Baseline characteristics of the sample are shown in for the full sample (n=54) and for the final sample (n=42). There were no significant differences between conditions at baseline for the full sample or the final sample on demographic, HIV risk, VL or CD4 counts, drug use, psychiatric, adherence, percentage of days of cocaine use, or ASI Drug Composite Scores. Fifty-four patients on HAART participated in the study, including 28 women, 25 men, and 1 transgendered individual; 82% were African-American, and 1 participant was Hispanic. Their mean age was 45 (SD=6.4); the modal age was 41. Thirty-seven percent of the sample had less than a high school education, but 31.5% had high school and 31.5% had some college education. Most (82%) were not employed. Most (44; 82%) had a history of involvement with the criminal justice system. More than half of this sample was heterosexual (32, 59.3%). On average, participants had been diagnosed with HIV 11.3 years ago (SD=6.5); the most recent diagnosis was four months before, and the most distant diagnosis was 22 years before. On average, participants took 58.27% (SD =27.5) of their prescribed HAART pills in the two weeks prior to enrollment. At baseline, there was a significant inverse relationship between adherence and viral load (r = −.34, p<.05). The participants had detectable HIV VL (mean log VL 2.97, SD = 1.18) and compromised immune functioning as indicated by low CD4 cells (mean 433, SD = 317). While not shown in the table, the full sample reported significant health problems beyond HIV; 31.5% had a current diagnosis of AIDS, while other common current co-morbidities included Sexually Transmitted Infections (93%), Hepatitis A (5.6%), B (18.5%), or C (35.2%), hypertension (42.6%) and Type 2 Diabetes Mellitus (7.4%). Participants were prescribed 6.4 medications per day, with 2.5 for HIV disease. Despite high unemployment, 94.4% reported stable housing.
also shows the drug use and positive psychiatric screens for the sample by assignment condition. On average, participants used crack cocaine on a third of days in the past month prior to enrollment in the study. On urine toxicology screens at randomization, half screened positive for cocaine, consistent with their self-report of using crack cocaine on a third of days. Some also screened positive for marijuana, benzodiazepines, opioids, barbiturates, and methamphetamine. Most were classified as crack cocaine dependent on the M.I.N.I., and 38.5% were also classified with alcohol use disorders. Psychiatric problems were common, with nearly half experiencing current Major Depression. Current suicidal and homicidal ideations were also present in the sample; whenever these were endorsed, study therapists and supervisors reviewed responses while the participant was there, and assessed current risk using a standardized protocol.
3.2 Treatment and Follow-Up Retention
Retention rates did not vary by group at any of the assessment or treatment sessions. The 2 mistaken enrollments had both been randomized to the MI+ condition and had completed the study, but were dropped from analyses. Forty of 54 enrolled participants (74.1%%) received all 6 treatment sessions, 45 of 54 (83.3%%) completed FU1, and 42 of 54 (77.8%) completed FU2 (see for the study flow chart and completion by condition).
3.3. Primary Outcomes
shows the mean differences for the primary dependent variables (mean adherence and ASI Drug Composite Scores) and secondary dependent variables (days using cocaine and log VL) across time. There were main effects of both interventions on HAART adherence and ASI Drug Composite Scores. shows the ANOVA results for the sample of 54 at baseline, 45 at FU1, and 42 at FU2. In some instances, missing data resulted in changes to the n’s when variables for that analysis were missing as noted below.
Mean differences between conditions across time for adherence, ASI Drug Composite Scores, days using cocaine, and logVL
Analysis of Variance Results for Condition and Time Effects on Dependent Variables
There was a moderately negatively skewed distribution of adherence at baseline (skewness = −.837), but this deviation from the normal distribution was within the acceptable range to proceed with the repeated measures ANOVA. There was a significant main effect of time on adherence. In the MI+ condition, adherence improved from 60.2% (SD =26.7%; n =26) at baseline to 93.3% (SD =10%; n =22) at FU1 and 93.9% (SD =13.7%; n =19) at FU2. In the Video+ condition, adherence improved from 56.4% (SD =28.5%; n =28) at baseline to 87% (SD =25.3%; n =23) at FU1 and 86% (SD =29.1%; n =23) at FU2. There was no significant between-groups effect on adherence. Between groups effect sizes for adherence were small and nonsignificant at FU1 [d =.33 (−.26, .91); p =.28] and FU2 [d =.34 (−.28, .95); p =.28]. When collapsing across conditions, participants improved their adherence from 58.27% (SD =27.5%) at baseline (n =54) to 90% (SD =19.5%) at FU1 (n =45) and 90% (SD =24%) at FU2 (n =42). Follow-up contrasts showed that mean adherence was lower at baseline than at FU1, F (1, 37) =32.11, p<.0001 and FU2, F (1, 37) =43.81, p<.0001; repeated measures effect sizes are shown in .
There was also a significant main effect on ASI Drug Composite Scores, which range from 0–1, with lower scores indicating fewer problems from drug use. In the MI+ condition, ASI Drug Composite Scores decreased from .18 (SD =.11) at baseline (n =26) to .10 (SD =.08) at FU1 (n =20) and rose slightly to .11 (SD =.09) at FU2 (n =18). In the Video+ condition, ASI Drug Composite Scores decreased from .16 (SD =.09) at baseline (n =28) to .09 (SD =.09) at FU1 (n =21) and .08 (SD =.08) at FU2 (n =22). Mean composite scores were higher at baseline than at FU1, F (1, 35) =17.27, p<.001, and FU2, F (1,35) =14.32, p<.001. There was no significant between-groups effect on ASI Drug Composite Scores. Between groups effect sizes for composite scores were small, nonsignificant, and in different directions at FU1 (d =.12 (−.72, .48); p =.72) and FU2 (d =−.12 (−.72, .48); p =.70). Across conditions, participants decreased their ASI Drug Composite Scores from .17 (SD =.10) at baseline (n =54) to .09 (SD =.09) at FU1 (n =41) and .09 (SD =.09) at FU2 (n =40).
3.4. Secondary Outcomes
There was a significant main effect of the interventions on percent of days using cocaine, but not on log VL or the proportion of participants with detectable or undetectable VL. In the MI+ condition, percent days using cocaine decreased from 35.4% (SD =31%) at baseline (n =26) to 15.4% (SD =15.1%) at FU1 (n =20) and 10.9% (SD =14.7%) at FU2 (n =19). In the Video+ condition, percent days using cocaine decreased from 30.4% (SD 29%) at baseline (n =28) to 13.8% (SD =19.7%) at FU1 (n =23) and 12.1% (SD =20.4%) at FU2 (n =23). Percent days using cocaine was higher at baseline than at FU1, F (1, 37) =22.32, p<.0001, and FU2, F (1, 37) =26.49, p<.0001. No significant between-groups effects were identified for any of these secondary outcomes. Across conditions, participants decreased the percent days they used cocaine from 32.8% (SD =30%) at baseline (n =54) to 14.6% (SD =17%) at FU1 (n =45) and 11.6% (SD =18%) at FU2 (n =42).
There was no significant main effect of time on log VL and no significant between-groups effect. Thirty-five percent of 26 MI+ participants and 41% of 27 Video+ participants had detectable VL at baseline, (χ2(1 df) =.21, p=.65), while 26% of the 19 MI+ and 43% of the 23 Video+ had detectable VL at FU2, (χ2(2 df) =.2.32, p=.31). Similarly, the proportion of people in each condition with decreased, unchanged, or increased log VL did not differ. Only 1 MI+ participant and 3 Video+ participants had significantly decreased log VL, while 14 MI+ and 14 Video+ participants had unchanged log VL, and 4 MI+ and 3 Video+ participants had increased log VL (χ2(2 df) =1.12, p =.57).
Both recruitment and retention were feasibility challenges. It required over 2 years to enroll 56 participants, 2 of whom were mistakenly enrolled despite reporting 100% adherence at baseline. Screening 201 potential participants yielded 76 eligible participants, a 38% gross eligibility rate and a net rate of participants to screens of 28%. The accrual rate was 2 new enrolled participants per month with approximately 8 screened per month. The primary reason for ineligibility was that potential enrollees reported over 85% HAART adherence, and in many cases, reported 100% adherence.
An additional challenge with this crack cocaine-using population was that it took considerable effort to retain participants in the study. Several participants were incarcerated long enough that they timed out of the study and missed either intervention or follow-up visits. Given both the drug dependence and psychiatric disorders that were present in the sample, we often engaged in crisis management (including two emergency hospitalizations for homicidality) and referrals to psychiatric care. Retention was quite good despite these challenges.
3.5.1. Treatment fidelity
As discussed previously, all sessions were reviewed weekly by therapists and supervisors to ensure that each condition was delivered as planned, and that drift across conditions was prevented. In addition, post-session Therapist Checklists indicated that the activities conducted within the 2 conditions differed as planned. Specifically, all of the activities occurred in the MI+ condition, varying as planned by session. In contrast, only role induction, education about HIV adherence, education about cocaine, showing videos, using stem questions to guide debriefing, and providing tip sheets and reading materials occurred in the Video+ condition, consistent with the protocol for that condition. In the Video+ condition, therapists showed no use of reflective listening or other MI skills or strategies; rather, they asked scripted questions as written. All Therapist Checklist items differed significantly by condition (data not shown for χ2 analysis of each activity by each session by condition; all p values <.05).
3.5.2. Treatment credibility
Pre-treatment, both MI+ and Video+ participants rated their assigned interventions favorably. MI+ participants scored a mean of 35.5 (SD = 5.4) out of 40, while Video+ participants had a mean score of 33.3 (SD = 7.5). Post-treatment, both groups rated their condition higher, with a mean for MI+ of 36.8 (SD = 4.3) and a mean for Video+ of 35.6 (SD = 4.8). These scores indicate that participants rated treatments as equally credible before and after they received them.
3.5.3. Treatment satisfaction
Both MI+ and Video+ participants reported total satisfaction scores indicating excellent satisfaction. MI+ participants’ satisfaction with their condition was a mean of 8.1 (SD = 3.0; 5 is most satisfied; 30 is least satisfied) at FU1, and was a mean of 8.3 (SD = 2.5) at FU2. Video+ participants’ satisfaction with their condition was a mean of 9.3 (SD = 2.7) at FU1 and 9 (SD =1.3) at FU2; none of these means were significantly different. These scores indicate that participants were satisfied with their assigned condition immediately and three months after completing treatment, and that the satisfaction rates did not differ by condition.