably the Rift Valley basin near Eritrea or highland areas, travel may be an even more important factor in other areas of the country . Travel history is also a potentially difficult variable to measure retrospectively; patients may not remember the dates or duration of their travel accurately over long recall periods. For this reason questions on travel were limited to the 30 days before the administration of the questionnaire. However, recall bias may still be present. Further, this study did not attempted to differentiate between travel in the previous week and travel occurring more than a week previously, which might help to focus results on the period during which an infection would have to have occurred in order to produce current symptoms.
The epidemiology of P. vivax is generally poorly explored  and given the current lack of a sound method for distinguishing relapses from current infections this study cannot distinguish between travel as a risk factor for new infection, likely diluting the potential effects. Nevertheless, P. vivax transmission is likely to be more resilient to interventions that reduce transmission pressure for several reasons including greater and more rapid production of gametocytes, shorter extrinsic incubation periods and the dormant liver stage . As such travel might be expected to be less of a factor in the acquisition of new P. vivax infections in this setting but could easily contribute to the maintenance of transmission even in very low transmission areas.
Derivation of travel history from individual recall is likely to be subject to recall bias. Recent technological developments including inexpensive small Global Positioning System data loggers and mobile phones may give rise to the ability to correct for these biases to some extent, by allowing for relatively fine scale temporal and spatial tracking of human or at least device movement [33, 34]. The results of this study indicate that increased risk of infection may arise not only from the movement of individuals themselves but also from changes in behaviour coupled with travel. Such behaviour changes cannot currently be tracked using mobile phone databases or GPS logging systems and as such individual survey methods may still be required to fully understand the epidemiology of malaria infection and its interaction with travel and human movement patterns.
These findings suggest that careful consideration should be given to travellers and perhaps especially adult men in areas of moderate transmission in Ethiopia. These individuals appear to be at the highest risk of malaria infection. While this study cannot determine if the risk evolves from travel to areas of higher risk or relaxation of behavioural preventative measures during travel, it indicates the potential for interventions on communicating the importance of maintaining behavioural prevention activities, such as use of bed nets during travel. It additionally suggests that such communication measures might be most advantageous if targeted towards men.