Setting
The study was conducted in a free STI testing centre in Fernand Widal Hospital in northern Paris. Care is entirely free of charge for all patients, even if they have no health insurance coverage. The tests performed include HIV, HBV, HCV, syphilis, gonorrhoea and chlamydia (urinary, vaginal, oropharyngeal and anorectal tests). HIV, hepatitis and syphilis are screened by serology. For gonorrhoea and chlamydia screening, self-swabbing is done by patients. Before the consultation, a self-administered questionnaire was proposed to all new patients visiting the centre and was completed in the waiting room. This questionnaire was included in the patient’s medical file and helped the doctor to direct the consultation according to the patient’s answers. However, only patients familiar with written French were able to complete the questionnaire. During pre-test counselling, the doctor evaluates each patient’s risk-taking behaviour and prescribes appropriate screening tests. Moreover, according to national recommendations for chlamydia/gonorrhoea, treatment is systematically started prior to results (presumptive treatment) when symptoms are strongly suggestive of chlamydia/gonorrhoea infections. Results are available approximately 1 week later. Patients can choose to be tested anonymously or not. Whether the patient chooses to remain anonymous for screening or not, he or she is given an anonymity number ensuring confidentiality throughout the process. The text messaging programme is proposed to patients by the reception agent who explains the programme with the help of a leaflet: 1) if all results are negative, the patient is informed by SMS and he/she does not need to return to the testing centre; 2) if at least one result is positive, the SMS invites the patient to come back to the centre to obtain the results. Up to three messages are sent if the patient fails to return. The SMS programme has been developed to respect medical confidentiality and is presented in detail elsewhere [16].
Study population
From October 2016 to May 2017, 2611 patients visited the free STI testing centre. During the 8 months of the study period, 2315 patients underwent STI testing, 219 came for post-exposure treatment and 77 came only for information with no testing. Among the 2315 patients who underwent STI testing, 266 (11%) patients had at least one positive test. Of these 266 positive patients, 214 patients filled in the self-administered questionnaire (participation rate 80%) and were included in this cross-sectional study.
Data sources
Data were collected from two complementary sources: the consultation database and the self-administered questionnaire.
The following data were routinely entered in the consultation database: date of first visit, anonymous testing, participation in the text messaging programme, STIs tested, provision of presumptive treatment for chlamydia or gonorrhoea or not, date of return, and results of the screening tests performed.
The questionnaire included sociodemographic factors: age, gender (woman, man, transgender), place of residence, place of birth, educational level and health insurance coverage, sexual behaviour and history (including sexual orientation and occasional and/or regular sex partners), previous HIV testing and factors related to the visit to the centre (who suggested testing, accompaniment by another person at the consultation, reason for attending indicated by one or several items that can be ticked among a list of 14 reasons), self-perceived risk of HIV infection (the patient perceived himself/herself at a higher risk of infection compared with the general population, equal risk, lower or no risk).
Outcome
In the literature, the time period used to define FTR varied between 1 month and 1 year [4, 9, 12, 14]. However, all studies showed that most patients returned for results between 7 days and 21 days after testing [6, 8, 10, 18, 19]. Thus, FTR at 30 days is a relevant indicator and was used in this study.
Factors analysed
Only 10 patients were aged over 25 years and had no previous HIV testing. Age and previous HIV testing were therefore combined in one variable with three categories: patients aged 25 years or older, patients under 25 years with no previous HIV testing and patients under 25 years with previous HIV testing.
In our study population, there were no transgender persons and only 7 patients were women who had sex with women. Gender and sexual orientation were therefore combined in a single variable with three categories: men who had sex with men (including bisexual men), heterosexual men, and women.
To explore both the potential role of self-swabbing and of presumptive treatment in FTR (patients may have thought they did not need to return because they had already been treated), data on testing for chlamydia/gonorrhoea infections and provision of presumptive treatment at the pre-test visit were combined in a single variable: not tested for chlamydia and gonorrhoea, tested without provision of presumptive treatment, tested with provision of presumptive treatment.
Consultation for risk-taking was based on the reason for attending given by patients in the questionnaire. We considered that the consultation was for risk-taking if the patient ticked at least one item among the 11 reasons corresponding to a risk (unprotected sexual intercourse, clinical signs of STI, HIV-positive sex partner, sex with sex worker …). The consultation was not for risk-taking if the patient reported that he/she wanted only information or reassurance, or if the patient wanted to be tested before stopping condom use with his/her regular partner.
Self-perceived risk of HIV infection was summarised into a binary variable: self-perceived risk higher or equal to the general population vs lower or no risk.
Statistical methods
Patient characteristics were compared according to FTR30 using the χ2 test or Fisher’s test. Univariate and multivariate analyses were conducted to assess factors associated with FTR30 using logistic regression models. Considering the limited number of events in our study, we adopted a parsimonious methodology using a stepwise backward selection with a threshold of 0.2 for removal from the model. Association was considered as statistically significant when p < 0.05. Statistical analyses were performed using STATA/SE 11.0 (Stata Corporation, College Station, TX, USA).