A strong association between HIV infection and migration is well established. Across sub-Saharan Africa (SSA), China, and South East Asia, migrants have higher HIV prevalence or greater HIV risk behaviors than nonmigrants [1–10]. Given this higher HIV prevalence, migrants are typically considered a high HIV risk group , and their mobility is often cited as one of the most important factors in the spread of HIV in SSA and elsewhere [11–13]. This may indeed have been the case during the early days of the epidemic when mobile men, such as long-distance truck drivers, seem to be spreading HIV from low prevalence areas to high prevalence areas . Once HIV became established throughout the general population, however, the direction of causality may have changed. Thus, it is important to examine more closely the pathways of transmission. This task, in turn, has higher data demands than did the earlier work, because migration is becoming more ubiquitous, and HIV more diffusely spread in the general population [15–18].
Despite important shifts in the patterns of migration and the epidemiology of HIV, assumptions regarding the underlying mechanisms connecting migration and HIV have persisted. Many continue to argue that risky behavior increases after individuals are released from the relatively more rigid social norms in rural premigration settings [19–21]. Other persistent hypotheses stem from enduring assumptions regarding urban settings, a common migration destination; as HIV prevalence is typically higher in urban than in rural areas, migrants would be at greater risk of infection even if their behaviors remained the same [16,22]. A third persistent hypothesis is that migration may also facilitate the acquisition of new partners by the migrants themselves and by the partners they left behind, due to time spent apart [23,24].
The assumption that migration causes HIV infection implies that HIV infection occurs after migration [1,4], but this assumption has seldom been tested because the required longitudinal data with pre and postmigration HIV status have seldom been available. Cross-sectional data are typically used to identify whether individuals who previously migrated are more likely to be HIV positive than those without a history of migration or mobility. These data, however, cannot tell us which event (migration or infection) preceded the other.
An examination of these important questions of timing has also been hampered by limited measures of migration and mobility . Migration differs by destination and origin (e.g., urban, rural), distance traveled, permanence, duration, and motivation, among other characteristics. Yet researchers have often had to rely upon simple binary measures of migrant/nonmigrant status [4,10,19,25,26]. When destination is measured for migrants, the focus of migration research has typically been on rural-to-urban migration [3,12,27] even though in SSA rural–rural migration (often related to marriage or work) is more common . This is important, as a major change in the epidemiology of HIV has been its spread to rural areas in SSA. Although rural-to-urban migration may have been critical in early phases of the epidemic to sustain its spread, this is no longer the case , and rural-to-rural migration has become a critical pathway though which HIV spreads throughout rural populations. Given the limited knowledge about these types of migration flows, as well as the limitations of common measures of migration types used in research on this topic, it is not surprising that leading scholars have called for examining migration patterns in greater detail .
To address several of these major shortcomings in the existing research that examines the links between migration and HIV infection, we use longitudinal panel data from Malawi to explore the causal pathway between HIV infection and migration. We also employ more detailed measures of migration than has typically been the case in earlier work. Contrary to the predominant assumption in the literature, we examine whether HIV-positive individuals are more likely to migrate than those who are HIV negative. We then examine the relationship between HIV infection and migration using more detailed measures of migration; disaggregated by destination (rural–rural, rural–town, and rural–urban) and permanence of stay (return/permanent migration).
Study design and data
Our longitudinal panel data enable us to overcome important limitations of previous efforts. We use data from two related sources: the Malawi Longitudinal Study of Families and Health (MLSFH), and the Migration and Health in Malawi (MHM) projects [29,30]. The MLSFH was designed as a couples’ survey, targeting a population-based representative sample of approximately 1500 ever-married women aged 15–49 and 1000 of their husbands in three rural sites of Malawi, starting in 1998. The first follow-up, in 2001, sought to re-interview all respondents from the first wave, along with any new spouses. The MLSFH returned to re-interview all previous respondents and new spouses in 2004, 2006, 2008, and 2010. Over the course of the MLSFH, two additional samples were added: in 2004, a new sample of approximately 1500 young adults (15–27) was added; the 2008 wave of the MLSFH interviewed approximately 800 parents of respondents. Comparisons of background characteristics between the MLSFH data and the rural sample of the Malawi DHS found relatively few substantive differences [31,32]. Other MLSFH sample characteristics and analyses of data quality have been previously published [29–33].
MLSFH initiated home-based HIV testing for all respondents in 2004, with follow-up testing in 2006 and 2008 . Respondents were given the opportunity to find out their test results, and the vast majority elected to do so: 68, 98, and 93% in waves 3–5, respectively, learned their HIV test results.
In all waves of MLSFH, the most common reason for noninterview was migration out of the sample area. The MHM project was designed to trace and re-interview MLSFH migrants . From its initiation, the MHM attempted to trace respondents who had migrated out of the sample after the first survey wave in 1998. To do this, the study team returned to the migrant's previous MLSFH village of residence, where they asked family members or friends where the migrant was currently living. For the majority of migrants, the team collected detailed information on the migrant's current location. The team then traced the migrant to his or her new residence.
The data collected by the MHM permit us to identify migration stream and destination. Rural–rural migrants are those who definitively moved to another rural part of Malawi, at least 20 km outside of an MLSFH sample area. Individuals moving to one of Malawi's three regional capitals (Mzuzu in the North, Lilongwe in the Central, and Blantyre in the South), and the third largest city (and former capital), Zomba, were considered rural–urban migrants. Finally, rural–town migrants are those who moved to the capital of one of Malawi's 22 districts. The MHM also identified return migrants, who were reported to be permanent migrants in migration tracking but were found in their MLSFH village of previous residence by the MHM in 2013.
The data collection and research conducted by MLSFH and MHM was approved by the Institutional Review Boards at the University of Pennsylvania and Tulane School of Public Health and Tropical Medicine, respectively. Both studies were approved in Malawi by the College of Medicine Research Ethics Committee or the National Health Sciences Research Committee.
We conduct our analysis in three steps. First, we examine whether HIV-positive individuals are more likely to migrate than those who are HIV negative. To do so, we use the longitudinal MLSFH data from 2004 (the year when HIV testing started for the MLSFH) through 2010. In this analysis, we focus on two variables: to establish the time-order between these measures, the dependent variable, migration, is measured from a future wave and HIV infection is measured at a prior wave, that is, before migration.
Using this approach, we pool the four most recent waves of MLSFH data (from 2004 to 2010) and run random effects logistic regressions in which the dependent variable is future migration (measured as migrant or nonmigrant), and the independent variable of primary interest is HIV status before migration. Random effects regression is used to account for correlation in the residual due to multiple observations of the same individual over time. In our first regressions, we control for age, sex, and a time trend in migration (measured by MLSFH survey interval). We then add two additional variables: previous migration experience, measured as having lived elsewhere for 6 months or more since the age of 15 and an interaction term for HIV positive status and previous migration history. The second regression permits us to examine whether those who were HIV positive and previously moved are more likely to move again. Results are shown in odds ratios (ORs).
The next step is to identify the destination for HIV-positive migrants. We use a similar approach as mentioned above, in which HIV status is measured prior to migration. But instead of using the binary measure of migrant or nonmigrant, we divide migration by stream: rural–rural, rural–town, and rural–urban migration (reference category is nonmigrants). With this four-category-dependent variable, we run random effects multinomial logistic regressions, using the same pooled MLSFH data from 2004 to 2010. As before, we also control for time trend, age, and sex.
Finally, we examine whether return migrants are also more likely to become HIV positive than those who remain in rural areas of origin. We again use the same lagged-variable regression approach, with HIV status measured before migration, and control for age, a quadratic age measure, sex, and a time trend. The dependent variable for these regressions is a three-category measure of nonmigrant (reference category), permanent migrant, and return migrant. Results for the final two sets of regressions are shown in relative risk ratios (RRRs).
The most recent wave of MLSFH, in 2010, sought to interview 5914 individuals, of whom 3701 (62.5%) successfully completed the survey, and 1096 (18.5%) had moved elsewhere by 2010. In previous MLSFH waves, 8.4% (464) moved by 2008 and 4.4% (217) by 2006. A total of 4265 respondents have full HIV testing and survey information from at least one MLSFH wave since the start of HIV testing in 2004. HIV prevalence in the most recent wave of testing, 2008, was 9.8% (Table 1).
The 1096 migrants by 2010 formed the initial target sample for the MHM. As shown in Table 1, the location was unknown for 77 migrants (7.0%), an additional 83 (7.6%) moved internationally, and 39 (3.6%) died. After omitting these, the final target sample was 897. Data collection took place in 2013; the study successfully found and interviewed 723 migrants (80.6%), of whom 94.1% (680) were tested for HIV; the final response rate is therefore 62.5% (680/1096). Nearly one-third (31.0%) were return migrants. Most moved to other rural areas (67.9%), 170 (23.5%) moved to towns, and 62 (8.6%) moved to one of Malawi's four largest cities. HIV prevalence among migrants was 14.3% in 2013. The total sample size used in our regression analysis is 4265, which includes all MLSFH men and women who were interviewed and tested for HIV at least once between 2004 and 2008 by the MLSFH, among whom a total of 897 eventually moved (and had available traceable information) by 2010 and formed the final target sample for the MHM study (i.e., a total of 897 migrants and 3368 nonmigrants by 2010).
HIV-positive individuals selected into migration
Our first set of results show that HIV-positive individuals are significantly more likely to migrate than those who are HIV negative. As shown in Table 2, model 1, HIV-positive individuals had 2.75 times greater odds of migrating by the next MLSFH wave than those who were HIV negative [OR 2.75; 95% confidence interval (CI) 1.89–4.01]. The relationship between HIV-positive status and migration persists even after controlling for previous migration history: results for model 2 show that, although those who migrated before have greater odds of moving again, individuals who moved before and are HIV positive are not significantly more likely to migrate. We also find a nonlinear relationship between age and future migration, in which the likelihood of migration first decreases for young and then increases at older ages.
Migration of HIV positive by destination
Table 3 shows results for multinomial random effects regression, in which the dependent variable is categorized by migration destination (rural, town, and urban). Our results compare these migration destinations with the nonmigrants. Results show strong associations between HIV-positive status and all migration destinations; those who are HIV positive are more likely to move to another rural area, a town, and a city (compared with not moving). The RRR that an HIV-positive individual will be a rural–rural migrant rather than a nonmigrant is more than four times greater than it is for an HIV-negative respondent (RRR 4.09; 95% CI 1.68–9.97). Similarly, being HIV positive significantly increases the RR that respondent will be a rural–town migrant (RRR 3.62; 95% CI 1.24–10.54) and a rural–urban migrant (RRR 6.28; 95% CI 1.77–22.26) instead of nonmigrant. Regressions comparing HIV infection across migration destination did not show statistically significant differences (results not shown). Taken together, therefore the results in Table 3 indicate that an HIV-positive status raises the odds of migrating to all the different destinations – rural areas, towns, and cities – rather than increasing only the odds of one specific migration flow (such as rural-to-urban). We also find the same convex nonlinear relationship between age and all three migration destinations. As indicated by the measures of MLSFH wave, the likelihood of migration increases over time for all destinations.
Migration of HIV positive by permanent or return migration
An examination of migration by permanence shows that HIV-positive individuals are more likely to be both return migrants and permanent migrants than nonmigrants. Table 4 shows results of multinomial random effects regressions, in which the RRR that an HIV-positive individual will be a permanent migrant rather than a nonmigrant is more than two times greater than it is for an HIV-negative respondent (RRR 2.58; 95% CI 1.67–3.37); being HIV positive significantly increases the RR that a respondent will move and return to the MLSFH village of origin instead of not moving (RRR 3.21; 95% CI 1.59–5.18). Similar to the above finding, therefore, Table 4 suggests that an HIV-positive status increases the odds of both permanent and temporary migration, rather than merely one type of migration. For other measures in these regressions, we again find the same convex nonlinear relationship between age and all three migration destinations. The likelihood of return migration increases over time (RRR 1.71, 95% CI 1.50–1.95). We also find differences in migration permanence by sex, in which women are less likely to be both return and permanent migrants than men.
We find that migration is not simply an independent risk factor for HIV: There are significant differences in HIV status between nonmigrants and migrants prior to migration. HIV-positive individuals are more likely to leave their village than those who are HIV negative. We also find distinct patterns by destination and permanence: HIV-positive individuals are more likely to move to both rural areas and cities than they are to stay in rural areas of origin, and the HIV positive are more likely to move permanently and cyclically than to consistently reside at their rural homes. Hence, geographic mobility and migration per se, rather than one specific type of migration flow (e.g., rural-to-urban or permanent migration) are elevated subsequent to becoming infected with HIV in our rural Malawi study population, suggesting that migration from rural areas contributes both to the rural-to-rural spread of HIV and the higher HIV prevalence that is found in peri-urban and urban areas.
The goal of this study is not to refute the possibility that migration can cause HIV infection, as this may have been the predominant pattern early in the epidemic and may continue to be important in some settings. Instead, we seek to demonstrate the importance of a longitudinal approach to studying the relationship between HIV infection and migration, and the value of distinguishing between different types of migrants. With cross-sectional postmigration data only, one might mistakenly conclude that the differences in HIV infection between migrants and nonmigrants were primarily due to migration instead of differences that existed prior to migration. Our results based upon our longitudinal data support the assertion that higher HIV prevalence among migrants is driven at least in part by their HIV status prior to migration. Therefore, higher HIV prevalence among migrants is likely explained by two pathways: greater likelihood of HIV-positive individuals to migrate and the greater likelihood of individuals who move to become HIV infected. Past literature has primarily focused on the latter group, under the assumption that individuals typically become HIV positive after migration, rather than before.
We also demonstrate the leverage that more detailed information on migrant destination and permanence can provide in studies of migrant health. Not only are HIV-positive individuals more likely to move to urban areas, they are also more likely to move to other rural areas. Moreover, they are more likely to return to the village in which they were first interviewed. The analysis of HIV infection and migration destination also suggests that greater HIV prevalence among migrants in cities may not be due to changes in risk behavior or exposure, as many are HIV positive before moving. These patterns of mobility are likely to be important for informing adherence to antiretroviral treatment (ART).
It is beyond the scope of this research to identify why HIV-positive individuals are more likely to migrate. Previous studies suggest that the higher HIV prevalence among migrants before moving is due to HIV-positive individuals who disproportionately move after marital dissolution [18,28]. As migration and marriage are closely linked in Malawi , this relationship may not hold in settings with different patterns. The link between marriage, HIV infection, and migration is supported by the information on reason for migration from the MHM study, in which marriage-related reasons made up a substantial percentage: 38.5% of all migrants moved to start a new marriage or as a result of marital dissolution (24.8% of men and 48.5% of women) (see  for a full tabulation and more details). It is also possible that the greater rural–urban migration among HIV positive could be motivated by better access to ART in urban areas . ART access may also explain the greater HIV prevalence among return migrants, who go into the city for treatment and later return: With successful treatment, they remain in the numerator. In addition to better ART access, HIV-positive individuals could seek anonymity in urban areas and freedom from scrutiny about HIV-positive status that characterizes rural Malawi . However, our findings clearly show that it is not only such rural-to-urban migration that is elevated among HIV-positive individuals, but it is also rural-to-rural and rural-to-town migration, indicating that better health infrastructure and the anonymity of urban areas are not the only aspects driving the migration decisions of HIV-positive individuals.
Despite our more robust measures of migration, some potentially important aspects of the migration process are not included here. We cannot fully capture migration streams, because our population does not include migrants originating in urban areas. We also acknowledge that other features of migration not included here may be importantly related to the relationship with HIV infection (duration at destination, distance traveled, and reason for migration); these should be the topic of future research.
Our results have important implications for HIV/AIDS research and prevention programs in SSA. Recent modeling studies have shown that the epidemic can continue to be fueled by circular migration if migrants engage in high-risk behavior at both origin and destination  and when HIV-positive individuals marry someone who is HIV negative ; thus, change in marriage and sexual behavior after migration may affect the future course of the epidemic in SSA. Although some have called for targeting migrants in HIV-prevention campaigns, our results suggest that it may be too late to prevent HIV infection if many migrants are already HIV positive. Instead of identifying migrants as a population at greater risk of becoming HIV positive, programs may instead seek to inform migrants of locations to access HIV antiretroviral therapy and other resources to support their HIV-positive status. In addition, our results suggest that, although migrants to urban areas have been the predominant focus of the literature, it is not only rural-to-urban migrants who have higher HIV prevalence, but rural migrants to other destinations as well.
The authors thank Invest in Knowledge Initiative, which collected data for the MHM and MLSFH, as well as two reviewers, whose helpful comments improved this research. This research was funded by National Institute of Child Health and Development (NICHD) (MHM grant no. R21HD071471-01), (MLSFH grant nos. R03 HD05 8976, R21 HD050652, R01 HD044228, R01 HD053781, and R24 HD-044964). All authors contributed to the study design and data collection. P.A. developed the research question, conducted the statistical analysis, and wrote the first draft of the article. All authors provided comments and edits on the manuscript at each stage.
Conflicts of interest
There are no conflicts of interest.
1. Boerma JT, Urassa M, Nnko S, Ng’weshemi J, Isingo R, Zaba B, et al. Sociodemographic context of the AIDS epidemic in a rural area in Tanzania with a focus on people's mobility and marriage
. Sex Transm Infect
2002; 78 (S1):i97–i105.
2. Lagarde E, Van Der Loeff MS, Enel C, Holmgren B, Dray-Spira R, Pison G, et al. Mobility and the spread of human immunodeficiency virus into rural areas of West Africa
. Int J Epidemiol
3. Li X, Zhang L, Stanton B, Fang X, Xiong Q, Lin D. HIV/AIDS-related sexual risk behaviors among rural residents in China: potential role of rural-to-urban migration
. AIDS Educ Prev
4. Lurie MN, Williams BG, Zuma K, Mkaya-Mwamburi D, Garnett G, Sturm AW, et al. The impact of migration on HIV-1 transmission in South Africa: a study of migrant and nonmigrant men and their partners
. Sex Transm Dis
5. Maticka-Tyndale E, Elkins D, Haswell-Elkins M, Rujkarakorn D, Kuyyakanond T, Stam K. Contexts and patterns of men's commercial sexual partnerships in northeastern Thailand: implications for AIDS prevention
. Soc Sci Med
6. McGrath N, Eaton JW, Newell ML, Hosegood V. Migration, sexual behaviour, and HIV risk: a general population cohort in rural South Africa
. Lancet HIV
7. Nunn A, Wagner H, Kamali A, Kengeya-Kayondo J, Mulder DW. Migration and HIV-1 seroprevalence in a rural Ugandan population
8. Pison G, Le Guenno B, Legarde E, Enel C, Seck C. Seasonal migration: a risk factor for HIV in rural Senegal
. J Acquir Immune Defic Syndr
9. Welz T, Hosegood V, Jaffar S, Batzing-Feigenbaum J, Herbst K, Newell ML. Continued very high prevalence of HIV infection in rural KwaZulu-Natal, South Africa: a population-based longitudinal study
10. Yang X, Derlega VJ, Luo H. Migration, behaviour change and HIV/STD risks in China
. AIDS Care
11. Barongo LR, Borgdorff M, Mosha F, Nicoll A, Grosskurth H, Senkoro KP, et al. The epidemiology of HIV-1 infection in urban areas, roadside settlements and rural villages in Mwanza region, Tanzania
12. Faria NR, Rambaut A, Suchard MA, Baele G, Bedford T, Ward MJ, et al. The early spread and epidemic ignition of HIV-1 in human populations
13. Lurie MN, Williams BG. Migration and health in Southern Africa: 100 years and still circulating
. Health Psychol Behav Med
14. Orubuloye IO, Caldwell P, Caldwell JC. The role of high-risk occupations in the spread of AIDS: truck drivers and itinerant market women in Nigeria
. Int Family Plann Perspect
15. Crush J, Williams B, Gouws E, Lurie M. Migration and HIV/AIDS in South Africa
. Dev Southern Africa
16. Deane KD, Parkhurst JO, Johnston D. Linking migration, mobility and HIV
. Trop Med Int Health
17. Palk L, Blower S. Brief Report: mobility and circular migration in Lesotho: implications for transmission, treatment, and control of a severe HIV epidemic
. JAIDS J Acquir Immune Defic Syndr
18. White RG. Commentary: what can we make of an association between human immunodeficiency virus prevalence and population mobility?
. Int J Epidemiol
19. Cassels S, Manhart L, Jenness SM, Morris M. Short-term mobility and increased partnership concurrency among men in Zimbabwe
. PLoS One
20. Kwena ZA, Camlin CS, Shisanya CA, Mwanzo I, Bukusi EA. Short-term mobility and the risk of HIV infection among married couples in the fishing communities along Lake Victoria, Kenya
. PLoS One
21. Setel PW. A plague of paradoxes: AIDS, culture and demography in northern Tanzania
. Chicago, IL: The University of Chicago Press; 1999.
22. Coffee M, Lurie MN, Garnett GP. Modelling the impact of migration on the HIV epidemic in South Africa
23. Helleringer S, Kohler H-P, Chimbiri A. Characteristics of external/bridge relationships by partner type and location where sexual relationship took place
24. Wolffers I, Fernandez I, Verghis S, Vink M. Sexual behaviour and vulnerability of migrant workers for HIV infection
. Cult Health Sex
25. Floyd S, Crampin A, Glynn J, Mwenebabu M, et al. The long-term social and economic impact of HIV on the spouses of infected individuals in northern Malawi
. Trop Med Int Health
26. Khan MR, Patnaik P, Brown L, Nagot N, Salouka S, Weir SS, et al. Mobility and HIV-related sexual behavior in Burkina Faso
. AIDS Behav
27. Coast E. Local understandings of, and responses to, HIV: rural–urban migrants in Tanzania
. Soc Sci Med
28. Anglewicz P. Migration, marital change and HIV infection in Malawi
29. Kohler H, Watkins S, Behrman J, Anglewicz P, Kohler I, Thornton RL, et al. Cohort profile: the Malawi Longitudinal Study of Families and Health (MLSFH)
. Int J Epidemiol
30. WPS, Anglewicz P, VanLandingham M, Manda-Taylor L, Kohler HP. Cohort profile: the Migration and Health in Malawi (MHM) study. PSC Working Paper Series
31. Watkins S, Behrman JR, Kohler HP, Zulu EM. Introduction to “Research on demographic aspects of HIV/AIDS in rural Africa”
. Demogr Res
32. Anglewicz P, Adams J, Obare F, Kohler H, Watkins S. The Malawi Diffusion and Ideational Change Project 2004–2006: data collection, data quality, and analysis of attrition
. Demogr Res
33. Bignami-Van Assche S, Reniers G, Weinreb AA. An assessment of the KDICP and MDICP data quality: interviewer effects, question reliability and sample attrition
. Demogr Res
34. Obare F, Fleming P, Anglewicz P, Thornton R, Martinson F, Kapatuka A, Kohler H-P. HIV incidence and learning one's HIV status: evidence from repeat population based voluntary counseling and testing in rural Malawi
. Sex Trans Infect
35. Reniers G. Divorce and remarriage in rural Malawi
. Demogr Res
36. Watkins SC. Navigating the AIDS epidemic in rural Malawi
. Popul Dev Rev
37. Reniers G, Armbruster B. HIV status awareness, partnership dissolution and HIV transmission in generalized epidemics
. PLoS One