Two hundred thirty nine patients were approached to participate in the study and 200 agreed to enroll. The 39 patients who declined enrollment were similar in age, sex, race/ethnicity, and site of diagnosis to the 200 enrolled patients (data not shown). After enrollment, five patients were excluded because their HIV confirmatory test was negative. After medical record review, four were excluded because they were found to have been previously diagnosed, and two were excluded because they had already linked to care before the baseline survey was completed. One patient was excluded after enrollment because he moved from the Houston area and thus received medical care elsewhere. Over 80% of the remaining participants were recruited within one month of their diagnosis. Of the 188 eligible participants, 1 died before completing a baseline survey, 3 withdrew consent and 3 did not complete the trust scales in their entirety. After medical record review, 10 participants never had a CD4 cell count obtained and were excluded, leaving 171 participants in the final analysis.
Characteristics of the sample population are presented in ; 68 % were male; 53% were between the ages of 30 and 50 years; 51% were non-Hispanic Black (Black); and 45.0% had not completed high school. Of note, 61% of the study sample did not identify as either a man who has had sex with men (MSM) or as an injection drug user (IDU). Just below half the participants (44%) were diagnosed in an inpatient or emergency department setting. The majority of these participants (62%) were interviewed before discharge, and an additional 23% were interviewed in the field. The remaining 15% were interviewed in an outpatient medical setting. Half of the participants (50%) diagnosed in the outpatient setting were interviewed in the field, outside of any healthcare facilities, while the other half were interviewed in an outpatient medical setting (47%) or in the hospital (3%).
Characteristics of 171 participants in the Steps Study, overall and stratified by late (CD4 cell count <200 mm3) or early (CD4 cell count ≥200 mm3) diagnosis.
The mean baseline CD4 count was 278 cells/mm3, and 51% had CD4 counts under 200 cells/mm3 and so were considered diagnosed late. The sample characteristics are also shown in stratified by CD4 cell count (< or ≥ 200 cells/mm3). The late diagnosis group was more likely to be diagnosed in an emergency room or hospital. Other baseline characteristics were not statistically significantly different between the two groups.
Results for the trust in physicians and trust in the healthcare system scales are presented in . Collectively, the participants reported high trust in physicians and the healthcare system. The mean (SD) trust in physicians score was 42.2 (8.4) with a range of scores from 14 to 58. The scale exhibited good psychometric properties with a Cronbach’s alpha of 0.91. Participants with late and early diagnosis had similar scores (43.1 [7.3] and 41.3 [(9.4], respectively). The difference was not statistically significant (t=1.39, P= 0.17). The mean score for the trust in the healthcare system scale was 16.1 (3.3), with a range of 6 to 20. The scale demonstrated good psychometric properties with a Cronbach’s alpha of 0.87. Participants with a CD4 cell count <200 cells/mm3 had significantly greater trust in the healthcare system (16.7 [3.1]) than participants with higher CD4 cell counts (15.5 [3.4], t=2.32; P= 0.02).
Trust in physicians and trust in the healthcare system scores in the Steps Study, overall and comparing participants with late (CD4 cell count <200 mm3) and early (CD4 cell count ≥200 mm3) diagnosis.
Because there likely was colinearity between the trust variables, we created 2 separate multivariate logistic regression models of early diagnosis of HIV infection, one containing trust in physicians, and one containing trust in the healthcare system. We considered the variables in for inclusion in the models. In both models, the only significant predictor of early diagnosis was the presence of a HIV risk factor (MSM or IDU versus neither MSM nor IDU). In the model that included trust in the healthcare system, identifying as an MSM or IDU had an adjusted odds ratio of 2.3 (95% CI 1.08–5.02, Wald Χ2=4.68, P=0.03) for early diagnosis. Similarly, in the model that included trust in physicians, the adjusted odds ratio was 2.3 (95% CI 1.06–5.0, Wald Χ2=4.48, P=0.03). Neither trust in physicians nor trust in the healthcare system was an independent predictor of early diagnosis of HIV infection (Wald Χ2=1.42, P=0.24; Wald Χ2=0.00, P=0.99, respectively).
In a linear regression model of the trust in physicians score that included gender, race/ethnicity, age, education, MSM and IDU status, and CD4 cell count, race/ethnicity was the only variable to have significant effect (). Hispanic participants were the most trustful of physicians (adjusted mean 45.1), followed by Black (adjusted mean 41.0) and then non-Hispanic White (White) participants (adjusted mean 35.1). The differences between the race/ethnicity groups were significant (Black to White P=0.01; Hispanic to White P<0.0001; Black to Hispanic P=0.01). Similar results were found in the trust in the healthcare system multivariate analysis (). In a model adjusted for the same factors as above, race was the only variable to be statistically significantly associated with trust in the healthcare system. Again, Hispanic participants had the highest trust scores (adjusted mean 17.5), followed by Black (adjusted mean 15.4) and White participants (adjusted mean 14.8). The difference between Hispanic and White participants was significant (P=0.007), as was the difference between Black and Hispanic participants (P=0.002), while the scores of the Black and White participants were similar (P=0.5). In this adjusted multivariate model, there was no longer any difference in trust in the healthcare system by early or late diagnosis group (P=0.39). Logistic regression models of the trust variables dichotomized at the median value yielded similar results, with early or late diagnosis not predictive of trust.
Multivariate linear regression analysis of mean trust in the healthcare system and trust in physicians scores in the Steps Study.
We conducted a number of additional analyses to better understand our results. It is possible that dichotomizing early and late diagnosis at ≥ or < 200 cells/mm3 obscures differences in trust. We constructed scatter plots of the trust scores and CD4 cell counts at diagnosis. There was no evidence to suggest that different cut-points would have yielded different results, and the Pearson correlation coefficients for CD4 cell count and trust in physicians and trust in the healthcare system were −0.12 (P=0.11) and −0.19 (P=0.01), respectively, which suggested that there was little explanatory power for either scale. It is noteworthy that the correlations were negative, contrary to our hypothesis. We also used 4 clinically meaningful categories of CD4 cell count results in multivariate models of trust, and did not observe any apparent trends in trust as CD4 cell counts decreased (see ). It is also possible that the 10 participants excluded from the analyses because they did not have CD4 cell count results were affecting the findings. The mean trust in physician and trust in healthcare system scores for these 10 excluded participants were 40.5 (6.3) and 15.8 (2.6), not statistically different from the trust scores of the 171 participants with CD4 cell counts (t=0.64, P=0.53 and t=0.33, P=0.74, respectively). It may be that the relation of trust to HIV diagnosis is different by race, and to test this possibility we included interaction terms (race by trust) in the logistic regression analyses, but none of the interaction terms were statistically significant (data not shown). Finally, the counterintuitive finding that participants with a CD4 count <200 cells/mm3 had higher trust in the healthcare system may be associated with the fact that participants with lower CD4 counts were more likely to have been diagnosed in an emergency room or hospital. It may be that this recent, intensive interaction with healthcare systems and providers boosted trust. The mean trust in the healthcare system and trust in physician scores did not differ by site of diagnosis, lending little support this hypothesis, however (t=0.66, P=0.51 and t=1.38, P=0.17, respectively).