This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Public Health and Surveillance, is properly cited. The complete bibliographic information, a link to the original publication on http://publichealth.jmir.org, as well as this copyright and license information must be included.
Although the prevalence of HIV among men who have sex with men (MSM) in Vietnam has been increasing in recent years, there are no estimates of the population size of MSM based on tested empirical methods.
This study aimed to estimate the size of the MSM population in 12 provinces in Vietnam and extrapolate from those areas to generate a national population estimate of MSM. A secondary aim of this study was to compare the feasibility of obtaining the number of users of a mobile social (chat and dating) app for MSM using 3 different approaches.
This study used the social app multiplier method to estimate the size of MSM populations in 12 provinces using the count of users on a social app popular with MSM in Vietnam as the first data source and a questionnaire propagated through the MSM community using respondent-driven sampling as the second data source. A national estimation of the MSM population is extrapolated from the results in the study provinces, and the percentage of MSM reachable through online social networks is clarified.
The highest MSM population size among the 12 provinces is estimated in Hanoi and the lowest is estimated in Binh Dinh. On average, 37% of MSM in the provinces surveyed had used the social app Jack’d in the last 30 days (95% CI 27-48). Extrapolation of the results from the study provinces with reliable estimations results in an estimated national population of 178,000 MSM (95% CI 122,000-512,000) aged 15 to 49 years in Vietnam. The percentage of MSM among adult males aged 15 to 49 years in Vietnam is 0.68% (95% CI 0.46-1.95).
This study is the first attempt to empirically estimate the population of MSM in Vietnam and highlights the feasibility of reaching a large proportion of MSM through a social app. The estimation reported in this study is within the bounds suggested by the Joint United Nations Programme on HIV/AIDS. This study provides valuable information on MSM population sizes in provinces where reliable estimates were obtained, which they can begin to work with in program planning and resource allocation.
Since the first reported case of HIV in Vietnam in 1990 [
Although the overall number of new infections has declined, the prevalence of HIV among MSM has been increasing in recent years [
Estimates of the size of populations at risk of HIV are necessary to understand the scale of the epidemic and in planning appropriate interventions and allocation of resources. A number of estimates of the MSM population size have been attempted in Vietnam, targeted to limited provinces [
Besides the inherent imprecision of the methods to arrive at the MSM population size based on profiling of urbanization of regions and modeling, the previous accepted population size of MSM in Vietnam does not reveal much about how this population can be reached, how MSM network together, what regional variations exist, or the age demographics of the reachable MSM. This study attempted to directly estimate the size of the MSM population in 12 provinces in Vietnam using the social app multiplier method and extrapolate from those areas to generate a national population estimate of MSM. A secondary aim of this study was to compare the feasibility of obtaining the number of users on a social app for MSM using 3 different approaches.
This study used the social app multiplier method to estimate the population size of MSM in 12 provinces of Vietnam. This method was piloted in Ho Chi Minh City and Nghe An province in 2016 and subsequently updated and improved based on learnings from the pilot study [
The social app Jack’d was selected to provide the first count. Overall, 3 methods were used to count the total number of users on Jack’d. First, following the method used during the pilot, in each of the 12 provinces, the total number of active users on Jack’d was enumerated over a 1-month period, and the final list of active users was deduplicated using the public profile information of app users such as age, pseudonym, and avatar. When counting active users, only profiles that appeared at least twice and spaced by several days in between were included in the final count to minimize the possibility of counting short-term visitors. A second method was a capture-recapture procedure that matched active users on Jack’d at 2 different time points, using the same public profile information as used in the first method, to estimate the total number of active users on the app in each of the respective provinces. The third method was to procure the total aggregate, unduplicated, and nonidentifiable number of app users in the respective province over a period of 1 month directly from the Jack’d social app administrator.
Immediately after 1 month of counting users on the Jack’d social app, an online survey using respondent-driven sampling (RDS) recruitment strategy was conducted in the MSM community in each of the 12 provinces to find out about their use of Jack’d in the past month. RDS is a form of chain referral sampling that uses a mathematical model to approach a true random sample [
The following study procedures were followed for the RDS online survey. Between 6 and 8 seed individuals were identified in each province and given 3 coupons each to recruit their peers to respond to the RDS online survey, who would in turn be given 3 coupons each, and so on. Recruits who did not have access to the internet to complete the RDS online survey or preferred to provide answers offline were provided with a telephone number to contact a member of the investigation team who would collect information from them over the phone or in person and enter it online.
The nonidentifying information collected from the 12 RDS online surveys was analyzed to calculate the proportion of respondents who answered
Homophily tests were conducted to assess the extent of random referrals from respondents to their personal networks [
Convergence plots in RDS were 1 indicator of having sufficient data collected to get a reliable estimate. When the key estimator remained stable within 2% of the sample proportion, we predicted that additional responses collected would yield insignificant changes to the estimate [
The RDS-I estimator tends to underestimate the result if the sampling does not converge, and the Gile’s SS bootstrap tends to underestimate the result if the homophily test value is less than 1 or there is bias in seed selection [
The population size estimates were converted to a percentage of the general adult male population (15-49 years) and compared with the range of percentages reported in the 2018 Spectrum Quick Start Guide [
Ethical approval for this study was obtained from the Institutional Review Board of the Hanoi School of Public Health (Approval no. 298/2016/YTCC-HD3, IORG no. 0003239; FWA number: 0009326). The protocol was also reviewed and approved by the US Centers for Disease Control and Prevention human subjects research office.
Hanoi had the highest number of active Jack’d users among the 12 provinces at 12,848 persons, and Binh Dinh province had the lowest number of active users at 260 persons. The median age of Jack’d users in the provinces was 25 years, with 75% of the users in the 19 to 35 years age group. The median age of the RDS survey respondents was 24 years, with 75% of the users in the 19 to 32 years age group. There was no statistically significant difference between the mean age or the mean network size of respondents who answered
Number of respondent-driven sampling survey participants and respondents, their social and demographic characteristics, and summary measures of their Jack’d usage.
Provinces | Participants | Characteristics of participants | Summary measures | ||||
Survey participants, n | Eligible survey respondents, n | Mean age (years) | Mean network size | Participants who did not have sex with another man in the last 12 months but who did not prefer sex with women only among eligible survey respondents, n | Participants with a Jack’d account, n | Participants using Jack’d in the last 30 days, n | |
An Giang | 173 | 167 | 22 | 8 | 39 | 33 | 18 |
Bac Giang | 132 | 125 | 28 | 46 | 28 | 77 | 52 |
Binh Dinh | 134 | 126 | 27 | 5 | 26 | 101 | 43 |
Can Tho | 195 | 167 | 23 | 6 | 28 | 37 | 19 |
Da Nang | 193 | 167 | 24 | 5 | 17 | 114 | 80 |
Dak Lak | 123 | 121 | 24 | 8 | 52 | 41 | 21 |
Dong Nai | 238 | 227 | 25 | 7 | 15 | 69 | 17 |
Dong Thap | 217 | 212 | 24 | 9 | 61 | 47 | 33 |
Hanoi | 296 | 264 | 23 | 13 | 22 | 209 | 130 |
Hai Phong | 345 | 279 | 25 | 10 | 70 | 155 | 96 |
Nam Dinh | 132 | 122 | 25 | 10 | 21 | 84 | 65 |
Thanh Hoa | 244 | 200 | 25 | 9 | 34 | 107 | 67 |
Total | 2422 | 2177 | 24.3 | 10.6 | 413 | 1074 | 641 |
Among the 12 provinces, 4 (Dak Lak, Dong Thap, An Giang, and Can Tho) had sensitivity ratios that were outside the 0.2 to 0.8 range when using Jack’d use in the last 30 days as the estimator. Among the 12 provinces, 2 (Dong Nai and An Giang) had sensitivity ratios that were outside the 0.2 to 0.8 range when using all-time Jack’d usage as the estimator. Among the 12 provinces, 4 (Bac Giang, Can Tho, Dak Lak, and Dong Nai) had homophily values less than 1 when using Jack’d use in the last 30 days as the estimator and 1 (Nam Dinh) had homophily value less than 1 when using all-time Jack’d use as the estimator. The convergence plots in 10 provinces showed that the RDS survey samples converged; however, in 2 provinces, Dak Lak and Thanh Hoa, there was bottlenecking between seeds (see
Analysis of the data shows that on average 37.5% of MSM in the 11 provinces with reliable estimates had used Jack’d in the last 30 days (95% CI 27.0-47.9). Among these provinces, Can Tho had the lowest percentage of active Jack’d users in the last 30 days at 11.4% (95% CI 3.3-19.4) and Da Nang had the highest percentage at 42.9% (95% CI 28.3-57.6). The average weighted percentage of MSM ever using Jack’d in the 11 provinces was 56.7% (95% CI 47.8-65.5). In 1 province, Dong Nai, the 30-day RDS-I and Gile’s SS estimates did not produce statistically meaningful results, and the all-time Jack’d use estimator was used instead in this province.
The highest population size of MSM aged 18 to 49 years among the 11 provinces with reliable estimates was in Hanoi at 30,417 persons (95% CI 24,656-39,691), and the lowest MSM population size was estimated in Binh Dinh at 743 persons (95% CI 559-1108). The average weighted percentage of MSM among males aged 15 to 49 years in the 11 provinces was 0.96%, with a range of percentages from 0.70% to 2.47%. The complete results are presented in
Estimated number of men who have sex with men aged 18 to 49 years and weighted percentage of men who have sex with men among males aged 15 to 49 years in 11 provinces of Vietnam.
Province | Participants counted on Jack’d over 30 days, n | Proportion of MSMa using Jack’d (last 30 days, RDS-Ib) | 95% CI for proportion of MSM using Jack’d (last 30 days, RDS-I) | Estimated population size of MSM aged 18-49 years | 95% CI for estimated population size of MSM aged 18-49 years | Weighted percentage of MSM among males aged 15-49 years |
Bac Giang | 360 | 0.42 | 0.28-0.55 | 864 | 653-1274 | 0.23 |
Binh Dinh | 260 | 0.35 | 0.23-0.47 | 743 | 559-1108 | 0.21 |
Can Tho | 713 | 0.11 | 0.03-0.19 | 6276 | 3677-21,418 | 1.22 |
Da Nang | 1713 | 0.43 | 0.28-0.58 | 3990 | 2974-6059 | 1.70 |
Dak Lak | 279 | 0.15 | 0.07-0.23 | 1895 | 1226-4180 | 0.46 |
Dong Nai | 1191c | 0.15d | 0.02-0.28 | 7759 | 4216-68,370 | 1.10 |
Dong Thap | 303 | 0.14 | 0.06-0.21 | 2181 | 1420-4711 | 0.48 |
Hanoi | 12,848e | 0.42 | 0.32-0.52 | 30,417 | 24,656-39,691 | 1.81 |
Hai Phong | 1141 | 0.34 | 0.25-0.43 | 3336 | 2645-4515 | 0.73 |
Nam Dinh | 446 | 0.39 | 0.26-0.52 | 1131 | 850-1687 | 0.29 |
Thanh Hoa | 670e | 0.22 | 0.11-0.33 | 3017 | 2032-5846 | 0.40 |
Total | 19,924 | 0.37 | 0.27-0.48 | 61,609 | 44,909-158,860 | 0.96 |
aMSM: men who have sex with men.
bRDS: respondent-driven sampling.
cCount is for all-time use of Jack’d using the capture-recapture method.
dEstimate is for all-time use of Jack’d (RDS-I).
eCount obtained directly from social app service provider.
Extrapolation of the results from the 11 provinces with reliable estimates to the national MSM population size resulted in an estimate of 178,000 MSM (95% CI 122,000-512,000) in Vietnam. The percentage of MSM among adult males aged 15 to 49 years in Vietnam is 0.68% (95% CI 0.46-1.95).
Our estimates are the first comprehensive national estimation of the MSM population size conducted in Vietnam that use an empirical method. The point estimate of the 15- to 49-year-old MSM population in Vietnam produced in this study is 178,000, with an estimated range from 122,000 to 512,000. The corresponding estimated percentage of MSM among adult males aged 15 to 49 years of 0.68% is within the range of 0.09% to 4.06% suggested for the Asia and Pacific region by the Joint United Nations Programme on HIV/AIDS (UNAIDS) Spectrum guideline; however, the guideline does not provide a functional definition of MSM, which limits the comparability of the results [
The specific definition of MSM adopted in this study based on behavior and sexual preference has broader inclusion criteria than internet-based surveys, which only include men who have been sexually active with a man in the past year in their analysis [
The provincial estimates in this study were lower in Bac Giang, Binh Dinh, Dak Lak, Dong Thap, Nam Dinh, and Thanh Hoa; within the range in Da Nang, Dong Nai, Hanoi, and Hai Phong; and higher in Can Tho, in comparison with the expected range of MSM population sizes reported in past HIV and AIDS estimates and projection reports [
According to our estimates, during a given month, nearly 2 out of 3 people with a Jack’d account are actively connecting with other MSM on the social app, and there is a significant relationship between use of the social app and having sex with other men. These results are consistent with other recent studies that indicate increasing use of online social networks by MSM to find partners, while bypassing social stigma [
Overall, 3 different approaches were used to obtain the count of Jack’d users for the multiplier method in this study. Each of these approaches comes with its own advantages and disadvantages. The direct counting of users on social apps is a resource-intensive process. It requires investigators to manually record characteristics of active users on the social app daily or more frequently to not miss any peak periods when users log into the apps. In peak periods and in large cities with thousands of users, this process may not be feasible. Moreover, this approach also requires the deduplication of users in the records, which is also a resource-intensive step in the process. However, the advantage of this method is that the resulting number of active users during a brief period of 1 month reduces the recall bias when survey respondents in the second part of the multiplier method are asked about their use of the app in the past month.
The second approach to obtain the count of Jack’d users for the social app multiplier method in this study was obtaining the data directly from the app service provider. The advantage of this approach is that it requires little time, there is no need for deduplication, there is higher accuracy in the count than the manual count, and there is greater privacy as 1 integer figure is collected as the sum of all active users. However, these advantages are traded off with the cost of purchasing the aggregate, unduplicated, and nonidentifiable number of active users from the service provider. The third approach to obtain the count of users was a capture-recapture method on Jack’d. This approach required only 2 counts on 2 distinct days on Jack’d and did not require any deduplication. The disadvantage of this method is that it produces a count of all active users at any time on the social app, which in turn requires a less precise question in the RDS survey that is prone to recall bias. Future research should consider the reliability and precision of the data generated by these approaches as additional criteria to decide on the approach for use with the social app multiplier method.
The multiplier method requires the independence of the 2 data sources, the population in the 2 data sources to be defined the same way, and the 2 data sources to have aligned time periods and geographic areas [
Although the age profile of Jack’d users and RDS survey respondents in the 12 provinces included in this study was comparable, these age profiles are skewed toward a younger age range than Vietnam’s total male population pyramid [
Intraprovince migration of MSM risks some bias in the population size estimates as the RDS online survey excludes individuals who report having moved within the past 3 months even though there is no way to be sure if the Jack’d users had moved to the province in the past 3 months. We attempt to reduce the possibility of counting short-term migrants and visitors by only counting users who appear at least twice on the social app at 2 distinct days spaced over 1 month.
The estimation of MSM in Vietnam reported in this study is within the bounds suggested by UNAIDS for countries in the Asia and Pacific region, and the range produced in this study comfortably includes the estimated number of MSM in Vietnam arrived at through the national technical working group profiling and modeling process. The current estimation is based on an empirical method that relies on well-known and tested techniques along with innovative use of social apps used by the MSM population in Vietnam. This study highlights the feasibility of reaching a large percentage of MSM through a social app with programmatic and health promotion interventions. It is also the first time that population size estimations have been conducted in the provinces included in this study, and where reliable estimates were obtained, this study provides those provinces with valuable information on MSM population sizes that they can begin to work with in program planning and resource allocation. In other provinces where the population size estimates were extrapolated to but not directly observed, this study recommends that the extrapolated estimates be validated using locally appropriate, empirical size estimation methods, including the reliable methods and technologies that were introduced in this study. In provinces where there was a degree of homophily, bottlenecking, and sensitivity in the RDS survey results or where the estimators failed to produce reliable results, alternative methods should be attempted to assess and validate the MSM population sizes. Although the national estimation in this study gets closer to defining the potential range of the number of MSM in Vietnam, future studies will be needed to validate the range and further specify the estimated number. As the MSM population size is one of the key inputs to the national AIDS epidemic modeling and projection process, the AIDS epidemic model needs to be reviewed and updated with the new estimation.
Method of extrapolation from the study provinces and calculation of the national population size of men who have sex with men.
Convergence, bottlenecking, sensitivity, and homophily in the respondent-driven sampling survey.
Gile’s sequential sampling
men who have sex with men
respondent-driven sampling
Joint United Nations Programme on HIV/AIDS
The authors would like to extend a special thanks to Kirk Dombrowski for guidance and review of the methodology and insights provided during the training and analysis; Jinkou “Button” Zhou for review and advice on the outline of the manuscript; and Marie-Odile Emond for invaluable support, review, and advice throughout the size estimation study. This study was made possible by the generous financial support of The Global Fund to fight AIDS, Tuberculosis and Malaria and the technical support of UNAIDS. The findings and conclusions in this study are those of the authors alone and they do not necessarily represent the views, decisions, policies, or official position of the institutions or funding agencies with which they are affiliated.
VHS was the lead investigator, conceptualizing the research. AS reviewed the literature and drafted the manuscript aided by VHS and AAQ. VHS designed and had the overall responsibility for the quantitative analysis, with technical support from VML, LTCT, and AS. NTN, LTCT, VML, and VHS were responsible for the translation and localization of the survey. All authors contributed to the development of the argument, interpretation of the results, and revising the manuscript critically for important intellectual content. All authors read and approved the final paper.
None declared.