Primary Care Outcomes in Patients Treated by Nurse Practitioners or Physicians:  A Randomized Trial FREE

The many pressures on the US health care system and greater focus on health promotion and prevention have prompted debates about primary care workforce needs and the roles of various types of health care professionals. As nurse practitioners seek to define their niche in this environment, questions are often raised about their effectiveness and appropriate scope of practice. Several studies conducted during the last 2 decades^1- 4 suggest the quality of primary care delivered by nurse practitioners is equal to that of physicians. However, these earlier studies did not directly compare nurse practitioners and physicians in primary care practices that were similar both in terms of responsibilities and patient panels.

Over time, payment policies and state nurse practice acts that constrained the roles of nurse practitioners have changed. In more than half the states, nurse practitioners now practice without any requirement for physician supervision or collaboration, and in all states nurse practitioners have some level of independent authority to prescribe drugs.⁵ Additionally, nurse practitioners are now eligible for direct Medicaid reimbursement in every state, direct reimbursement for Medicare Part B services as part of the 1997 Balanced Budget Act,⁶ and commercial insurance reimbursement for primary care services within limits of state law. Finally, state law determines whether nurse practitioners are eligible for hospital admitting privileges, either by regulating access at the state level or by allowing local hospital boards to decide. The combination of authority to prescribe drugs, direct reimbursement from most payers, and hospital admitting privileges creates a situation in which nurse practitioners and primary care physicians can have equivalent responsibilities. The present study is a large randomized trial designed to compare patient outcomes for nurse practitioners and physicians functioning equally as primary care providers.

The opportunity to compare the 2 types of providers was made possible by several practice and policy innovations at the Columbia Presbyterian Center of New York Presbyterian Hospital in New York City. In 1993 when the medical center sought to establish new primary care satellite clinics in the community, the nurse practitioner faculty were asked to staff 1 site independently for adult primary care. This exclusively nurse practitioner practice was to be similar to the clinics staffed by physicians. All are located in the same neighborhood, serve primarily families from the Dominican Republic who are eligible for Medicaid, and follow the policies and procedures of the medical center. The nurse practitioner practice, the Center for Advanced Practice, opened in the fall of 1994.

New York State law allows nurse practitioners to practice with a collaboration agreement that requires the physician to respond when the nurse practitioner seeks consultation. Collaboration does not require the collaborating physician to be on site and requires only quarterly meetings to review cases selected by the nurse practitioner and the physician. The state also grants nurse practitioners full authority to prescribe medications, as well as reimbursement by Medicaid at the same rate as physicians. The medical board granted nurse practitioners who were faculty members in the school of nursing hospital admitting privileges, thereby making the basic outpatient services, payment, and provider responsibilities the same in the nurse practitioner and physician primary care practices. Additionally, nurse practitioners and physicians in the study were subject to the same hospital policy on productivity and coverage, and a similar number of patients were scheduled per session in each clinic.

While it has been posited that nurse practitioners have a differentiated practice pattern focused on prevention with lengthier visits,⁷ this study was purposely designed to compare nurse practitioners and physicians as primary care providers within a conventional medical care framework in the same medical center, where all other elements of care were identical. Nurse practitioners provided all ambulatory primary care, including 24-hour call, and made independent decisions for referrals to specialists and hospitalizations. The Spanish language ability of the nurse practitioners and physicians was similar, although the physicians had somewhat better Spanish facility on average. All of the nurse practitioners (n = 7) and most of the physicians (n = 11) had limited knowledge of Spanish, and 6 physicians were either fluent or bilingual. Staff who served as interpreters were available at each study site. The central hypothesis was that the selected outcomes would not differ between the patients of nurse practitioners and physicians.

Between August 1995 and October 1997, adult patients were recruited consecutively at 1 urgent care center and 2 emergency departments that are part of the medical center. Patients who reported a previous diagnosis of asthma, diabetes, and/or hypertension, regardless of the reason for the urgent visit, were oversampled to create a cohort of patients for whom primary care would have an impact on patient outcomes, as has been postulated in previous studies.^8- 9 Patients were screened by bilingual patient recruiters and asked to participate if they had no current primary care provider at the time of recruitment and planned to be in the area for the next 6 months. The study was approved by the institutional review board of Columbia Presbyterian Medical Center. After an oral explanation of the consent form, written informed consent was obtained from each patient (both English and Spanish explanations and forms were available).

Those who provided informed consent were randomly and blindly assigned to either the nurse practitioner or 1 of the physician practices. Different assignment ratios were used during the recruitment period. Initially the ratio was 2:1, with more patients assigned to the nurse practitioner practice, because it opened after the physician practices and was able to accept more new patients. Subsequently, the ratio was changed to 1:1 as the nurse practitioner practice's patient panel increased. Despite this change, the mean number of days between the urgent visit at which patients were recruited and the follow-up appointments was similar (8.6 days for patients assigned to nurse practitioners compared with 8.9 days for patients assigned to physicians).

Recruited patients were then offered the next available appointments at the assigned clinic, and project staff made reminder calls the day before the appointments. Patients who missed their appointment were offered another appointment at the assigned practice. After patients kept their initial appointments, they were considered enrolled in the study and eligible for follow-up data collection.

Patients were told which provider group they were assigned to after randomization, and the type of provider could not be masked during the course of care. Patients who refused to participate or were deemed ineligible for the study were given follow-up primary care appointments by the study recruiters to the same practices. Additionally, during the study period, all practices received new patients from usual sources such as hospital discharges, recommendations from friends and family, referrals from other physicians, direct access by the patients themselves, and advertising. The study did not require a different process of care or documentation for enrolled patients.

At the initial visit, the patients became a part of the nurse practitioner or physician practices' regular patient panel, and all subsequent appointments, care, and treatments were arranged through the practice site of the assigned primary care nurse practitioner or physician. The primary care nurse practitioners and physicians had the same authority to prescribe, consult, refer, and admit patients. Furthermore, they used the same pool of specialists, inpatient units, and emergency departments. No attempt was made to differentiate study patients from other patients in the practice or to influence the practice patterns of the participating nurse practitioners and physicians. However, patients were free to change their source of medical care during the study. Medicaid in New York is currently fee-for-service and patients could go to other providers, go to a specialist directly, or use the emergency department without notifying their primary care provider. Approximately 3% of patients (n = 43) went to another clinic after keeping the first randomly assigned appointment, and 9% (n = 116) went to multiple primary care clinics during the 6-month period.

At the time of recruitment, patients provided demographic and contact information and completed the Medical Outcomes Study 36-Item Short-Form Health Survey (SF-36). After the initial primary care visit, interviewers contacted the enrolled patients either by telephone or in person, if necessary, to administer a satisfaction questionnaire. Six months after this initial appointment, the enrolled patients were again contacted and asked to complete a second, longer interview. The decision to interview patients 6 months after the initial primary care visit was based on prior survey experience with this patient population.¹⁰ The primary care patients served by the medical center are primarily immigrants and frequently change residences, travel between New York and their countries of origin, and have interruptions in telephone service. Attempts were made to locate all enrolled patients for this follow-up, including those who could not be located for the initial satisfaction interview. At the 6-month interview, the SF-36 and the satisfaction questionnaire were repeated, and additional questions were asked about health services utilization. A research nurse accompanied the interviewers, and for patients who reported a diagnosis of asthma, diabetes, or hypertension, physiologic data were collected.

Data on all health services utilization at the assigned practice and all other medical center sites were obtained from the medical center computer records for both the 6 months prior to recruitment and for 6 months and 1 year after the initial primary care appointment. These data were collected for all patients who were enrolled, including those who could not be located for the 6-month follow-up interview. Utilization data were also available for patients who were recruited but who did not keep their initial primary care appointment and therefore were not enrolled in the study. For these patients, the data were collected for the 6 months prior to recruitment and 6 months and 1 year after the date of the missed appointment they were given at recruitment.

The SF-36 was used as a baseline and follow-up measure of health status. This instrument elicits patient responses to 36 questions designed to measure 8 health concepts (general health, physical function, role-physical, role-emotional, social function, bodily pain, vitality, and mental health)¹¹ or to create 2 summary scores (physical component summary and mental component summary).¹² The origin and logic of the item selection, as well as the psychometrics and tests of clinical validity, have been reported by the survey's developers.^13- 14 Additionally, the survey's utility for monitoring general and specific populations, measuring treatment benefits, and comparing the burden of different diseases has been documented in 371 studies published between 1988 and 1996.^15- 16 For example, the SF-36 has been used to measure differences in function between chronically ill patients with and without comorbid anxiety disorder¹⁷; has demonstrated that it can detect changes in health status that correspond to clinical profiles for 4 common conditions¹⁸; and has shown that it reflects changes in health status that correspond to a predicted clinical course for elective surgery patients.¹⁹

Patient satisfaction was measured by using "provider-specific" items from a 15-item satisfaction questionnaire used in the Medical Outcomes Study.²⁰ Three items related to clinic management were included in the survey to provide the medical center administration with information about patients' perceptions of the clinic, but those items were not intended for use in the comparison of providers.

The survey instruments used in the study were written in English and then translated into Spanish. The bilingual members of the study team reviewed the Spanish versions to ensure that the meaning had not been changed. Approximately 80% (78.8% at recruitment and 83.7% at 6 months) of the interviews were conducted in Spanish.

Physiologic measures included disease-specific clinical measurements taken by a research nurse at the time of the 6-month follow-up interview. Blood pressure was determined for patients with hypertension, peak flow for those with asthma, and glycosylated hemoglobin for those with diabetes.

Utilization data included hospitalizations, emergency department visits, urgent care center visits, visits to specialists, and primary care visits within the Columbia Presbyterian Medical Center system. Only visits with a nurse practitioner or physician at a primary care site were counted as primary care. Specialty visits were defined as visits to a medical specialty clinic or specialist physician office. Emergency department and urgent care center visits were combined before analysis.

Recruitment and enrollment goals were established based on estimates of the sample size needed to detect a difference of 5 points on a 100-point scale for the SF-36 scores on all scales when comparing 2 groups with repeated measures. As the randomization ratio was projected to change during the course of the study with availability of appointments, it was projected that the final ratio between the 2 groups would be 1 patient in the physician group for every 1.5 patients in the nurse practitioner group. The sample size estimates for unequal groups were extrapolated from those presented by the instrument's developer for equal groups, assuming α = .05, 2-tailed t test, and power of 80%. Differences of more than 5 points are considered clinically and socially relevant, according to the guidelines for the interpretation of the survey.¹¹

Baseline demographics and health status for the nurse practitioner and physician groups at randomization and following enrollment were compared using χ² and t tests. Ten of the 12 satisfaction questions were factor analyzed (the 11th question that asks whether the patient would recommend the clinic to family and friends was left as a separate item; an item about medication instructions was dropped, as it was not applicable to the majority of respondents who were not prescribed any medications at their first visit). There were 3 factors with eigenvalues greater than 1, indicating that they represented reasonable constructs. The first, "provider attributes"(Cronbach α = .80) rated the provider on technical skills, personal manner, and time spent with the patient on a 5-point scale from poor to excellent. "Overall satisfaction" (Cronbach α = .86) was the factor created from 2 items addressing the quality of care received and overall satisfaction with the visit. The "communications" factor (Cronbach α = .59) combined 5 areas in which patients may have had problems understanding the provider's assessment and advice. Mean scores were computed for each factor.

Using the data collected at recruitment, mean baseline scores on the SF-36 for the scales and summary scores were used to establish the comparability of the nurse practitioner and physician groups in terms of health status. Four types of analyses were conducted using the SF-36 as an outcome measure. The first 2 included t tests to compare mean scores for nurse practitioner and physician patients at 6-month follow-up (both unadjusted and adjusted for baseline demographics and health status) and baseline to 6-month change scores. The third was a subgroup analysis designed to compare the sickest patients. Patients whose baseline score on the physical component summary of the SF-36 was in the bottom quartile (sickest) of the study sample were selected, and 6-month follow-up SF-36 scores were compared using the same analyses used for the total sample.

The fourth analysis classified patients into categories according to the change from baseline to follow-up in each patient's individual scores on the summary measures. This analysis was modeled on a comparison of patients treated in health maintenance organization and fee-for-service systems.²¹ The SE of measurement was used to create 3 categories: "same" (change not greater than what would be expected by chance), "better" (improved more than expected), and "worse" (declined more than expected).¹² While these definitions are based on a statistical construct, they provide results that may be more clinically relevant than mean scores or mean change in scores over time. A χ² test was then used to compare the distribution of the nurse practitioner and physician patients among these groups. In addition, the change from baseline to follow-up for the entire sample was compared using paired t tests.

Ranges and mean values for the physiologic measures were obtained, and mean values for the 2 groups were compared using t tests.

For the analyses of health services utilization, data were obtained for 6 months prior to the date of recruitment, 6 months after, and 1 year after the first primary care visit. Neither the recruitment visit nor the assigned primary care visit was included. Comparisons between the nurse practitioner and physician patients' health services utilization after enrollment were made using χ² tests (unadjusted) and Poisson regression (adjusted). To compare the utilization prior to recruitment with that following, signed rank tests were used.

The 159 patients (12.1%) who, after the first visit, either went to a clinic other than the one assigned or to multiple primary care clinics were maintained in the initially assigned group for the analyses, consistent with an intent-to-treat analysis. All analyses were repeated without these 159 patients, and the results were the same.

Of the 3397 patients screened and given follow-up appointments, 41.6% were not randomized because they either refused to participate (11.2%) or did not meet the screening criteria (30.4%). Of the 1981 patients who were randomized, 1181 (59.6%) were assigned to the nurse practitioner clinic and 800 (40.4%) to the physician clinics. The average age of the randomized patients was 44.4 years and 74.6% were female; 84.9% were Hispanic, 8.8% were black, and 1.1% were white. There were no statistically significant differences in the demographics or health status of the patients randomized to nurse practitioners or physicians (Table 1).

The 1316 patients (66.4%) who kept their initial primary care appointments following randomization were considered enrolled in the study. This rate is comparable to the normal rate of appointments (65%) kept at the participating clinics (P. Craig, MA, RN, e-mail message, August 4, 1999). Compared with the 665 patients (32.4%) who did not keep their appointments, those who did (the enrolled patients) differed significantly at baseline in several respects. Enrolled patients were older (45.9 vs 41.3 years); a higher proportion were female (76.8% vs 70.2%) and Hispanic (90.3% vs 82.9%); a higher percentage reported a history of 1 or more of the selected chronic conditions (53.7% vs 45.0%); and they had to wait fewer days for their follow-up appointments (7.8 vs 10.7). These findings are consistent with other studies of patient behavior relative to keeping or missing appointments.^22- 24

Our analysis of the data available on patients who did not keep their primary care appointments found no differences in health services utilization after 1 year among the patients assigned to the nurse practitioner group and physician group.

The difference in the retention rates between recruitment and enrollment for the nurse practitioner group (68.2%) and the physician group (63.8%) was statistically significant (χ²₁ = 4.3, P = .04). However, neither the patients who enrolled nor those who failed to keep their appointments differed significantly between the nurse practitioner and physician groups in terms of baseline demographics, SF-36 scores, or patient-reported prior diagnosis of the selected chronic conditions (Table 1).

Among the nurse practitioner patients, 59% saw the same provider for all primary care visits in the first year after the initial visits compared with 54% of the physician patients, and this difference was not statistically significant (χ²₁ = 2.7, P = .11).

Initial satisfaction interviews were completed for 90.3% (n = 1188) of all patients who made a first clinic visit (90.8% of the nurse practitioner group and 89.4% of the physician group). Almost 92% of all completed interviews took place within 6 weeks of the initial appointment.

Six-month interviews were completed for 79% of all enrolled patients (80.5% of the nurse practitioner group and 76.7% the physician group). This completion rate is considered high for a transient immigrant population and is comparable to or better than that achieved by other studies in the area served by the medical center. The majority of completed interviews (91.4%) took place between 180 and 240 days after the initial appointment. The most common reasons for loss to follow-up were the inability to locate the patient (65.9%) or that the patient had moved out of the area (17%). A small number of patients (23 [2.8%] in the nurse practitioner group and 16 [3.1%] in the physician group) refused to complete the interview when they were contacted. Five patients (2.9%) were located but were unable to complete the interview due to physical limitations or mental illness, and 3 patients (1.1%) were deceased. Figure 1 summarizes the participation rates at each major stage in the study.

There were no significant differences in the scores between nurse practitioners and physicians for any of the satisfaction factors after the first visit (Table 2). At the 6-month interview there were no statistically significant differences in "overall satisfaction" or "communications" factors or in willingness to refer the clinic to others. The difference in mean score for the "provider attributes" factor, however, was significant, with the physician group rating providers higher than the nurse practitioner group (4.22 vs 4.12 out of a possible 5; P = .05). The provider attribute consists of patients' ratings of the providers' technical skill, personal manner, and time spent with the patient. The clinical significance of a 0.1 difference on a 5.0 scale is unlikely.

Overall, the health status of the study group improved from baseline to follow-up, and the improvement was statistically significant on every scale (Table 3).

There were no significant differences between the nurse practitioner and physician patients on any scale or summary score at 6 months. This is true for both the unadjusted scores and scores adjusted for demographics and baseline health status. The additional analysis (not shown) of the summary scores, using the change categories of "same," "better," and "worse" to characterize the clinical course of each patient, also revealed no significant differences between provider types.

Finally, 152 nurse practitioner patients and 103 physician patients were defined as the sickest (health status scores in the bottom quartile of the sample at baseline) and their scores analyzed separately. Again, there were no differences between nurse practitioner and physician patients in scale scores or summary measures at 6 months (both unadjusted and adjusted), nor did the change in scores from baseline to follow-up differ between nurse practitioner and physician patient groups.

The physiologic measures taken at the time of the interview for patients who reported 1 of the selected chronic illnesses were not statistically significantly different between the nurse practitioner and physician patients for asthma and hypertension. The mean peak flow measurements for the 64 physician patients with asthma was 292 L/min, compared with 297 L/min for the 107 nurse practitioner patients (t test = −0.29, P = .77). Glycosylated hemoglobin mean value for the 46 physician patients with diabetes was 9.4% vs 9.5% for the 58 nurse practitioner patients (t test = −0.22, P = .82).

For patients with hypertension, there was no statistically significant difference in the systolic reading: 139 mm Hg for the 145 physician patients and 137 mm Hg for the 211 nurse practitioner patients (t test = 1.08, P = .28). The mean diastolic reading, however, was statistically significantly lower for the nurse practitioner patients at 82 mm Hg compared with 85 mm Hg for the physician patients (t test = 2.09, P = .04).

For our comparison of outcomes we analyzed utilization of health care services for nurse practitioner and physician patients who enrolled in the study by keeping their initial primary care appointment. There were no statistically significant differences between the nurse practitioner and physician patients for any category of service during either the first 6 months or the first year after the initial primary care visit for either unadjusted or adjusted use rates (Table 4). When the utilization analyses were repeated for the subsets of "sickest" patients as defined in the "Self-reported Health Status" section above, no differences were found in the health care services utilization between the nurse practitioner and physician patients (Table 5). In the 6 months and 1 year after the initial primary care visit, enrolled patients in both groups made significantly more primary care and specialty visits and fewer emergency/urgent visits than in the 6 months prior to recruitment. The percentage of enrolled patients hospitalized was not significantly different for either 6 months or 1 year after the initial primary care appointment.

This study was designed to compare the effectiveness of nurse practitioners with physicians where both were serving as primary care providers in the same environment with the same authority. The hypothesis predicting similar patient outcomes was strongly supported by the findings of no significant differences in self-reported health status, 2 of the 3 disease-specific physiologic measures, all but 1 of the patient satisfaction factors after 6 months of primary care, and in health services utilization at 6 months and 1 year.

The difference between the nurse practitioner and physician patients' mean ratings of satisfaction with provider attributes was small but statistically significant. It may be attributable to the fact that the nurse practitioner practice was moved to a new site after 2 years and before recruitment and data collection were completed; the physician practices were not moved during the study period. When the "provider attribute" subscale scores for the nurse practitioner and physician patients whose 6-month follow-up period overlapped this move were compared, the ratings by nurse practitioner patients were significantly lower than those of the corresponding physician patients (4.16 vs 4.36; P = .04). There was no significant difference in ratings among patients not affected by the move. Additional research will be needed to determine whether this is a persistent difference or if it results from conditions unique to this study.

A statistically significant, but small, difference was discerned in the mean diastolic blood pressure of patients with hypertension, with the nurse practitioner group having a slightly lower average reading at 6 months. Given the size of this change and the lack of differences in self-reported health status, there does not seem to be an obvious reason for this difference.

Although insufficient statistical power to discern differences has been a problem in much of the previous research comparing nurse practitioners and physicians, the sample size in this study was adequate to test the hypothesized similarity of nurse practitioner and physician groups. At the end of the study, power calculations were repeated using final sample size and the means and SDs from these data. These revealed that the sample size was adequate to detect differences from baseline to follow-up between the 2 patient groups of less than 5 points for 6 of the 8 scales (3.2 for general health; 3.3 for vitality; 3.4 for mental health; 3.4 for social function; and 4.2 for bodily pain) and less than 6 points on 2 scales (5.9 on role-physical and role-emotional). This magnitude of difference is similar to differences commonly reported in studies comparing groups^21,25 and in studies of change over time within 1 group.^17,26

There is evidence that the outcome measures chosen were sensitive enough to discern any important differences. The SF-36 is a widely used outcome measure and its sensitivity has been documented in several studies.^11,18,27 In this study, there were sizable and statistically significant changes for both nurse practitioner and physician patients in all scale scores and summary measures from baseline to follow-up. Some improvement would be expected, even over a 6-month period with or without primary care, following the urgent care visits at which subject recruitment occurred; the SF-36 did detect improvement. The utilization indicators are in widespread use in cross-sectional and longitudinal studies. With the exception of number of hospitalizations, which stayed the same in both groups, these measures also changed significantly over time.

Strengths of this study included adequate sample size and the ability to randomize patients to equivalent clinical settings and to providers with equal responsibilities. However, there were also several limitations.

Patients could not be randomized at the point of initial contact with the provider. Because the nurse practitioner and physician practice sites were geographically separate, patients had to be randomized when they were recruited in the emergency department or urgent care center to give them follow-up appointments at various locations with different appointment schedules. This time and location gap likely contributed to the loss of almost one third of the sample between randomization and enrollment. Although this is substantial, it is within the range reported in similar randomized trials.²⁸

While the loss rate was significantly different for the nurse practitioner and physician groups, there is no reason to suspect that this represents a systematic violation of the protocol or any compromise of randomization. Patients dropped out before receiving care, and the dropout rate was higher for those assigned to the traditional model of physician care. This suggests that assignment to the new model of nurse practitioner care did not negatively influence patient behavior. There is no evidence of selection bias in that there were no significant differences in demographics, baseline health status, or prerecruitment health services utilization patterns between nurse practitioner and physician randomized patients, for either those who enrolled or those who did not keep their appointments.

A 1-year follow-up for SF-36 and patient satisfaction would have been more useful than taking these measures at 6 months. In part, we believed a population with limited access to health care would show changes in these measures in 6 months. But more influential in the decision regarding follow-up was the knowledge that this population is difficult to track because of changing addresses, changing eligibility for Medicaid, and frequent extended trips out of the country. Although we do have service utilization data for both 6 months and 1 year, data on satisfaction and self-perceived health status were not collected for 1 year.

Finally, the study had some characteristics that limit the generalizability of results. It was conducted in medical center–affiliated, community-based primary care clinics, which may differ from individual providers or small group practices. The providers were faculty from a university medical center, hence were not necessarily typical of those in nonacademic practice settings. The patients were predominantly immigrants from the Dominican Republic who were eligible for Medicaid and many did not speak English. This differs from the setting in which many commercially insured patients receive primary care but does resemble other academic, public and safety net providers, and the Medicaid populations they serve. While the setting and patient population are limitations, they are also what permitted randomized assignment and an environment in which nurse practitioners and physicians were able to function equally as primary care providers. The ability to do this type of study, even in a setting atypical for some patients, adds significant weight to the results from prior studies that have demonstrated the competence of nurse practitioners.

Who provides primary care is an important policy question. As nurse practitioners gain in authority nationally with commercially insured and Medicare populations now accessing nurse practitioner care, additional research should include these populations. As cost and quality issues pervade the public debate on managed care, those who are the first-line health care providers become pivotal resources in the emerging health care system. Nurse practitioners have been evaluated as primary care providers for more than 25 years, but until now no evaluations studied nurse practitioners and physicians in comparable practices using a large-scale, randomized design. The results of this study strongly support the hypothesis that, using the traditional medical model of primary care, patient outcomes for nurse practitioner and physician delivery of primary care do not differ.