Significant differences in the migration patterns of migrant workers were found from Cambodia and Myanmar on the Thai-Cambodia border. Migrants from Myanmar had a longer duration of residence in Thailand, rarely if ever returned to Myanmar, and were more likely to consider Thailand home than Cambodians. The vast majority had came from Mon state, had used the services of a broker to enter Thailand, and were more likely to have lived in other locations in Thailand before arriving in their current location. The majority of migrants from Cambodia are short-term migrants, and more than half have been in Cambodia in the last three months, usually for holidays or to visit friends and family. While less than a third of Cambodians who have lived in Thailand for more than six months had plans to return to Cambodia, more than two-thirds of short-term migrants had plans to return, and almost three quarters had been in Cambodia in the past three months. The provinces listed as province of origin and of next destination covered most of Cambodia, though the majority were western and central provinces.
RDS has historically been used to study other hidden or difficult-to-reach populations, including men who had sex with men, commercial sex workers, and injection drug users, often in the context of HIV prevention, and has been shown to be a flexible and robust method that can produce a sample representative of the heterogeneity of hidden populations [16
]. This methodology was successful in recruiting a sufficient sample of both Cambodian and Myanmar migrants living and working on the Thai-Cambodia border, despite initial concerns that social networks might not be large enough, and that restrictions in travel would hinder the ability to recruit new participants. While the overall design of this study was very similar to RDS studies done with other populations, several additional factors in recruiting this population had to be taken into account. First, while most RDS studies have usually been done in urban areas, this population is largely rural, and issues around funding for transportation had to be considered. Location of the study sites was extremely important to assure access. Due to language barriers in these migrant populations, translators were necessary. Because the agricultural work that many migrants come to do is highly seasonal, it was critical to time the study with the peak in the need for agricultural work.
In the Cambodian community, there were strong networks between recent and more long-term migrants, resulting in the recruitment of more short-term (M2) migrants than long-term (M1) despite starting with more M1 seeds. In the Myanmar community, very few short-term migrants were recruited, which was confirmed by discussions with key informants stating that there were in fact few short-term migrants from Myanmar, perhaps reflecting the amount of time required for migrants to travel from the western to the eastern side of Thailand. In addition, the homophily analysis showed that M2 migrants from Myanmar were recruited exclusively by and exclusively recruited M1 migrants, showing that they are well-integrated into the long-term migrant community and confirming that they are likely few in number. While RDS methodology is not immune from sampling bias [20
], and discrepancies between network compositions or the recruitment behaviours (including social cohesion in recruiting from only one's own types) may impact the nature of the sample, this does not appear to have been responsible for the preponderance of long-term migrants from Myanmar.
While respondent driven sampling proved to be an effective sampling methodology to study mobile migrant populations, there were some challenges and limitations in the implementation. While not inherent to RDS, staff struggled with the length of the questionnaire, and this as well as the need for interpreters may have limited data quality. All answers were self-reported and not based on observation, possibly resulting in some recall or reporting biases. Blood samples were not collected. They would have given more information not only as to the movement of the human population, but of parasites as well.
The migration patterns demonstrated here have implications for containment of artemisinin resistance. There was a concern that migrants from Myanmar may carry the resistant parasite back to their country of origin, which is highly endemic, has not achieved the levels of malaria control that other SE Asian countries have, and has limited access by international agencies due to the political situation. However, those residing along the Thai-Cambodia border, primarily in Trat, have in fact largely settled in Thailand, and do not return often, if at all. The Cambodian population in the three border provinces of Trat, Chantaburi, and Sa Kaeo differ in three important ways. First, there is considerably more frequent cross-border mobility, with even long-term migrants returning on a regular basis. Second, while migrants come primarily from western and central Cambodia, there has been migration from and back to almost all provinces in Cambodia, some with very low transmission levels and largely non-immune populations. Interestingly, no migrants claimed an origin in Pailin Province, which has typically been thought to be an epicenter of development of anti-malarial resistance. Finally, while most migrants had been in Cambodia in the last three months, very few crossed the border more frequently than monthly. Whether this reflects the true reality of the situation or is due to sampling bias if the networks of those who cross more frequently did not intersect with those of the labourers in this sample, it seems as if the population resides in Thailand long enough to benefit from a treatment follow-up programme. Frequent population movements, both across the Thailand-Cambodia border and from the border area across Cambodia, indicate the need for heightened surveillance for artemisinin tolerance outside what has been designated as the containment zone, as well as close cooperation amongst Thai and Cambodian authorities.
Migration contributes to re-emergence of malaria in previously malaria-free areas [22
], and cases of malaria in migrants are often reported among people who recently returned to their countries of origin to visit friends and family [23
]. In Thailand, 46% of all malaria cases reported were in short term migrants in 2003; this increased to 55% in 2006 [2
]. As national reports do not take into account reports from non-state organizations dealing with migrants, these figures are likely to be an underestimate. As these individuals may then return to homes across Cambodia, they may carry resistant parasites with them. Effective containment depends on identifying and promptly treating all cases of malaria. However, unregistered migrants are in a vulnerable position as they may be subject to arrest and deportation, and do not have health insurance [7
], discouraging health care-seeking in Thailand. Furthermore, even if they seek treatment from Thai malaria clinics that do not require insurance, they often do not receive the same follow-up to assure success of treatment, under the assumption that they are too mobile to follow through the 28 days follow-up period. While data were not collected on how frequently migrants changed employers (and potentially residences), these data suggest that only a small minority return to Cambodia at least monthly, and that the majority likely remain long enough to complete a 28-day follow-up period. Case follow-up is crucial for artemisinin resistance containment; further research should be done to elucidate the future movement plans of short-term migrants diagnosed in malaria clinics and determine the best way to assure follow-up.