The HIV epidemic in India displays significant heterogeneity with respect to geographical location and risk behaviour, with female sex workers (FSW), their male clients, men who have sex with men and injection drug users being at highest risk [1–7]. The south-west Indian state of Karnataka is considered one of the most severely affected [1,2,8], with heterosexual transmission thought to be the main infection route . With an estimated 2.6% adult HIV prevalence in 2004 , Bagalkot district in northern Karnataka is considered a high-prevalence district. The HIV prevalence in this primarily rural district varies across talukas (subdistrict administrative blocks), with the 2004 HIV prevalence being 4.9%, 2.9% and 1.2% in three adjacent talukas . Seasonal migration and the widespread practice of sex work have been suggested as important drivers of the epidemic [10–15].
The link between migration and HIV is complex and not well understood. Mobility per se does not necessarily favour the spread of HIV, but often creates a context (prolonged separation from family, isolation, greater anonymity) in which migrant men may engage in risky sexual practices, such as visiting FSW . In addition, younger migrants returning with more disposable income may seem more attractive to women and have more sexual partners . Several studies have found that HIV prevalence or risky sexual practices are higher among migrants compared with non-migrants [14,17–22] although others have not . One Indian study also suggested that female partners of male migrants may be at higher risk of HIV infection as a result of risky local sexual behaviour while partners are away .
In northern Karnataka, seasonal migration for employment is an important component of livelihood strategies, because recurring droughts have led people to move in search of employment instead of agriculture [14,16,23]. Evidence suggests that migration is common among both men and FSW, 19% of FSW in Bagalkot and approximately 45% of men and 11–31% of FSW in neighbouring districts reported being migrants [16,24]. Understanding the impact of migration is important for programmatic reasons; however, empirical data on migrant sexual behaviour at home and in the migration destination are limited. This study thus uses mathematical modelling to explore the potential impact of changes in local sexual behaviour and migration-related risk factors on HIV prevalence using six different migration scenarios in Bagalkot district. We also explore the extent to which migration could explain heterogeneity in prevalence between talukas and assess the potential impact of migration on the effectiveness of core group interventions.
In this paper, the three talukas in Bagalkot noted above are designated A, B and C, with taluka A having the lowest HIV prevalence (1.2%) and taluka C the highest (4.9%).
The study used a deterministic compartmental model of heterosexually transmitted HIV infection. The full model structure and equations are described in the Appendix. The model included four stages of infection (initial/acute HIV infection, asymptomatic, symptomatic and AIDS), and divided men and women into two levels of sexual activity representing individuals who did not engage (low risk) or did engage in commercial sex (FSW/clients) in their home taluka; the population grew exponentially at 2% per year [25–28] without AIDS. Susceptible individuals became infected with a force of infection that depended on their rate of sexual partner acquisition, the HIV prevalence in their sexual partners, and the probability of HIV transmission per partnership. Once infected, individuals progressed through HIV stages at rates equivalent to the inverse of the average duration spent in each stage. Individuals left the population when they ceased sexual activity or died of AIDS.
Clients and FSW were assumed to have many one-time contacts with each other . Low-risk women had few partnerships with many contacts with low-risk men or clients, whereas low-risk men only formed partnerships with low-risk women when in their home taluka. FSW and clients ceased high-risk behaviour, returning to low-risk activity at a fixed rate, and were replaced by low-risk men and women in equal quantity proportionate to the distribution of infection in each group. A fraction of FSW, clients or low-risk men were assumed to migrate once per year to the migration destination for a fixed period of time and engage in commercial sex before returning.
Six seasonal migration scenarios were defined based on three risk populations migrating: (1) all men; i.e. men who visit FSW both in their places of origin and destination (clients) and who visit FSW only in the migration destination (low-risk clients) (MIG-M); (2) clients only (MIG-CL); and (3) FSW (MIG-SW); and on two behaviour scenarios for the local population while migrants are away (Table 1) developed using the following reasoning.
When FSW migrate, the seasonal decline in local FSW supply either results in: (1a) the remaining FSW seeing more clients to meet client demand, which remains unchanged; or (1b) clients reducing their number of visits to FSW; when clients/low-risk clients migrate, the seasonal decline in the client population size either results in: (2a) local clients/low-risk men increasing their number of FSW visits to meet the unchanged FSW supply; or (2b) local FSW/low-risk women reduce their number of partners. In 1a/2a, the total number of FSW–client contacts and low-risk men/low-risk women partnerships remain constant, whereas in scenarios 1b/2b, the number of FSW contacts and low-risk men/low-risk women partnerships decline proportionately to migrant population size. In scenarios 1a/2b and 1b/2a, clients or low-risk men (M_demand) or FSW and/or low-risk women (F_demand), respectively, control the local demand for sex while migrants are away.
Non-migration related parameter values
Table 2 summarizes demographic, behavioural, epidemiological and migration parameter values, based on empirical survey data in FSW and the general population in Bagalkot district [10,24] and Karnataka (2005) (Integrated Behavioural and Biological Surveys, unpublished, 2005; Sexual Behavioural Surveys, unpublished, 2005) and literature reviews.
By 2006, the number of FSW in taluka A was enumerated as 835 (∼1.6% of sexually active female population) , and 11.4% of men reported ever visiting FSW in anonymous polling booth studies . Assuming that clients buy sex for 20 years (Integrated Behavioural and Biological Surveys, unpublished, 2005), we estimated that approximately 8.8% (3550) of sexually active men were clients in 2004 (see Appendix). The mean number of client visits per FSW was estimated to be approximately 510–520 per year . The number of FSW visits per client (∼120) was derived by dividing the total number of client contacts reported by FSW by the client population.
Considerable uncertainty surrounds the amount of condom use by FSW and clients in Karnataka before 2004–2005. As a result of increased exposure to interventions, condom use has become more common. In 2005, the fraction of FSW reporting always using condoms with non-regular clients varied between 64 and 75% in districts of Karnataka, compared with 50–70% for repeat clients, 10–53% for regular partners and 6–13% for husbands/cohabiting partners (Sexual Behavioural Surveys, unpublished, 2005). We assumed an overall condom use of 30% of FSW–client contacts from 2004.
Migration-related parameter values
In Bagalkot, 87% of FSW live and work in rural areas and are concentrated in the talukas with a greater density of irrigated land , possibly reflecting the density and movements of agricultural labourers. A total of 18.6% of FSW reported working outside the district; 15.5% reported this in taluka A . In other districts within Karnataka, 10.5–31.3% of FSW reported ever doing sex work in other places (Sexual Behavioural Surveys, unpublished, 2005). As districts nearest Bagalkot had the highest fraction of migrants, we set the default fraction of yearly FSW migrants to 10% (sensitivity analysis range: 5–20%). In Bagalkot district (2004), 34% of men reported travelling for work (only 3% for more than one month in the past year) . This may be an underestimate if migrants were away during the survey. In Bijapur district (of which Bagalkot was part until 1997), 41% of married and 47% of unmarried men reported being migrants in 2005 . The default fraction of client migrants was set to 30% (varied 15–60%). As the estimated duration away ranged from 3 to 8 months among migrants in Bijapur , the default time away was set to 4 months (varied 2–8 months). A smaller fraction of low-risk men, 10% (varied 5–20%) were assumed to migrate.
The model was initially calibrated to reflect overall male, female and FSW/client HIV prevalence in the lowest prevalence taluka (A) in 2004, without migration. Each migration scenario was then simulated at the start of the epidemic to assess its effect on the overall epidemic, using default parameter values.
To assess the independent impact of migration (different sexual behaviour assumptions of the local population while migrants are away: M_demand/F_demand), the default rate of commercial sex by FSW and clients in the migration destination was set to that in the place of origin, and the HIV epidemic in the migration destination was similar to the local epidemic (Fig. S2, Fig. S3, see Appendix).
The sensitivity of the 2004 and equilibrium HIV prevalence to each migration parameter was assessed by varying each through their predefined ranges (Table 2) for each scenario. The preventive potential of targeting interventions at migrants was assessed with the population attributable fraction (PAF) due to migration, defined as the difference in the cumulative number of new infections since the beginning of the HIV epidemic with and without migration, divided by the former value; and the prevented fraction, defined as the difference between the cumulative number of new infections (2004–2015) with and without the intervention, divided by the former value.
To assess the extent to which migration could reduce intervention impact, the prevented fraction was calculated for the most influential migration scenario (with the highest PAF: MIG-M/F_demand) when a maximally effective intervention was introduced in high-risk groups in 2004 (87% effective condoms used always) in (1) all migrants' contacts locally and in the migration destination, (2) all migrants' contacts only in the migration destination, (3) all non-migrants' contacts locally and (4) an equivalent fraction of non-migrants' contacts as migrants' locally. Results are shown for both default migration and an extreme high-risk scenario (i.e. clients have twice the number of FSW contacts in the migration destination as in the place of origin and are exposed to FSW with HIV prevalence that is 1.5 times higher than in the place of origin in 2004 and at equilibrium; Fig. S2, Appendix).
Figure 1 shows the modelled HIV prevalence in the absence of migration, which agreed quite well with the observed overall FSW and client HIV prevalences in taluka A in 2004, when HIV prevalence began to plateau.
Default migration scenarios: men migrating
Figure 2 shows the 2004 and equilibrium HIV prevalences in the local populations for the six default migration scenarios. For different risk populations (local clients/FSW, Fig. 2a; low-risk women, Fig. 2b), migration had the largest impact when men or only clients migrated (MIG-M/MIG-CL), and local FSW determined the demand for commercial sex (F_demand) (Fig. 2). Here, local clients increased their demand for commercial sex by 43% (compensating for their 30% population size decrease). The impact was more pronounced earlier in the epidemic than at equilibrium. With default migration parameter values, the 2004 local client prevalence increased from 13% without to 23% with migration (77% increase), and the equilibrium HIV prevalence increased from 17% to 22% (29% increase) (Fig. 2a). Similarly, local 2004 and equilibrium FSW HIV prevalence increased from 29% and 39% without migration, to 46% and 43%, respectively, (by 50% and 10%) (Fig. 2a). Interestingly, the local low-risk female 2004 HIV prevalence doubled if clients and low-risk men migrated, and increased to a smaller extent if only clients migrated (Fig. 2b).
Male migration had a smaller impact on local client, FSW and low-risk female HIV prevalence if local clients rather than FSW drove the demand for commercial sex (M_demand). In this scenario, there was even a marginal decline in epidemic size, as a result of a 30% decrease in the number of local client visits to FSW (equal to the default migrant client fraction). In this case, local client 2004 and equilibrium HIV prevalence decreased from 13% without to 12% with migration, and local FSW 2004 HIV prevalence was reduced from 30% to 27%, and from 41% to 35% at equilibrium, respectively (Fig. 2a).
Default migration scenarios: female sex workers migrating
Compared with the previous male migration scenarios, FSW migration had a smaller impact on the HIV epidemic (Fig. 2c,d). The largest impact occurred among local FSW and clients when the demand for commercial sex was driven by local clients (M_demand), because this forced local FSW to increase their frequency of sexual contacts by 11% (compensating for the 10% population size decrease). Local FSW and client HIV prevalence increased from 30% and 11% without migration to 36% and 15% with migration in 2004, but increased by only approximately 1% at equilibrium (Fig. 2c). When the demand for sex work was driven by local FSW (F_demand), FSW migration reduced the local epidemics, albeit negligibly (Fig. 2c). The impact of both FSW migration scenarios on low-risk women was very modest (<5%) (Fig. 2d).
Sensitivity analysis of varying migration parameters
Figure 3 shows the independent impact of varying each migration parameter through its prespecified range on the 2004 overall HIV prevalence.
The largest change in the overall 2004 HIV prevalence occurred when either a fraction of all men (MIG-M) or only clients (MIG-CL) migrated, and local FSW determined the demand for sex (F_demand). In both scenarios, HIV prevalence was most sensitive, in order, to migrant population size, duration away, rate of FSW contacts and FSW HIV prevalence in the migration destination. When local clients determined the demand for commercial sex (M_demand), however, varying migrant size and duration away had a smaller effect than varying the rate of FSW contacts and FSW HIV prevalence in the migration destination (Fig. 3). When FSW migrated, the modelled 2004 HIV prevalence was less sensitive to any of the parameters varied (Fig. 3). Although the equilibrium HIV prevalence was less sensitive to variation in the migration parameters than in 2004, similar qualitative relationships were obtained (results in Appendix).
Only when men or clients migrated, and women determined the local demand for sex, was the overall 2004 modelled HIV prevalence brought within range of taluka B (2.9%; 2.2–3.6%). Even with the highest range values explored, none of the migration scenarios brought 2004 prevalence in taluka A within range of taluka C (4.9%; 3.5–6.3%).
Fraction of infections caused by migration
Figure 4a shows the PAF (1976–2015) caused by migration for each default migration scenario. The largest PAF was obtained when all men (MIG-M) or only clients (MIG-CL) migrated and women determined the local demand for sex (F_demand). These scenarios produced approximately two times and 1.6 times more infections than without migration, respectively. All other migration scenarios produced less than 16% more infections than the scenario without migration. Figure 4b shows the prevented fraction (2004–2015) when all men migrated and women determined the local demand for sex, the most influential (highest PAF) migration scenario. With default migration, targeting all migrants in the migration destination or an equal number of locals (∼30% of the total population) reduced infections by 18–20% over the 11-year period. Interestingly, protecting all local contacts perfectly was 1.5 times more effective than protecting all migrant contacts locally and in the migration destination, and almost three times more effective than protecting all migrant contacts in the migration destination only. As expected, with a very high migration-related risk, the benefit of targeting migrant contacts increased; protecting all migrant contacts in the migration destination and at home prevented the most infections (half), followed by protecting all local and all migrant contacts only in the migration destination. Preventing an equal number of locals and migrants prevented fewest infections, as a result of reduced risk behaviour among locals.
This study showed that the impact of different seasonal migration patterns on HIV prevalence in Bagalkot district, Karnataka, varied substantially across scenarios. As expected, migrant population size, duration of time away, the number of contacts while in the migration destination, and the HIV prevalence in the migration destination influenced prevalence. The extent to which these factors could have a substantial impact on the HIV epidemic or explain heterogeneity in HIV prevalence across talukas depended strongly on who migrated (all men, clients or FSW), and also on who determined the demand for commercial sex in the local population while migrants were away.
Influence of male/clients migration
Migration of men from the general population or clients could cause the largest increase in HIV prevalence among high-risk groups and low-risk women if local FSW determine the demand for commercial sex, if local clients increase their visits to FSW and low-risk men interact with FSW when migrating. The migration of these populations could, however, also result in reduced HIV transmission if the local demand for sex work is driven by local clients, because in this case local FSW have fewer client partners while migrant clients are away.
Influence of female sex worker migration
Even when local FSW recruit more clients when migrant FSW are away, the impact of FSW migration is likely to be modest. Given that local FSW and clients experience the same proportional change in numbers of partners during migration, the smaller impact of FSW relative to client migration (when locals recruit more partners while migrants are away) happened because based on the available information, we assumed that the population size of FSW was smaller and fewer FSW (10%) than clients (30%) migrated. In sensitivity analysis, when the same fraction of clients as FSW migrated (10%; 5–20%), similar overall and equilibrium HIV prevalence were observed.
Potential impact of prevention targeted to migrant population
When all men or just clients migrate, and FSW determine the demand for sex, migration increased the number of HIV infections by 50–100% compared with no migration, but in other cases migration produced fewer infections. The potential usefulness of interventions targeted to areas where migration is common thus depends on the migration pattern. As HIV transmission was most sensitive to migrant size in the most affected scenario, an effective intervention's impact would probably be substantial when numbers of migrants are high.
Increased condom use among migrants both in their place of migration and at home should have the most impact on HIV transmission. Even when migration caused a high fraction of infections, however, results indicated that a core group intervention introduced only locally because of the difficulty of reaching migrant populations could still be very beneficial. It is important to recognize that these results are an estimate of maximum impact with 100% condom use (87% effective) and the actual impact of an intervention would be reduced.
Migrants or migration
Our findings suggest that local changes in sexual behaviour resulting from migration can be more important for determining the effect of migration on HIV transmission than the possible increase in risky behaviour by migrants while away. Interestingly, even if migrants do not engage in riskier behaviour in their place of migration, migration can still increase the spread of HIV, especially earlier in the epidemic, because of the influence on sexual behaviour of the local population. Increases in risk while in the migration destination had less of an impact on overall HIV prevalence compared with a larger migrant population, or amount of time away, in the two scenarios with the largest impact (clients/men migrated, FSW determined the demand for sex). Migration could explain a large fraction of the variability in HIV prevalence across different talukas, but only under these two migration scenarios.
These results do not mean that risk behaviour in the migration destination is not important. Existing information suggests that the relative importance of local sexual behaviour change and of migrant risk behaviour may be site specific, because the presence and impact of these factors varies in different situations. Risk behaviour was similar in migrants and non-migrants in a study in Rajasthan, north India, although migrants had higher partner change rates and opportunities for high-risk behaviour in the migration destination . Migration to areas of high HIV prevalence was also common for this population . Migrants were not more likely to have sex with non-regular partners or FSW than non-migrants in Bijapur district . Empirical studies from South Africa have found that migrant status was associated with HIV infection in men  and women ; and a modelling study, on which this study builds, found that increased risk behaviour in the migration destination had a substantial impact on HIV prevalence earlier in the epidemic . Although the assumptions made in the latter study regarding how local sexual behaviour changed while migrants were away is unknown, our study also found that migration had a larger impact earlier in the epidemic. Future studies should explicitly quantify the difference in partner numbers that non-migrants have when migrants are away, and if these contacts are made locally, as in this study, or outside the home taluka. The fraction of men who only visit FSW when migrating should be quantified, as HIV prevalence was much more sensitive to migrant size when these ‘low-risk clients’ were included. Collecting these data, as well as the relative difference in numbers of high-risk partners locally and in the migration destination would strengthen projections of the impact of migration and increase study validity.
It is currently difficult to quantify precisely the contribution of migration to HIV transmission, as data are limited. The extent to which FSW or clients in northern Karnataka determine the demand/supply for commercial sex as population sizes fluctuate is unknown. In this study, local FSW or clients were assumed to determine the demand for commercial sex, whereas it is likely to be somewhere between these extreme assumptions. We have not considered in-migration scenarios in which the out-migrant population is effectively replaced or the local population is increased, the effects of which would probably be between our extreme assumptions unless in-migrants had a much higher HIV prevalence or prompted increases in local commercial sex. We looked at annual migration, but further research could examine an increased frequency of return by migrants, as this has been found to be an important factor in increasing HIV prevalence locally in the South African setting , possibly because migrants may be in the acute, very infectious phase of HIV infection upon return. It is therefore important also to collect information on the frequency of travel and average duration spent in migration sites.
In summary, our findings suggest that migration could play a significant role in the HIV epidemic in northern Karnataka under very specific assumptions. Understanding how local sexual networks change in areas where migration is common is crucial for understanding the overall impact of migration, and for designing HIV prevention interventions locally and in the destination of migration.
K.N.D. conducted the analysis and wrote the paper. P.V. provided extensive feedback and edits and helped with the project's progression. S.M. helped with the project's conception and implementation in different stages and provided extensive feedback and edits; S.M. was also extensively involved in the organization of the projects from which data were gathered and used to estimate model parameters. B.M.R. organized the data collection for data used to estimate model parameters and provided feedback. J.F.B. helped with the project's conception and provided feedback at various stages; J.F.B. was also extensively involved in the organization of the projects from which data were gathered and used to estimate model parameters. M.-C.B. helped with the project's conception, methodology and implementation in different stages and provided extensive feedback and edits.
Sponsorship: Support for this study was provided by the Bill and Melinda Gates Foundation. K.N.D. was supported by a Manitoba Health Research Council graduate studentship during this research.
The views expressed herein are those of the authors and do not necessarily reflect the official policy or position of the Bill and Melinda Gates Foundation.
Conflicts of interest: James F. Blanchard receives funding from Canada Research Chairs, Health Canada.
All other authors declare no conflict of interest.
3. Arora P, Cyriac A, Jha P. India's HIV-1 epidemic. Can Med Assoc J 2004; 171:1337–1338.
4. Nagelkerke NJ, Jha P, de Vlas SJ, Korenromp EL, Moses S, Blanchard JF, et al. Modelling HIV/AIDS epidemics in Botswana and India: impact of interventions to prevent transmission. Bull WHO 2002; 80:89–96.
5. Brahme R, Mehta S, Sahay S, Joglekar N, Ghate M, Joshi S, et al. Correlates and trend of HIV prevalence among female sex workers attending sexually transmitted disease clinics in Pune, India (1993–2002). J Acquir Immune Defic Syndr 2006; 41:107–113.
8. Karnataka State AIDS Prevention Society (KSAPS), Government of Karnataka, and India–Canada Collaborative HIV/AIDS Project (ICHAP). Karnataka HIV status reports: report on HIV sentinel surveillance 2004. Bangalore: KSAPS and ICHAP; 2004.
9. Population Foundation of India (PFI), Karnataka State AIDS Prevention Society (KSAPS), India–Canada Collaborative HIV/AIDS Project (ICHAP) and Population Reference Bureau (PRB). HIV/AIDS in Karnataka, situation and response. Bangalore: PFI; 2004.
10. Becker ML, Ramesh BM, Washington RG, Halli S, Blanchard JF, Moses S. Prevalence and determinants of HIV infection in South India: a heterogeneous rural epidemic. AIDS 2007; 21:739–747.
11. Singh YN, Malaviya AN. Long distance truck drivers in India: HIV infection and their possible role in disseminating HIV into rural areas. Int J STD AIDS 1994; 5:137–138.
12. Kumar P. Migration induced HIV epidemic in Kerala, India. In: The XIIIth International AIDS Conference. Durban, South Africa, 9–14 July 2000 [Abstract MoPeD2662].
13. Changedia S, Gilada I. Relationship of migration and displacement patterns with HIV/AIDS in India. In: The XIIIth International AIDS Conference. Durban, South Africa, 9–14 July 2000 [Abstract MoPeD2661].
14. Gupta K, Singh SK. Social networking, knowledge of HIV/Aids and risk-taking behaviour among migrant workers. International Union for the Scientific Study of Population (IUSSP) Regional Population Conference on Southeast Asia's Population in a Changing Asian Context. Bangkok, Thailand: IUSSP; 2002. Available at: http://www.iussp.org/Bangkok2002/S06Gupta.pdf
. Accessed: 13 December 2007.
15. Blanchard J, O'Neil J, Ramesh BM, Bhattacharjee P, Orchard T, Moses S. Understanding the social and cultural context of female sex work in Karnataka, India. J Infect Dis 2005; 191(Suppl. 1):S139–S146.
16. Halli SS, Blanchard J, Satihal DG, Moses S. Migration and HIV transmission in rural South India: an ethnographic study. Culture, Health Sexuality 2007; 9:85–94.
17. Desmonda N, Allen CF, Clift S, Justinea B, Mzugua J, Plummera ML, et al. A typology of groups at risk of HIV/STI in a gold mining town in north-western Tanzania. Soc Sci Med 2005; 60:1739–1749.
18. Clift S, Anemona A, Watson-Jones D, Kanga Z, Ndeki L, Changalucha J, et al. Variations of HIV and STI prevalence within communities neighbouring new goldmines in Tanzania: importance for intervention design. Sex Transm Infect 2003; 79:307–312.
19. Lagarde E, Schim van der Loeff M, 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 2003; 32:744–752.
20. Bronfman MN, Leyva R, Negroni MJ, Rueda CM. Mobile populations and HIV/AIDS in Central America and Mexico: research for action. AIDS 2002; 16(Suppl. 3):S42–S49.
21. George S, Jacob M, John TJ, Jain MK, Nathan N, Rao PSS, et al. A case–control analysis of risk factors in HIV transmission in South India. J Acquir Immune Defic Syndr 1997; 14:290–293.
22. Bryan AD, Fisher JD, Benzinger TJ. Determinants of HIV risk among Indian truck drivers. Soc Sci Med 2001; 53:1413–1426.
23. Deshingkar P, Start D. Seasonal migration for livelihoods in India: coping, accumulation and exclusion. London, UK: Overseas Development Institute; 2003.
24. Blanchard JF, Halli S, Ramesh BM, Bhattacharjee P, Washington RG, O'Neil J, et al. Variability in the sexual structure in a rural Indian setting: implications for HIV prevention strategies. Sex Transm Infect 2007; 83:30–36.
25. Das NP, Bhavsar S. Population growth rate in India: emerging trend in the light of 1991 Census results. Demography India 1991; 20:227–241.
27. Census of India. Office of the Registrar General and Census Commissioner, India. Bagalkot census data 2001
. New Delhi, 2001. Available at: http://bagalkot.nic.in/2001_census.html
. Accessed: 9 May 2005.
28. Census of India. Office of the Registrar General and Census Commissioner, India. Chapter 3: Size, growth rate and distribution of the population
. New Delhi, 2001. Available at: http://www.censusindia.net/data/chapter3.pdf
. Accessed: 4 October 2005.
29. Population Research Centre (PRC), JSS Institute of Economic Research. Female commercial sex workers in Karnataka: a baseline survey. Report no. 120. Dharwad, Karnataka, India: PRC; 2003.
30. India–Canada Collaborative HIV/AIDS Project (ICHAP). Community-based HIV prevalence study in ICHAP demonstration project area, key findings. Bangalore, India: ICHAP; 2004.
31. Washington RG, Singh D, Adrien A, Blanchard JF, Moses S, Doctor R. An evidence-based approach to understanding and addressing rural to urban migration in India. In: The XVth International AIDS Conference. Bangkok, Thailand, 11–16 July 2004; [Abstract TuPeD5238].
32. Singh D, Adrien A, Washington R, Moses S, Blanchard J. Baseline results from HAMARA, a project on HIV/AIDS and rural migration in Rajasthan, India. In: The XVth International AIDS Conference. Bangkok, Thailand, 11–16 July 2004; [Abstract D10789].
33. Lurie MN, Williams BG, Zuma K, Mkaya-Mwamburi D, Garnett G, Sturm A, et al. The impact of migration on HIV-1 transmission in South Africa: a study of migrant and non-migrant men and their partners. Sex Transm Dis 2003; 30:149–156.
34. Zuma K, Gouws E, Williams B, Lurie M. Risk factors for HIV infection among women in Carletonville, South Africa: migration, demography and sexually transmitted diseases. Int J STD AIDS 2003; 14:814–817.
35. Coffee M, Lurie MN, Garnett GP. Modelling the impact of the HIV epidemic in South Africa. AIDS 2007; 21:343–350.
36. Tovanabutra S, Robison V, Wongtrakul J, Sennum S, Suriyanon V, Kingkeow D, et al. Male viral load and heterosexual transmission of HIV-1 subtype E in northern Thailand. J Acquir Immune Defic Syndr 2002; 29:275–283.
37. Wawer MJ, Gray RH, Sewankambo NK, Serwadda D, Li X, Laeyendecker O, et al. HIV-1 transmission per coital act, by stage of HIV infection in the HIV positive index partner, in discordant couples, Rakai, Uganda. J Infect Dis 2005; 191:1403–1409.
38. Baggaley RF, Garnett GP, Ferguson NF. Modelling the impact of antiretroviral use in resource-poor settings. PLOS Medicine
: e124, doi:10.1371/journal.pmed.0030124.
39. Ramesh BM, Washington R, Mondal S, Moses S, Alary M, Blanchard JF. Sex work typology and risk for HIV in female sex workers (FSWs): findings from an integrated biological and behavioural assessment in the southern Indian state of Karnataka.
In: XVIth International AIDS Conference.
Toronto, Canada, 13–18 August 2006 [Abstract WEAC0305].
high-risk sex; HIV/AIDS; mathematical models; out-migration; risk factors
© 2008 Lippincott Williams & Wilkins, Inc.