Of the 936 participants enrolled in RWISA at baseline, 698 (163 HIV-negative and 535 HIV-positive) subjects completed the baseline HTQ and at least one postbaseline HTQ (visits 2, 3, and 4) and were included in this analysis. The 238 subjects without postbaseline HTQ data were excluded. Participants included in the analysis were similar to those not included in terms of age, marital status, income, education, HIV status, experience of genocidal rape and traumatic events, and baseline HTQ and CES-D scores. However, those included in the analysis reported less depressive symptoms and, for the HIV infected, had higher CD4 cell counts at baseline than did those excluded.
shows the demographic and clinical characteristics of the participants by HIV status. The women were mostly poor and had only primary school education; almost half were widowed. Compared to HIV-infected women, uninfected women were significantly older and more likely to be widowed, to have a lower monthly income, and to have attended secondary school. Both HIV-infected and uninfected participants experienced a high number of traumatic events during the genocide, and 60% of the HIV-infected women and 68% of the uninfected women had HTQ scores at baseline consistent with PTSD. Depressive symptoms were significantly more prevalent among the HIV-infected than the HIV-negative participants at baseline and were reported in more than three quarters of the whole study cohort. Women with HIV were significantly more likely to have experienced postgenocidal violence (including sexual or physical abuse or threats) and to have used healthcare and trauma counselor services. At study entry, about one third of the women with HIV had CD4 cell count <200 cells/μL.
Baseline Characteristic of Rwanda Women's Interassociation Study and Assessment Participants with Posttraumatic Stress Disorder Measured for Four Visits
shows the changes in HTQ scores from the baseline visit to visits 2, 3, and 4. There was a continuing reduction in the mean HTQ scores at each follow-up visit, reflecting a sustained reduction in PTSD among both HIV-positive and HIV-negative women. Using the HTQ cutoff of 2.0, the prevalence of PTSD fell from 61% in the entire cohort at baseline to 24% (24.7% HIV-infected and 23.9% uninfected women) at visit 4 (data not shown). Although less striking, participants reported fewer depressive symptoms at visit 4 compared to baseline (77% reported CES-D ≥16 at baseline compared to 57% at visit 4).
Change in posttraumatic stress disorder (PTSD), by visit (with standard error).
There was also a reduction in the number of participants who reported seeking care from a trauma counselor over the study period. At the baseline visit, 25% (28.4% of the HIV-infected women and 15.1% of the uninfected women) reported seeing a trauma counselor. This was reduced at each of the next three visits, with only 6.8% (8.2% of HIV-infected women and 1.9% of uninfected women) (p<0.0001) seeking a trauma counselor in the 6 months before visit4
(data not shown).
To better understand which factors were associated with this reduction in PTSD, we analyzed the associations of patient characteristics at baseline with change from the baseline visit to visit 2, 3, and 4 in linear regression models using change as the outcome (). For example, the variable marital status shows change from study entry to visit 2 of −0.05 for widowed, which means that PTSD declined from baseline to visit 2 by 0.05 units more for those who are widowed than it did for other groups. For the variable of experienced postgenocidal violence, the coefficient of 0.07 from study entry to visit 2 means that for those who experienced genocidal violence, the decline in PTSD from baseline to visit 2 was 0.07 units less than for those who did not.
Univariate Linear Regression of Changes in Harvard Trauma Questionnaire Score from Baseline to Visits 2, 3, and 4 in Rwanda Women's Interassociation Study and Assessment Participants
For reduction in HTQ score between the baseline visit and visit 2 (where the largest reduction between visits was observed), only high baseline PTSD was associated with a significantly reduced HTQ score. This pattern of improvement in PTSD in those participants with the highest baseline HTQ scores continued from baseline to visits 3 and 4. In addition, the likelihood of a reduced HTQ score in visit 3 was significantly increased if the participant reported experiencing more traumatic events. In changes from baseline to visit 4, reporting a higher number of traumatic events and experiencing genocidal rape were significantly associated with reduced HTQ scores. Importantly, those participants with more depressive symptoms (CES-D ≥16) were less likely to show declines in their HTQ scores at each visit. Eighty percent of the cohort had PTSD symptoms decrease from visit 1 to visit 4. In general, the same factors that were associated with linear change in PTSD from visit 1 to visit 4 () were also associated with lack of improvement in PTSD from visit 1 to visit 4 (data not shown).
Compared to uninfected women, women with HIV had similar changes in HTQ scores in visits 2 and 3 but a significantly smaller reduction in HTQ scores in visit 4. These changes were not related to their CD4 cell counts. As women with HIV initiated indicated ART, we also investigated the effect of its use on reducing HTQ scores. By the fourth study visit, two thirds of the HIV-infected women reported taking ARTs. As shown in , however, there was no correlation between antiretroviral medication use and reduced HTQ scores.
We then used a multivariate model to evaluate predictors for the sustained reduction in HTQ scores from the baseline visit to visit 4 (). We included baseline HTQ score and CES-D ≥16, which had the most consistent associations with PTSD change in the univariate models, along with variables for which there is theoretical evidence for association. We also included HIV status, as we had hypothesized that PTSD could be part of the psychologic response to being infected with HIV. After adjusting for HIV status, CD4 <200 vs. >200 had no association in univariate models and was not included. We used the number of genocidal traumatic events in the model instead of reported experience of genocidal rape, as genocidal traumatic events and genocidal rape were collinear, the number of traumatic events had better associations in univariate models, and genocidal rape is included as a genocidal traumatic event. Previous studies have also found the number of traumatic events to be associated with persistent PTSD. Other variables in were not included because of both lack of statistical association in unadjusted models and a priori theoretical evidence for association. In the final model, baseline PTSD continued to be the strongest predictor of improved HTQ score. Reporting depressive symptoms (CES-D ≥16) was again a major factor preventing improved HTQ score.
Final Model for Change in Posttraumatic Stress Disorder from Visit 1 to Visit 4