Study design and setting
We conducted this cross-sectional survey in Guangdong, China from March to Augustin 2018, which wasa sub-study of theGuangdong governmental sentinel surveillance programmes on STIs. The Guangdong governmental sentinel surveillance network was established in 2015 and comprised 10 cities (Zhuhai, Dongguan, Foshan, Jiangmen, Qingyuan, Shaoguan, Jieyang, Shantou, Maoming, Zhanjiang) with a high burden of STIs (see supplementary table S1).The 10 cities were located in the Pearl River Delta, East, West, and North of Guangdong, which can reflect economic and geographic diversity in Guangdong. Their locations are shown in supplementary figure S1. In the surveillance programme, 1-3 clinics at the prefecture-level city were selected according to the number of their outpatients. A minimum sample size of 200 to 250 wasrequired for each site.
According to the sentinel surveillance protocol16, participants were sampled by the sequential sampling method. In detail, allmale visitors in clinics were recruited into surveillance excluding men who visited a clinic for reproductive health, dermatology disease, and other research programmes (e.g. voluntarycounseling and testing). The eligible subjects need to be men at least 18 years old who had sex lives and were willing to test for HIV, syphilis, CT, and NG.In our study, we further excluded men who hadsex with meninprevious years, and subjects without STIs test results,usinga smartphone, andusing online sex-seeking. The flowchart of the study population was shown in supplementary files (Figure S2).
The sentinel surveillance programme annually observes the prevalence of STIs including HIV, syphilis, CT, and NG among STIs clinic visitors. Information from surveillance includessocial-demographic, sexual behaviors, previous STIs diagnosis, and the questions about HIV-related knowledge (see supplementary Table S3). Urine was self-collected by each eligible participant for CT and NG testing, and blood was collected by professional physiciansat the clinics to test HIV and syphilis. More information on testing kits and lab testing methods can be found in Supplementary methods (Table S4).
Local STIs clinic staff who assisted the survey were given intensive training. Interview settings had at least 1 private interview/counseling room and a waiting room. After providing verbal informed consent, those who met the eligibility criteria and agreed to participate in the study were referred to a separate,quiet room to complete a questionnaire.
The questions that referred toonline sex-seeking use and its related sex activities were as follows: duration of using online sex-seeking venues, the number of partners, condom use, and the period from initial online connection to an in-person meeting. They were also asked whether they inquired about condom use and HIV status with their sex partner before meeting, via the internet.
Participants were specifically asked whether they ever found sexual partners online. If participants reported any use of online venues to find sexual partners, they were classified as online sex-seeking users, while others were categorized as non-users. The online venues contained Weibo, Website, Blog, and Apps. Online partners only referred to female partners that our participants met online.
Data analyses were conducted in RX64 3.4.2 (R Foundation for Statistical Computing, Vienna, Austria). We used the chi-square test to compare the difference between categorical factors.Fisher’s exact test was used as an alternative to the Chi-square test when one or more of the cell counts in a cross-table is less than 5.
Univariate and multivariable logistic regression models were used to explore the factors associated with online sex-seeking use adjusting for age, residence, marital status, visiting clinic type, and the number of children. The potential factors associated with online sex-seeking and risky sex were included models as categorical variables. For the score of HIV-related knowledge (8 was the total score), we classified it into two categories (< 6, 6-8) and then included them in models. We also separately observed the effect of questions on right condom use and the number of sexual partners (see supplementary files Table S3).
Subsequently, we performed a sub-analysis to identify the characteristic of users with condomless sex and quick sex. Condomless sex was defined as the sexual behavior that online sex-seekers did not use acondom during the last sexual intercourse with their online partners. Quick sex meant that users met their online partners in person and engaged in sex within1 week from the first online meeting.