During the summer of 2006, the first author spent several days each week as a volunteer with the HARC NEP. The description of the operation of the program is based on her impressions and discussions with HARC staff.
Between 2003 and 2006, HARC NEP staff interviewed needle exchange participants at their first visit and again at 6-months using an intake and follow-up questionnaire designed by HARC. Interviews were conducted in a private room in the HARC outreach van to maintain privacy and confidentiality. A structured survey form was employed to gather information from NEP clients.
The forms were altered slightly over time to increase the accuracy of the interview responses and to more precisely reflect the NEP goals. In total there were five questionnaire versions. Responses to items on each version of the questionnaire were coded to combine identical questions. Identical questions that appeared on some versions with a numerical frequency scale (i.e., none, 1-3 times, 4-6 times, 7-9 times, 10+ times) and on other versions with a subjective response scale (i.e., never, rarely, sometimes, often, always) were also combined.
Variables representing frequency of using a previously used needle, giving used syringes to another IDU, exchanging needles for another IDU, sharing other injection materials, cleaning skin with alcohol, reusing the same syringe, bleaching needles before reuse, condom usage, and HIV testing were dichotomized into "never" (included "zero" and "never" responses) or "ever" (included responses indicating they had previously been tested, previously received results, or had consented to being tested at the time of the interview) to reflect the potential for measurement differences based on the scales used in different versions of the questionnaires. A dichotomized item to represent the number of sexual partners reported was also created: either one partner or any number greater than one (those NEP users with no sexual partners were excluded from the analysis of this variable).
HARC used the Transtheoretical Model to measure progress toward changing drug-using behaviors [12
]. Individual respondent's stage was measured using a single question regardless of the questionnaire version employed. Some versions explicitly stated the five stages (Pre-contemplation, Contemplation, Preparation, Action, or Maintenance) and asked respondents to pick the stage that best described them. Other versions used a proxy measure and asked respondents to select from a list of statements the one that best described them in order to represent the client's stage (i.e., not at risk, need to decrease risk, ready to change, started doing things, or have been doing things). Responses to all forms of this question were combined into a single scale and coded as follows: 1 = Pre-contemplation or Not at risk, 2 = Contemplation or Need to decrease risk, 3 = Preparation or Ready to change, 4 = Action or Started doing things, and 5 = Maintenance or Have been doing things.
The initial data analysis plan consisted of paired comparisons between baseline and follow-up measurements of injection-related and sexual behavior-related HIV risk behaviors. Descriptive analyses revealed that only a small number of subjects completed both baseline and follow-up interviews which changed the analysis plan significantly. As a result, the focus of the analysis was shifted to a description of the program and the population it serves.
Aggregated baseline interviews were compared with aggregated follow-up interviews using the dichotomized variables to examine general trends over time. Although there was potentially a correlation between the observations of the 14 individuals with both baseline and follow-up interviews, and a generalized estimating equation model was considered, independence was assumed because the correlated observations were only a small percentage of the total number of observations. Logistic regression was used to calculate unadjusted odds ratios for the dichotomized variables using baseline measurement as the reference category, and simple linear regression was used to calculate coefficients, confidence intervals, and p-values for continuous and ordinal categorical variables. The logistic regression models initially included only the baseline/follow-up variable as a predictor. Race and sex, when added to the model, were not significant and not retained in further models. Odds ratios were calculated for the follow-up group compared to the baseline group. The small paired sample was analyzed using paired t-tests on non-dichotomized variables (except for whether the individual had ever had an HIV test, which was dichotomized into never or ever) to assess differences between the two groups and to generate hypotheses for further investigation. SPSS and R statistical packages were used for statistical analysis.
This study was approved by the Medical School Institutional Review Board (IRBMED - HUM00004528) and reapproved by the Health Sciences and Behavioral Sciences IRB at the University of Michigan after a change in university affiliation.