Nurse Practitioners, Physician Assistants, and Physicians Are Comparable in Managing the First Five Years of Diabetes

Abstract

Background: Increasing use of nurse practitioners and physician assistants is a possible solution to the shortage of primary care providers in the United States, but the quality of care they provide is not well understood.

Methods: Because the scope of practice of the 3 provider types is similar in the Veterans Health Administration, we determined whether patients managed by primary care nurse practitioners, physician assistants, or physicians had similar hemoglobin A1c levels at comparable times in the natural history of diabetes. Our retrospective cohort study examined veterans with newly diagnosed diabetes in 2008, continuous primary care from 2008 to 2012, and more than 75% of primary care visits with nurse practitioner, physician assistant, or physician.

Results: Of the 19,238 patients, 95.3% were male, 77.7% were white, and they had a mean age 68.5 years; 14.7%, 7.1%, and 78.2% of patients were managed by nurse practitioners, physician assistants, and physicians, respectively. Median hemoglobin A1c was comparable at diagnosis (6.6%, 6.7%, 6.7%, P > .05) and after 4 years (all 6.5%, P > .5). Hemoglobin A1c levels at initiation of the first (7.5%-7.6%) and second (8.0%-8.2%) oral medications for patients of nurse practitioners and physician assistants compared with that of physicians was also similar after adjusting for patient characteristics (all P > .05). Nurse practitioners started insulin at a lower hemoglobin A1c (9.4%) than physicians (9.7%), which remained significant after adjustment (P < .05).

Conclusions: At diagnosis and during 4 years of follow-up, diabetes management by nurse practitioners and physician assistants was comparable to management by physicians. The Veterans Health Administration model for roles of nurse practitioners and physician assistants may be broadly useful to help meet the demand for primary care providers in the United States.

Keywords: Diabetes; Midlevel providers; Nurse practitioner; Physician assistant; Primary care.

Appendix Figure 1. Consort diagram

51,116 veterans had two years of primary care before a new diagnosis of diabetes (diabetes diagnosis was defined by (i) use of ICD-9 code 250.xx at a primary care visit, (ii) any use of 250.xx twice in 2008, and/or (iii) outpatient prescription of a diabetes drug). Of these, 35,643 had continuity of primary care from 2006–2012 defined as at least one primary care attending visit per calendar year, and 19,601 also had an HbA1c at diagnosis and at least one A1c per year from 2008–2012. Less than 2% of patients (n=363) lacked a majority of visits with one particular provider type. Because such a small percentage of patients had “mixed” provider type care, we included only the 19,238 patients who were managed either by NPs, PAs, or physicians. Of these, 2,821 were managed by NPs, 1,367 managed by PAs, and 15,050 managed by physicians. NP = nurse practitioner, PA = physician assistant, MD = physician (MD and DO).

Figure 1. Percent of patients prescribed diabetes medications, and referred to diabetes clinic, from 2008–2012 by managing provider type

Bar graphs depicting: percent of patients who completed at least one diabetes specialist clinic visit between 2008–2012; percent of patients who were prescribed any diabetes medication and/or insulin. DM = diabetes, Rx = prescription. Group P values shown were determined by Chi-square test.

Figure 2. Median HbA1c levels at systematic time points in the natural history of diabetes in patients newly-diagnosed in 2008

Box and whisker plots of median patient HbA1c levels compared between patients at various time points in the natural history of diabetes: 1) within 12 weeks before/6 weeks after diabetes diagnosis, 2) each patient’s mean HbA1c over 5 years from 2008–2012, 3) after four years of primary care follow-up in 2012, 4) at initiation of first oral antihyperglycemic medication, 5) at initiation of second oral antihyperglycemic medication, and 6) at initiation of insulin. Boxes include 25^th–75^th percentile, dashed lines indicate min-max range, dots represent outliers. Group P values shown were calculated using Kruskal-Wallis tests. Wilcoxon tests were used to compare HbA1c between two provider groups.