High-risk behaviors and their association with awareness of HIV status among participants of a large-scale prevention intervention in Athens, Greece

Aristotle was conceptualized, designed, and implemented by a research group in the Department of Hygiene, Epidemiology, and Medical Statistics at the Medical School of the National and Kapodistrian University of Athens, and ran between mid-2012 and end of 2013. The National Strategic Reference Framework 2007–2013, the European Social Fund, national resources, and the Hellenic Scientific Society for the Study of AIDS and Sexually Transmitted Diseases funded and supported this intervention. Aristotle focused on PWID and aimed to offer HIV testing, to increase the rate of HIV diagnosis, and to improve linkage to care.

In terms of design, Aristotle [11‐13, 26] consisted of five successive enrollment rounds (A, B, C, D, E) using respondent-driven sampling. Each enrollment round lasted 10–12 weeks with between-rounds breaks of 1 to 4 weeks. Selected PWID acted as seeds to begin the chain referral process in each round (5–11 per round). Seeds received coupons and were asked to deliver them to three PWID from their social environments. The coupons had unique numbers linking each recruiter with his/her potential recruits.

PWID who visited the Aristotle site were eligible to participate in an RDS round if they: (i) held a valid RDS coupon (all but seeds); (ii) had been involved in drug injection without prescription in the past 12 months; (iii) were aged ≥18 years; (iv) were residents of Athens Metropolitan Area; and (v) came to the Aristotle site for the first time in the current testing round with the intention to participate (PWID could participate in multiple enrollment rounds but only once in each round). All participants gave written informed consent.

Experienced Aristotle staff interviewed the participants (computer-assisted personal interviews) based on the National HIV Behavioral Surveillance System (NHBS) questionnaire for PWID, appropriately adjusted to the Greek setting and with additional items that were of interest [27]. The questionnaire included items on demographics, injecting network size, accommodation status, history of incarceration, history of previous HIV testing, HIV status, sexual practices, and drug use history and practices.

Participants gave blood that was tested for HIV by a microparticle EIA anti-HIV-1/2 assay (AxSYM HIV-1/2 gO; Abbott Laboratories, Abbott Park, IL). Western blot (MP Diagnostics, Singapore) was used to confirm reactive results.

After the interview and the blood collection, participants received their primary monetary incentive, sterile syringes and injection paraphernalia, condoms, and leaflets. The monetary incentive included 5 Euros for participation in Aristotle (interview and blood collection); 3 Euros when the participant came back to receive his/her HIV-test results; and up to 9 Euros in total when his/her recruits were enrolled in the program. Those who tested HIV positive were referred to infectious disease clinics and to Opioid Maintenance Treatment (OMT) programs of the Greek Organization Against Drugs. Needle and syringes were also distributed by the Greek Organization Against Drugs and several non-governmental organizations [12, 13]. Participants also received three coupons to distribute them to other PWID. This recruitment chain resulted in the enrollment of approximately 1400 PWID in each round.

Based on responses to relevant questionnaire items, high-risk drug injection-related behaviors were defined as follows: i) Injecting drugs at least once per day in the past 12 months; ii) Reporting receptive syringe sharing about half of the time, most of the time or all of the time the participant injected drugs in the past 12 months; iii) Reporting receptive syringe sharing the last time the participant injected drugs; iv) Practicing drug division with a syringe that someone else had already injected about half of the time, most of the time or all of the time the participant injected drugs in the past 12 months.

High-risk sexual behavior was operationalized as condomless sexual encounters (vaginal or anal sex) in the last year (participants answered “usually no” or “never” to the question about condom use). Males were asked whether they used condom with a female or male partner. Women PWID were asked about whether their male partner used condom when they had sex.

The exact wording of questions and responses for drug injection-related and sexual behaviors is given in the Additional file 1.

Participants were classified as aware of their HIV infection (AHS) if their self-reported HIV-positive status at the interview was in agreement with the positive result of their blood test. HIV-positive unaware (UHS) were those who did not know about their infection when they tested HIV positive. HIV-negative participants were those whose blood test was negative for HIV. Participants could be reclassified over the course of Aristotle if they had become aware of their HIV infection or had acquired HIV.

Differences in demographic parameters and high-risk behaviors among the different groups were assessed using chi-squared tests and Kruskal–Wallis tests. A non-parametric test for trend was used to assess changes over time.

There were some parameters that dictated the selection of a certain advanced statistical method. First, some PWID participated in more than one enrollment rounds, which resulted in clustered data with multiple measurements on some of the participants at different times. Observations from the same participants may be correlated, while those from different subjects are assumed to be independent. Failure to account for correlation of within-subject measurements can result in invalid standard errors and erroneous findings. Second, drug injection-related and sexual behaviors were binary outcomes (high-risk versus less risky behavior or no risk) in these analyses measured repeatedly for some Aristotle participants. Awareness of HIV serostatus was also a time-dependent covariate. Multiple conventional logistic regression analyses of outcomes and covariates at each time-point (round) could inflate type I error rate. Third, high-risk behaviors might not change regularly over time, i.e., the probability of change may differ between segments of Aristotle rounds. In this case, a uniform regression function cannot be applied to the data. In order to accommodate the above-mentioned issues, piecewise linear generalized estimating equation models (GEE with logit link function) were selected to regress binary repeated outcomes (separate for each outcome) against covariates allowing for different slopes of change [24, 25, 28‐30]. GEE are an extension of generalized linear models for longitudinal data and belong to the family of marginal models that take into account the averaged relationship in the population.

The choice of change point (or breakpoint) for the piecewise models was based on the graphical examination of the data and modelling procedures. Segmented GEE models were obtained by a two-piece definition of the regression function over two intervals of the domain of covariates (one segment from round A to round B and one segment from round B to E).

Given the potential correlation of measurements on the same subject, a working correlation (a hypothesized relationship between repeated observations on a participant) has to be defined a priori for the estimation of model parameters. These analyses assumed an unstructured correlation within each Aristotle participant.

All analyses were performed in Stata 11.1 [31].

Overall, 3320 PWID were enrolled in Aristotle across all 5 rounds. Of these, 53.8% participated only in one round, 20.5% in 2 rounds, 14.1% in 3 rounds, 11.3% in 4 rounds, and 7.9% in all rounds. Therefore, the total number of interviews was 7110. Sociodemographic characteristics and risk behaviors at first participation in Aristotle by gender are shown in Table 1. Participants were mostly males (2807, 84.5%) and from Greece; one third were homeless, currently or in the past 12 months; 64.1% said that they lacked health insurance; and almost half had history of imprisonment. The median age was 34.2 years.

The median duration of injection drug use was 12 years and the primary substance of use was heroin. More than one third of the participants reported at least daily injection of drugs with a median of 3 injections in a usual day; 20.6% reported receptive syringe sharing when they last injected; 8.0% reported receptive syringe sharing at least half the times they injected drugs in the past 12 months; and 37.2% said they divided drugs at least half of the times they injected drugs in the past 12 months with a syringe that someone else had used before.

More than half of the male participants (56.7%) reported that they consistently (always or usually yes) used condoms while 37.6% of female participants said so when they were asked about use of condoms by their male partners (Table 1).

At first participation in Aristotle, of 3320 PWID, 506 (15.2%) were positive for HIV. Of the HIV-positives, 127 (25%) were aware of their infection. Awareness of HIV infection at first participation was associated with female gender, Greek nationality, history of participation in OMT programs, and less frequent drug injection (less than once per day) in the past 12 months (Table 2). Moreover, HIV positive females who knew about their infection at their first visit to Aristotle were more likely to report that their male partners used condoms than HIV positive females who were not aware of their HIV status (p = 0.022). In males, condom use was not associated with awareness of HIV infection at their first visit to Aristotle (Table 2).

Figure 1a shows changes in high-risk injecting behaviors across Aristotle rounds. The reported prevalence of high-risk drug injection-related behaviors including injecting at least once per day in the last 12 months, dividing drugs at least half of the times the participant injected drugs in the past 12 months with a syringe someone else had used before, and receptive syringe sharing at least half of the times the participant injected drugs in the past 12 months, declined significantly (p for trend < 0.001). There was no change, however, in the proportion of PWID who reported that they had shared a syringe the last time they injected.

Males reported use of condoms more frequently than females throughout the course of Aristotle. The reported prevalence among males remained, more or less, the same in all rounds. In females, there was a small increase from round A to B but condom use gradually declined from the second to the last round. There was not however any statistically significant change in condom use among both males and females across the Aristotle rounds (Fig. 1b).

More than half of all HIV-positives (280/506; 55.3%) in Aristotle were found in the first enrollment round (Fig. 2). The prevalence of HIV among first-time participants decreased across rounds (19.8% in round A to 10.6% in the last round).

The proportion of HIV-infected participants who knew about their infection at their first visit in Aristotle increased from 21.8% (61/280) in round A, to 31% (31/100) in round B and to 36.4% (12/33) in the last round of the program (Fig. 2).

The prevalence of drug injection-related behaviors among AHS, including injecting drug use at least once per day in the past 12 months, receptive syringe sharing at least half of the times the participant injected drugs in the past 12 months, and dividing drugs at least half of the times the participant injected drugs in the past 12 months with a syringe someone else had used before, decreased from round A to round Ε (Fig. 3a, b, c, d).

Multivariable analyses using GEE models were adjusted for gender, age, origin, homelessness, education, history of imprisonment, and history of OMT. The odds of injecting drugs at least once per day in the past 12 months among AHS decreased by 66% [Odds Ratio (OR); 95% Confidence Interval (CI): 0.34 (0.19, 0.61)] from round A to B (Table 3). Similarly, among UHS, the odds of injecting drugs at least once per day in the past 12 months were 63% less in round B than in round A [OR (95% CI): 0.37 (0.25, 0.57]. Between rounds B and E, the odds of injecting drugs at least once per day in the past 12 months were slightly reduced by 16% [OR (95% CI): 0.84 (0.74, 0.96)] and by 6% [OR (95% CI): 0.94 (0.76, 1.18)] among AHS and UHS, respectively.

Comparing round A to B, the odds of receptive syringe sharing at least half of the times the participant injected drugs in the past 12 months decreased by 47% in the round B among AHS [OR (95% CI): 0.53 (0.31, 0.91)] and by 33% among UHS [OR (95% CI): 0.67 (0.45, 0.99)]. Between rounds B and E, the odds of receptive syringe sharing at least half of the times the participant injected drugs in the past 12 months were reduced by 19% [OR (95% CI): 0.81 (0.71, 0.93)] among AHS and by 4% [OR (95% CI): 0.96 (0.77, 1.20)] among UHS.

HIV negatives experienced smaller decreases in the odds of high-risk drug injection-related behaviors from round A to B than AHS or UHS.

Reported condom use did not significantly change in both males and females between rounds A and B. From round B to E, condom use among AHS males increased significantly by 24% [OR (95% CI): 1.24 (1.01, 1.52)]. A non-significant increase in condom use (21%) was observed among UHS males [OR (95% CI): 1.21 (0.88, 1.67)]. However, this difference in changes of condom use between AHS and UHS did not reach statistical significance. Interestingly, between rounds B and E, there was a slight 6% decrease in condom use among male HIV-negative participants [OR (95% CI): 0.94 (0.89, 0.99)]. There were no changes in condom use from round B to E among AHS [OR (95% CI): 0.86 (0.64, 1.17)], UHS [OR (95% CI): 0.88 (0.50, 1.55)], and HIV-negatives [OR (95% CI): 0.96 (0.85, 1.08)] in the female population of Aristotle.