Quality of HIV Care Provided by Nurse Practitioners, Physician Assistants, and Physicians FREE

Nurse practitioners (NPs) and physician assistants (PAs) or physicians often deliver primary care to patients with HIV.

Researchers reviewed records of 6651 patients at 68 HIV clinics to compare NPs' and PAs' and physicians' quality of care. Nurse practitioners and PAs and physicians with HIV expertise performed similarly on 6 of 8 guideline-related quality measures, while NPs' and PAs' patients had higher rates of purified protein derivatives and Papanicolaou smears than expert physicians' patients. Nurse practitioners and PAs performed better on 6 of 8 quality measures than generalist physicians without HIV expertise.

Experienced NPs and PAs might provide basic guideline-related care similar to that of physician experts and better than that of physicians without HIV-related expertise.

—The Editors

Nurse practitioners (NPs) and physician assistants (PAs) often practice independently and can prescribe medications in most states (1). Their numbers have dramatically increased in recent years (1–2), and understanding how they perform is an important health policy issue. This question may be particularly important in rural and inner city areas with physician shortages and in countries with limited health care resources (3). Studies have suggested that both the quality (4–6) and the cost-effectiveness (7) of primary care provided by NPs are similar to that provided by physicians. These findings, however, may not be generalizable to other settings and, in particular, to specialty care (8–9).

Since the beginning of the HIV epidemic, NPs and PAs have played a critical role in caring for persons with HIV, both as members of care teams and as primary HIV care providers. To our knowledge, no studies have assessed the performance of NPs and PAs in the role of primary HIV care provider. Thus, we assessed the quality of care provided by NPs and PAs who may specialize in HIV care and have high HIV caseloads. Because the highest-quality care is probably provided by multidisciplinary care teams in organizations that have access to the many services needed by patients with HIV (10), we also compared the characteristics of organizations in which NPs and PAs are primary HIV care providers with organizations in which only physicians are primary HIV care providers.

We collected data for our study as part of a controlled evaluation of a quality improvement collaborative involving HIV care sites that received funds from the Ryan White Comprehensive AIDS Resources Emergency (CARE) Act Title III (11). Participants attended 4 learning sessions over a 16-month period and received training in quality improvement strategies. We selected control sites from CARE Act Title III clinics that did not participate in the collaborative (details to follow).

Of the 200 CARE Act Title III sites in the continental United States in May 2000, we excluded 16 sites that reported caseloads of fewer than 100 cases per year, 12 sites that were initially slated to participate in the collaborative but elected not to participate, and 1 site that lost its CARE Act Title III funding. Of the 62 sites that agreed to participate in the collaborative, 44 (71%) sites provided chart review data. Of the 109 nonparticipating sites that were eligible to be control sites, 65 sites provided the information that we needed for matching. Of the 40 potential control sites, 25 (63%) sites agreed to provide medical records for chart review. More details about the collaborative and study are provided elsewhere (12). The Committee on Human Studies of Harvard Medical School approved the study protocol.

We randomly sampled 75 patients who had 1 visit or more during the review period from each site before and after the intervention, which took place between 30 June 2000 and 31 September 2001.

Study facilitators at each site identified all clinicians (including physicians, NPs, and PAs) with primary responsibility for caring for patients with HIV. For sites with 5 or fewer clinicians, all clinicians were eligible for the survey. For sites with more than 5 clinicians, we randomly sampled 5 clinicians. We asked selected clinicians about experience with quality improvement initiatives, physician background, and training. Of 279 clinicians surveyed, 243 (87%) clinicians responded. Response rates did not differ according to region, type of clinic, or survey wave.

Since we selected only up to 5 clinicians per clinic, we did not survey all clinicians. Thus, for our analyses, we limited our sample to patients linked to a clinician who returned a survey. We asked medical record reviewers, who were usually clinic nurses, to identify the name and qualifications (for example, physician, NP, or PA) of the clinician “who makes most major (for example, changing antiretroviral regimen) decisions regarding this patient's care.” Reviewers identified a responsible clinician for 8841 of 9020 (98%) patients. We linked 6551 (74.1%) of these patients with a surveyed clinician. Compared with patients linked with surveyed clinicians, unlinked patients were slightly more likely to be women (34% vs. 31%; P = 0.016), have more HIV-related diagnoses (0.20 vs. 0.17; P = 0.009), and have CD4 cell counts lower than 0.2 × 10⁹ cells/L (34% vs. 30%; P < 0.001). Rates of hepatitis C positivity, psychiatric disorders, and drug abuse were similar.

Each clinic selected 1 or 2 reviewers who were trained to use the chart abstraction tool that we developed for our study. Chart reviewers were blinded to the study hypothesis. Data abstracted included age and sex, history of HIV-related illnesses, comorbid medical or psychiatric conditions (including current substance abuse or psychiatric illness), screening and prophylaxis against HIV-related conditions, number and timing of visits, CD4 cell counts, viral loads, and antiretroviral medications prescribed. We asked medical record reviewers to specify whether each visit was with a physician, an NP or PA, a nurse, or some “other” clinician (for example, a nutritionist). Of the 2 review periods, the first covered the year before the intervention (1 June 1999 to 30 May 2000) and the second covered the year beginning 6 months after the start of the intervention and ending 3 months after the end of the intervention (1 January 2001 to 31 December 2001) (Figure 1).

We based the 8 quality-of-care measures on guidelines, which did not change over this time period (13–15). Our primary measures were rates of highly active antiretroviral therapy (HAART) use at the last visit during the review period and control of HIV viral load for appropriate patients. We considered patients to be eligible for HAART if their CD4 cell counts were less than 0.5 × 10⁹ cells/L or their viral loads were greater than 20 000 copies/mL and if they were already receiving HAART, as per guidelines in effect at the time (13). We considered viral load to be controlled if it was undetectable or less than 400 copies/mL. We also assessed whether patients were screened for tuberculosis, hepatitis C, and cervical cancer; underwent appropriate Pneumocystis carinii prophylaxis; and received influenza vaccinations during the 1-year review period. For hepatitis C screening, we accepted documentation of a previous positive hepatitis C test result. We defined appropriate access to outpatient care as having a visit during at least 3 of 4 quarters.

We combined NPs and PAs for our analysis. Physician assistants practice under the direction of a physician, which is not true for NPs, but both often have considerable autonomy in practice (1). We asked physicians about the highest level of training that they had received (for example, family medicine or infectious disease fellowship) and whether they considered themselves to be “specialists” (hereafter called “experts”) in treating HIV. For physicians, we created a composite training or expertise variable with 3 levels: infectious disease physicians, generalist HIV experts, and generalist non–HIV experts. Studies have shown that this variable is related to referral patterns, continuing medical education activities, knowledge levels about HIV care, and rates of prescribing HAART (16–17).

All analyses were cross-sectional. First, we compared the characteristics of physicians and NPs and PAs who were primary HIV care providers and the characteristics of their patients. Next, we compared the characteristics of the sites at which only physicians were primary HIV care providers (n = 32 [“physician sites”]) with those of the sites at which NPs and PAs were primary HIV care providers (n = 36 [“NP and PA sites”]). We used chi-square tests for categorical variables, t-tests for normally distributed continuous variables, and Wilcoxon rank-sum tests for variables that were not distributed normally. We used SAS, version 8.2 (SAS Institute, Inc., Cary, North Carolina), for these analyses.

Finally, we used hierarchical multiple logistic regression to compare the physician groups (infectious disease specialists, generalist HIV experts, and generalist non–HIV experts) with NPs and PAs on each of the 8 quality measures. All models controlled for patient age, sex, stage of disease (on the basis of lowest recorded CD4 cell count over the period of care), active psychiatric or substance abuse problems, history of HIV-related diagnoses, comorbid medical conditions, study period (preintervention vs. postintervention), and an intervention or control indicator. To assess possible confounding by site, we estimated an additional model that included variables that statistically significantly differed between physician sites and NP and PA sites. To account for the clustering effects in our sample, we specified a 3-level mixed-effects model, with patients nested within providers and providers nested within clinics. For these analyses, we used the GLLAMM (Generalized Linear Latent And Mixed Models) program with Stata, version 8.0 (Stata Corp., College Station, Texas) (18). We calculated adjusted rates and 95% CIs by using the estimated logistic regression coefficients and their covariance matrix.

The Agency for Healthcare Research and Quality and the Lifespan/Tufts/Brown Center for AIDS Research funded this study. The funding sources had no role in the design, conduct, or reporting of the study or in the decision to submit the manuscript for publication.

Of the 243 primary HIV care providers, 66 were NPs or PAs. They were the primary HIV clinicians for 20% (1304 of 6651 patients) of the study patients. Compared with physicians, NPs and PAs were more often women (76% vs. 39%; P < 0.001) and were more often white (79% vs. 66%; P = 0.004) (Table 1). Nurse practitioners and PAs tended to see fewer patients per week (36 patients vs. 49 patients; P = 0.014), reported being the primary provider for fewer patients with HIV (107 patients vs. 164 patients; P = 0.025), and reported caseloads with higher fractions of patients with HIV (85% vs. 61%; P < 0.001). Of the 66 NPs and PAs, 51 were NPs and 15 were PAs (data not shown). Nurse practitioners were more likely than PAs to be women (84% vs. 47%; P = 0.003) and to be “very satisfied” with their practice (45% vs. 13%; P = 0.026), but all other characteristics were similar (Table 1).

Compared with patients primarily cared for by physicians, patients primarily cared for by NPs and PAs were slightly younger (39.7 years of age vs. 40.7 years of age; P < 0.001) (Table 2), were more often women (43% vs. 28%; P < 0.001), had fewer non–HIV-related comorbid conditions (0.37 vs. 0.41; P = 0.028), and were more likely to have active substance abuse documented in their medical record (22% vs. 15%; P < 0.001).

We compared the characteristics of the 32 sites at which physicians were primary HIV care providers (physician sites) with those of the 36 sites at which NPs and PAs were primary HIV care providers (NP and PA sites) (Table 3). The mean number of full-time–equivalent NPs and PAs was greater in NP and PA sites than in physician sites (3.1 vs. 2.3; P = 0.005), but the mean number of full-time–equivalent physicians did not significantly differ between the sites (4.1 vs. 7.4; P = 0.061). The NP and PA sites were more likely to have at least 1 substance abuse counselor (63% vs. 34%; P = 0.020). Clinic type, clinic staffing, HIV specialization, presence of a multidisciplinary HIV team, or how frequently the team met did not differ between the sites.

The mean number of clinic visits per year was 6.6 when a physician was the primary HIV care provider and 7.4 when an NP or PA was the primary HIV care provider (P < 0.001) (Figure 2). Patients with a physician as the primary HIV care provider had an average of 0.4 visit per year with NPs and PAs and 0.8 visit per year with other providers. Patients with an NP or PA as the primary HIV care provider made an average of 1.3 visits per year with physicians and 1.1 visits per year with other providers. Of the patients primarily seen by NPs and PAs, 46% visited a physician at some time during the study period. Of the patients primarily seen by physicians, 16.2% had at least 1 visit with an NP or PA (data not shown).

Performance rates for NPs and PAs were similar to or higher than physicians' rates for all 8 quality measures (Table 4). Rates of P. carinii prophylaxis and hepatitis C testing did not statistically significantly differ between NPs and PAs and physicians. Nurse practitioners and PAs had higher performance rates for purified protein derivative testing (0.63) than infectious disease–trained physicians (0.53), generalist HIV experts (0.47), and generalist non–HIV experts (0.49) (P < 0.050 for all). They also had higher performance rates for Papanicolaou (Pap) smears (0.71) than infectious disease–trained physicians (0.56), generalist HIV experts (0.62), and generalist non–HIV experts (0.52) (P < 0.050 for all). For the remaining measures (HAART use, viral load control for patients receiving HAART, influenza vaccine use, and visits), rates were higher for NPs and PAs than for generalist non–HIV experts (P < 0.050 for all) and were similar to infectious disease–trained physicians and generalist HIV experts. Models that controlled for the 2 variables that statistically significantly differed between sites (number of full-time–equivalent NPs and PAs and the presence of substance abuse counselors) yielded similar results.

In our national study of patients who were cared for in CARE Act Title III HIV clinics, 20% of patients received most HIV care from NPs and PAs. The performance of NPs and PAs in caring for patients with HIV was similar to that of physicians trained in infectious diseases and general medicine HIV experts for 6 of the 8 quality measures that we examined and was superior to that of HIV expert physicians for the other 2 quality measures. Nurse practitioners and PAs performed statistically significantly better than generalist non–HIV experts on 6 of the 8 quality measures. Performance rates substantially differed from a clinical standpoint, from 7 to 19 percentage points.

One possible explanation for our findings is selection bias. Patients cared for by physicians may have had more complications than patients cared for by NPs and PAs. Our data do not suggest this. Although patients cared for by NPs and PAs were slightly younger, were more likely to be women, and had slightly fewer comorbid conditions, they were also more likely to have drug abuse problems, and we adjusted for these and other patient characteristics in the multivariable model. Other unmeasured patient characteristics may relate to both selection into an NP and PA practice and the quality measures that we assessed. However, the consistency of the effects that we observed across measures of treatment, prophylaxis, prevention, and access supports the validity of our findings.

A second possibility is that NPs and PAs practiced at sites that were better organized for the care of patients with HIV. Our analyses of site characteristics, however, revealed only 1 statistically significant difference—more sites at which NPs and PAs were primary HIV care providers had substance abuse counselors. These sites had slightly fewer full-time–equivalent physicians, but this should have made it more difficult for NPs and PAs to perform well. When we controlled for these site-level factors, the findings did not change.

A third possibility is that NPs' and PAs' practices were less busy and were more focused on HIV care, allowing them to focus on the processes that we assessed. Nurse practitioners and PAs did see fewer outpatients per week than physicians and a higher percentage of patients with HIV. However, when we adjusted for these factors in multivariable models, our findings did not change, suggesting no confounding by site.

Our previous work showed that HIV expertise is strongly associated with higher-quality care for both physicians (12, 17) and HIV care sites (19). The NPs and PAs we studied were experienced HIV care providers, following an average of 108 patients with HIV, and most of their patients had HIV. In addition, they typically practiced in environments with several supports for HIV care, including HIV care teams and access to expert HIV physicians. Furthermore, more than 46% of patients cared for primarily by NPs and PAs also saw a physician during the year we studied. These data suggest that key elements in the performance of these NPs and PAs may be high levels of experience, focus on a single condition, and either participation in HIV care teams or other easy access to physicians with HIV expertise. Nurse practitioner or PA performance may also have been higher because these clinicians saw fewer patients per week than physicians.

Our data show that under certain conditions, NPs and PAs can function as lead HIV clinicians and can provide care that is similar to, and for some measures better than, that provided by physicians. Between 1992 and 2000, the number of NPs, PAs, and certified nurse midwives has increased by 160% (2). National data show that patients are seeing nonphysician providers in addition to, and not instead of, physicians (20–21), and our study suggests that this is also true for patients who see NPs and PAs for their HIV care.

Our findings may be important for HIV care sites located in rural, underserved, or resource-constrained care settings. In care settings where access to physician providers is limited, properly trained and supported NPs and PAs may help address this access problem. We should note, however, that HIV expert physician backup is almost certainly necessary for NPs and PAs to provide high-quality HIV care. Whether NPs and PAs are more cost-effective than physicians is unclear. Although NPs and PAs usually earn less than physicians (22), some data suggest that they have longer visits (6–7) and order more tests (6). Many countries with limited resources are scaling up the delivery of antiretroviral medications (23–29), and our data suggest that properly trained and supported nonphysician health professionals can play a useful role in these settings.

We examined quality measures that could be assessed by a medical record review. All but 1 (nondetectable viral load) of these measures are process measures. Our findings may have differed if we could have assessed mortality rates, hospitalization rates, appropriate management of opportunistic infections, changes in health status, medication adherence, or patient reports about or ratings of care (30). Similarly, our findings might have differed if we had measured the performance of more complex HIV-related care decisions (such as managing treatment failure) or had assessed other types of quality in these patients (such as appropriate care of depression, diabetes, hypertension, or hyperlipidemia). Four care processes that we assessed (purified protein derivative testing, hepatitis C screening, Pap smears, and influenza vaccinations) are simple for anyone with basic clinical training to implement, and the remaining 4 care processes (HAART, viral load control, P. carinii prophylaxis, and visit frequency) are the subject of detailed clinical practice guidelines (31–32). Nonetheless, we found similar results across all 8 quality measures. Finally, we studied patients receiving care in clinics that have CARE Act Title III funding, and we cannot know whether our findings are generalizable to other settings. Sites that receive this funding, however, disproportionately care for patients with low incomes and complex social problems. We therefore expect that NPs and PAs would have less, not more, difficulty caring for more advantaged populations.

The preconditions for the NP or PA performance observed in our study are high levels of experience, focus on a single condition, and either participation in teams or other easy access to physicians and other providers with HIV expertise. These results may be particularly salient for care settings in which access to physician HIV experts is limited, including countries with limited resources for health care.