Understanding HIV Care Provider Attitudes Regarding Intentions to Prescribe PrEP

Amanda D. Castel, MD, MPH; Daniel J. Feaster, PhD; Wenze Tang, MPH; Sarah Willis, MPH; Heather Jordan, MPH; Kira Villamizar, MPH; Michael Kharfen, BA; Michael A. Kolber, MD, PhD; Allan Rodriguez, MD; Lisa R. Metsch, PhD


J Acquir Immune Defic Syndr. 2015;70(5):520-528. 

In This Article


Participant Characteristics

There were slightly more respondents from Miami than from Washington (Table 1). There were slightly more male respondents (59%), with modal age category of 40–49. Most participants were non-Hispanic white (49%), followed by Hispanic (25%), and black (13%). Half of the respondents self-identified as infectious disease specialists, and more than 25% as primary care physicians (18% internal medicine, 11% family medicine) who provided some HIV care. Nearly half (47%) of providers had been practicing for more than 20 years, 82% had seen more than 200 HIV-positive patients in their practices, and 73% had seen more than 20 HIV-positive patients in the previous 3 months. More than half of providers (53%) agreed that PrEP was effective or most effective and 24 (17%) had prescribed PrEP before completing the survey. Those providers who had previously prescribed PrEP were more likely to come from practices with a written PrEP protocol, had more patients ask for PrEP, and had lower scores on the "lack of PrEP knowledge scale" (data not shown).

Latent Class Analysis

The LCA identified 2 distinct classes of providers. The comparisons of LCA model fit (Table 2) show that no solution was favored by all 3 information criteria, but the 2-class solution was favored by 2 of the 3 criteria (AIC, ABIC) over the single class solution. The 3-class solution had lower AIC and ABIC than did the 2-class solution. However, whereas the 2-class solution was replicated in 41 of 50 solutions of the random-start process, the 3-class solution was only replicated in 6 of 1000 random starts. Furthermore, when moving from 1000 to 1500 and then 3000 initial random starts, a new maximum was found each time. As this is indicative of local maxima, we focused on the 2-class solution. The entropy of the 2-class solution was good (0.904); the average probability of being in class 1 for those classified in class 1 was 0.968 and that of being in class 2 for those classified in class 2 was 0.978. There were no differences in the proportions from each site across the 2 classes (χ2(1) = 0.003, P = 0.958).

Comparison of the 2 Provider Classes

Table 3 shows the probability and 95% confidence intervals of either agreeing or strongly agreeing with the statements by the 2 classes of respondents. This information is presented as a response profile in Figure 1. Class 1, the larger class (95 respondents), tended to agree less with statements that oral PrEP and microbicides can decrease the risk of HIV acquisition than did class 2 (47 respondents). A significantly higher proportion of class 2 vs. 1 agreed that PrEP was feasible in their clinics and that they had adequate time to prescribe PrEP. A higher proportion of class 2 vs. 1 respondents also agreed that they would prescribe PrEP to serodiscordant couples and that it might empower women unable to negotiate condom use. With respect to perceived barriers, class 2 also had a slightly higher probability of agreeing that cost might pose a significant barrier.

Figure 1.

Probability of agreement with statements by latent class analysis groups.

PrEP Knowledge and Experience Scale

There were no differences across classes with respect to demographic characteristics, medical specialty, years of, or size of practice (Table 4). There was, however, a significant difference in the PrEP knowledge/experience scale. Class 2 showed the higher score, indicating less experience with PrEP (t(22.7) = 2.88, P = 0.009). Differences were explained by class 2 being significantly more likely than class 1 to be working in practices without written PrEP protocols (96% vs. 76%; χ2(2) = 11.41, P = 0.003); significantly less likely to have had PrEP requests in the previous 6 months (71% vs. 41%; χ2(2) = 13.62, P = 0.004); and significantly less likely to have ever prescribed PrEP (90% vs. 63%; χ2(2) = 18.74, P < 0.001).

Likelihood of Prescribing PrEP to Certain Patients

There was a moderate but not statistically significant difference in the likelihood of prescribing PrEP to patients of differing characteristics scale (t(21.5) = 1.95, P < 0.07). More clinicians in class 2 than class 1 were likely to prescribe to individuals with multiple sex partners (43% vs. 20%; χ2(4) = 10.13, P = 0.04), and a history of noninjection drug abuse (24% vs. 7%; χ2(2) = 18.08, P = 0.001). Both classes, however, reported low likelihood of intending to prescribe to patients with a history of missing medical visits (4.0%–4.8%) or a history of medication nonadherence (2.4%–4.0%).