Sample and Comparison Group Characteristics at Baseline and Attrition Assessment
The recruitment and referral process resulted in a study sample of 523 who completed the baseline survey, including 176 PHA candidates and 347 of their contacts. Of the 176 PHA candidates, 112 (63.6%) completed the first five training sessions to become trained PHAs. They referred 223 CRs into the study. Additionally, 64 PHA candidates recruited into the study either never initiated the training program (n=55, 33.3%) or completed only 1–4 sessions (n=19, 10.8%). These untrained PHAs referred an additional 124 CRs into the study.
For analyses of baseline equivalence, we retained the distinction between four groups (trained PHAs, CRs of trained PHAs, untrained PHAs and the CRs of untrained PHAs). However, in conducting tests of behavioral, attitudinal, and other change between baseline and follow-up of the returned sample to assess RAP intervention outcomes, we generally used three comparison groups: trained PHAs, CRs of trained PHAs, and the rest of the sample, referred to as “Others.” This latter group included all others referred into the study as CRs and all those recruited as PHAs who never initiated the training, but it excluded ten PHA candidates who returned for the follow-up survey who attended any of training Sessions 1–4. Excluding these ten PHAs who dropped out before Session 5 allowed us to make a greater distinction between those PHAs who received both first level (staff-delivered) and possibly also second level (PHA-delivered) intervention, from those who received only second level intervention from the trained PHAs.
It is important to note that all participants in the project, including trained and untrained PHAs and all CRs, could potentially have been contacted by a trained PHA in their network between baseline and 6-month assessments, given the close geographic proximity of the neighborhoods in which Hartford drug users live and move and the known close network connections among drug users across the city (Weeks et al., 2002
). The diffusion aspect of the design indicated that this would indeed occur. We therefore did not use the group of “Other” participants as a non-intervention comparison group. Instead, we treated the members of this group as potential recipients of the second level PHA-delivered intervention.
Demographic and health characteristics at baseline of comparison samples are indicated in . We tested baseline differences among the four comparison groups on demographic characteristics, drug and sex risks, PHA efficacy attitudes, and risk reduction practices reported at baseline. Analysis of baseline equivalence on key characteristics indicated few significant differences among groups, with some important exceptions. Although we over recruited women to be PHAs at intake, both male and female PHAs tended to recruit male CRs, thereby generating a total CR sample that had fewer than expected women. Thus, there is a statistically significant difference in the percentage of women PHAs compared to CRs. However, the total sample reflected a gender ratio similar to that in previous Hartford studies, which indicated street-recruited drug users to be about 25% women (Singer & Weeks, 1992
; Weeks et al., 2001
; Weeks et al., 1996
). The sample also included four individuals who self-identified as transgender. There were virtually no differences between PHA recruits and CRs by ethnicity; however, significantly more African Americans than Hispanics or non-Hispanic Whites and other ethnic groups completed the PHA training.
Baseline Demographics and Health Characteristics of the RAP Sample (percentages except where indicated)a
PHAs and CRs did not differ significantly by socio-economic characteristics of educational attainment, employment status, income, and homeless status at baseline. All these indicators showed significant poverty and poor economic prospects in the study sample. In health and treatment histories, more trained PHAs reported having a history of STI and HIV, and more untrained PHAs reported having had hepatitis C. There were no significant differences among comparison groups in recent drug treatment history, including use of detoxification, in-patient, or outpatient programs.
Baseline differences in drug use in the prior thirty days between trained and untrained PHA groups were statistically significant (more PHAs who completed the training were crack users, more untrained PHAs were frequent injectors), but differences in their general sexual risk behaviors were not significant (). We also found differences in baseline PHA efficacy scale scores, with all PHA candidates (trained and untrained) averaging higher scores than CRs, and PHAs who completed the training entering the study with the highest baseline scores on their beliefs about the potential effectiveness of drug users to bring about risk and harm reduction in their communities and with their peers (p < .001).
Baseline Risk Characteristics and Attitude Scores of RAP Participants (percentages except where indicated)a
We were able to relocate and interview 367 (70.2%) of all RAP participants for the 6-month follow-up assessment. Our tracking records indicated that those lost to follow-up included 10 incarcerated, 11 who moved out of the area, 2 in residential drug treatment, and 2 who had died; the rest had unknown reasons for attrition. Those who returned included 134 (76%) of all original PHA recruits and 233 (67%) of all original CR referrals. However, differences in retention among the study subgroups was significant (p<.001), with 87.5% of trained PHAs (n=98) and 70.7% of their CRs (n=157) returning for the 6-month survey, compared to 56.3% of PHAs who completed less than 5 training sessions and 61.3% of the CRs of untrained PHAs. We compared baseline demographic and risk characteristics of the follow-up sample with the sample lost to follow-up to look for attrition bias. This included comparisons by sex, ethnicity, age, homeless and employment status, drug use (types and amount), and sex risk (number of partners and unprotected sex) in the prior 30 days. No significant differences were found between the retained and lost samples in sex, ethnicity and baseline risk characteristics. However, those lost to follow-up were more likely at baseline to have been homeless, unemployed, and younger.
PHA Exposure to Intervention
A quarter (25.6%) of the 176 PHAs recruited and interviewed at baseline never initiated the training program, either because they did not successfully refer two CRs (n=16) or they did not arrive at the scheduled training start date (n=29). Additionally, 19 PHA candidates (10.8%) who started the training program dropped out before completing the five sessions needed to become fully trained PHAs. In addressing the “intent to treat” question for the RAP PHA training curriculum, we found that those who did not initiate or who completed less than five sessions tended to be heavy injection drug users (), for whom an intensive training program such as RAP's PHA curriculum may have presented too great a burden. However, retention in the PHA training of those who initiated the program was very high; 86% (n=112) who started the program completed five sessions to become trained PHAs, and 51% (n=66) of those who started completed all 10 sessions over a three month period (Weeks et al., 2006
Changes in Behaviors and Attitudes Between Baseline and Follow-up
Overall Risk Reduction in the Total Sample: Hypothesis 1
We compared baseline and 6-month follow-up data of all participants who completed both surveys (n=367) to assess changes in reported risk behaviors between these time points. indicates significant reduction in the percentage of all returned RAP study participants who reported engaging in injection drug use and sharing syringes, crack use, and non-injection opiate, cocaine and amphetamine use, increased rubber tip use among crack users, and reduction in all sexual risks. The percent of injectors who shared equipment (cookers, cotton, rinse water) and drug solutions also decreased by 29% and 42%, respectively, though this was not statistically significantly. Bleach use is not included in because few RAP participants reported using it. When indicating having used a syringe that had previously been used by someone else, 13 out of 26 at baseline and 6 out of 11 at follow-up said they used bleach.
Percent of RAP Participants Reporting Risk Behaviors in Prior 30 Days at Baseline and 6-month Follow-up (participants who completed both measures, n=367)
In addition to a reduction in the percentage of the sample who reported engaging in risk behavior at follow-up, we also found a significant reduction in frequency of drug and sex risk incidents (), including injection rates and times used crack cocaine in the prior 30 days, as well as the number of unprotected sexual encounters overall and with non-primary (e.g., casual and/or paying) partners. We also found notable reductions in the rate of injection equipment sharing, non-injection heroin, cocaine and/or amphetamine use, and number of unprotected sexual encounters with an IDU.
Risk Behaviors in Prior 30 Days Reported at Baseline and 6-month Follow-up (Mean times reported by participants who completed both measures, n=367)
Risk Reduction among Comparison Groups: Hypothesis 2
We assessed risk reduction outcomes among the study comparison groups by looking at categorical change from baseline to follow-up in reported prior 30 day behaviors comparing trained PHAs, their CRs, and the “Others” (excluding the ten retained participants who attended PHA training Sessions 1 – 4, as noted above). gives percentages of participants who maintained no risk, reduced, or ceased risk. A high percentage of participants in each of the three comparison groups reduced or ceased each of the drug and sex risk practices or maintained low risk, though there is evidence that some participants initiated or increased risk during the test period. On nearly all measures, PHAs indicated the best outcomes, followed by their CRs, and then Others. However, statistical comparisons of these data did not support our hypothesis that PHAs, who received level one intervention (5–10 intensive staff-delivered sessions in the RAP PHA training) and possibly also level two intervention from another trained PHA, would show significantly better outcomes than study participants who were only potentially exposed to level two intervention (the PHA-delivered program). CRs and Others reported nearly the same risk reduction as each other, and both groups came close to the risk reduction or low-risk maintenance of the PHAs. Comparing PHAs with both of the other groups combined indicated that the only significant difference was in having ceased/reduced unprotected sex in exchange for money/drugs (p<.05). When Others were removed from these analyses, reducing number of sex partners also varied significantly (p<.05) in the comparison between PHAs and their CRs. This lack of significant difference across participant types held with analysis of risk incidents as well (data not shown).
Drug and Sex Related Risk Reduction or Low Risk Maintenance Among RAP Participants from Baseline to 6-month Follow-up, Prior 30 Day Reported Behavior (percentages who completed both surveys, n=367)
We conducted several types of analysis to explore reasons for this lack of difference in outcomes by comparison groups. These included: a) assessing RAP intervention exposure and RAP influence in all comparison groups, b) exploring evidence of diffusion of intervention effects from PHAs to CRs and to Others, and c) changes in the urban environment that may have affected the whole sample outside of RAP influence.
RAP Intervention Exposure and Influence
Several indicators of exposure to the level two RAP Peer-delivered Intervention are presented in . While some of these are direct indications of exposure to the peer intervention (e.g., recognition of the RAP Flip-book and slogans and receipt of reading materials, condoms and other prevention materials from “someone from the RAP project”), others are indirect (e.g., receiving prevention information, demonstration or materials from “any other drug user” and talking with other drug users about HIV and other health issues). As mentioned above, these latter were necessary measures because some study participants may not have been clearly aware of the RAP project nor able to identify the PHAs.
Indication of Exposure to RAP Peer Intervention at Baseline (B) and Follow-up (F) (percentages within each category unless otherwise indicated)a
These data revealed some important indicators of PHA delivery of the RAP peer interventions. For example, receiving reading materials, condoms and other prevention materials from someone in the RAP project increased significantly between baseline and follow-up for the total sample. Differences at baseline between PHAs and the other two comparison groups were significant in these measures (p<.05) and remained so at follow-up for reading materials and condoms. However, by follow-up the difference between PHAs and both of the other comparison groups in reporting RAP as a source of other prevention materials (i.e., bleach kits and crack health kits) was no longer significant, suggesting that these groups had caught up with PHAs in receiving prevention materials from someone in RAP.
Likewise, receipt of prevention interventions from “another active drug user” (most of whom were someone the participant knew) was significantly different among comparison groups at baseline, but by follow-up was no longer significant between PHAs and their CRs. Differences in this measure at follow-up between Others and the PHAs and CRs remained significant (p<.05), though the increase from baseline to follow-up within this group was also significant, suggesting that the group of Others had also been exposed to the RAP intervention. This was confirmed with the measure of recognition of the RAP Flip-book and the slogans, with the greatest recognition indicated by CRs, but also significant recognition among Others in the study sample. Reported having talked with drug users about HIV and other health issues likewise pointed to significant change in the behaviors of all drug users in the study, including CRs and Others, which was sustained even up to two weeks before the follow-up survey.
To assess the relationship between exposure to RAP interventions and the positive outcomes reported by all comparison groups in the study, we analyzed responses to a direct question asked at follow-up regarding changes participants had made in the prior 6 months “as a result of having talked to someone in the RAP project” (specified as a PHA or a drug user handing out materials from a backpack, not project staff). confirms the strongest reported risk reduction in most measures associated with talking to a PHA was among the PHAs themselves, followed by their CRs, and then Others. However, only reduction in drug use, adoption of rubber tips on crack pipes, reduction in sharing cookers, and reduction in number of sex partners as a result of talking to someone from RAP varied significantly by participant type, with PHAs reporting better outcomes on these items than either of the other two groups. Additionally, the two health promotion measures differed significantly among groups, with more PHAs reporting having talked to other drug users about HIV prevention and other health issues (i.e., the work of a “Peer Health Advocate”). This pattern points to interaction among the PHAs themselves, as well as potential PHA contact and influence over both the CRs and Others in the study.
6-Month Follow-up Reported Risk Reduction Behavior Change as a Result of Talking with Someone from RAP
Another indicator of RAP project influence was the PHA efficacy scale, which measured participants' belief in the ability of active drug users to influence their peers to reduce risk and to have a positive impact on their community. All groups' scores increased between baseline and follow-up from a mean of 2.73 to 2.78 (scale of 1 = low efficacy beliefs to 4 = high efficacy beliefs), though PHAs' mean score increased by 3.69% (from 2.85 to 2.95), compared to CRs', which increased 1.38% (from 2.68 to 2.72), and Others', which increased 0.48% (from 2.70 to 2.71). Repeated measures ANCOVA indicated significance in both the time effect (baseline to follow-up, p=.001) and group effect (differences among comparison groups, p=.000), though not in the interaction between time and group (each group's change pattern over time, p=.059). We also found a significant correlation (p<.05) between follow-up PHA efficacy scale score and reported reduction in prior 30 day unprotected sex, number of sex partners, and drug use, and reported increase in use of rubber tips and having talked to other drug users about HIV or other health issues as a result of having interacted with a PHA.
Indication of RAP Intervention Diffusion Through Networks
As a first step in assessing RAP intervention diffusion, we sought to locate PHAs in the personal networks of study participants. Through an intensive review of the name lists generated in the network component of the survey and confirmed through street outreach and ethnographic observation, we identified trained PHAs and their CRs who were named on the lists of other participants. We then compared the mean number of PHAs and CRs named by members of each of the study subsamples at follow-up (). As expected, CRs were the most likely to name one or more PHAs in their personal networks, and PHAs were the most likely to name one or more CRs. However, it is also notable that the group of Others also named both PHAs and CRs, suggesting significant mixing of participants who were directly exposed to RAP intervention, and potential for diffusion of the intervention materials and effects beyond the PHAs and their own recruited CRs.
RAP Intervention in Personal Networksa
We also compared the groups in terms of the percent of network members to whom participants gave information, materials or demonstrations, and the percent of their network members from whom they received these, including at their primary drug use site (). Differences among comparison groups were significant for giving prevention information and for giving materials/demonstration to network members (p<.001 for both), as well as for receiving materials/demonstrations from network members (p<.05), including at the participant's primary drug user site (p<.001). In all cases, PHAs gave prevention to the highest percentage of network members, followed by their CRs. CRs received prevention materials from the highest percentage of network members, followed by Others. However, at the primary drug use site, CRs, followed by PHAs themselves, received prevention from the highest percent of their network members. Notably, differences were not significant across comparison groups for receiving HIV prevention information from network members (p=.082), suggesting similarity among subgroups in the degree to which network members were sharing information with each other.
To confirm the activities of the Peer Health Advocates, we conducted analyses to assess the correlation between presence of a PHA in the participant's personal networks and receiving or giving out prevention materials and information. We found a correlation at follow-up between number of PHAs in the network and having received prevention information, materials or a demonstration from another active drug user in the last 6 months (n=365, Pearson r = .21, p=.001). When PHAs themselves were excluded from these analyses, the correlation remained significant (n=267, Pearson r = .22, p<.01). We also found a correlation between the number of PHAs in the network and having received prevention in the participant's primary drug-use site at follow-up (n=365, Pearson r = .12, p<.05). However, when PHAs were excluded from these analyses, the correlation was no longer significant (n=241, Pearson r = .12, p=.101).
Non-RAP HIV Prevention and Community Changes During the Intervention Study Period
Both PHAs and CRs also used other HIV prevention services in the city prior to the baseline RAP interview as well as during the period between the baseline and 6-month surveys. We initially built into this study measures of exposure to other community interventions to look for potential confounders in the assessment of RAP intervention outcomes. However, it appears that use of these services in itself was potentially influenced by RAP intervention exposure.
In the 2½ year period during which we conducted RAP baseline surveys, ethnographic and outreach observations of service availability documented that there were no new service programs initiated and no other major outreach interventions going on in the city concurrent with the RAP study. We also examined the potential correlation between time enrolled in the study and having “received HIV prevention information from any other programs, agencies or institutions,” “received any materials concerning HIV or AIDS,” “received any HIV prevention materials other than condoms,” and “received any condoms” in the last 6 months from local prevention or health care agencies and programs. Most of the Pearson coefficients were very small and not statistically significant except for negative correlations between time enrolled in the study and having received any condoms from the Hartford Needle Exchange Program (NEP), from an AIDS service organization, and from a social service program (Pearson r = −.119, −.116, −.094; p= .006, .009, .036, respectively). However, participant use of existing prevention services increased significantly from baseline to 6-month in the total sample (p<.05), and among CRs (p<.05). This suggests that participants, especially CRs, were making greater use of existing services in the period prior to their follow-up interviews than they had been before the baseline. This explanation was confirmed in a focus group with PHAs, in which they suggested that both they and their contacts were making decisions to take greater advantage of these services after being exposed to the health advocacy work of PHAs or the RAP training.