Quality of Outpatient Care With Internal Medicine Residents vs Attending Physicians in Veterans Affairs Primary Care Clinics

Abstract

This assessment of common primary care measures from 10 Veterans Affairs clinics from 2014 compares outcomes in patients of resident physicians with patients of attending physicians.

Concerns persist that care provided by resident physicians is of lower quality than that provided by more experienced attending physicians.¹ In this study, we compared quality of outpatient care between internal medicine residents and attending physicians in US Department of Veterans Affairs (VA) primary care clinics.

Methods

As part of an evaluation of the Centers of Excellence in Primary Care Education, an interprofessional education initiative, we examined 10 geographically diverse VA medical centers with affiliated internal medicine residencies. This work was determined to be a quality improvement activity per Veterans Health Administration policies, with a waiver of informed consent. We identified patients seen in teaching primary care clinics at each site who were assigned to the panel of either a resident physician or attending physician (staff physician who supervised resident’s care) and who had at least 1 primary care visit in calendar year 2014. We collected demographic characteristics, comorbidity,² and quality of care and health service utilization measures from the VA Corporate Data Warehouse in 2014. Outcomes included measures of diabetes care quality (annual glycated hemoglobin [HbA_1c] testing, HbA_1c poor control [>9% or unmeasured (to convert HbA_1c concentration to a proportion of total hemoglobin, multiply by 0.01)], and annual renal testing [urine microalbumin to creatinine ratio or prescription of an angiotensin-converting enzyme inhibitor or angiotensin receptor blocker]), use of a high-risk medication³ in patients older than 65 years, hypertension control (blood pressure <140/90 mm Hg), emergency department visits, and hospitalizations (all-cause hospitalizations and from ambulatory care–sensitive conditions). We selected these measures because they are common primary care measures that we could extract reliably. We compared outcome measures using logistic mixed models adjusted for demographic characteristics, comorbidities, and years under VA care and included site as a random effect to control for site-level effects. We used a 2-sided P < .05 as a significance threshold.

Results

Of 76 392 patients, mean (SD) age was 62.3 (15.7) years, 69 677 (90.6%) were male, 55 082 (72.1%) were white, 10 986 (14.4%) were black, 2568 (3.4%) were Hispanic, and 7756 (10.1%) were of another race/ethnicity. We identified 19 324 patients cared for by resident physicians and 57 068 patients cared for by attending physicians in 2014. Residents cared for a greater proportion of younger, female, and black patients (Table 1). Patients of resident physicians had a higher mean comorbidity score (14.1 [16.1] vs 12.5 [15.1] for attending physicians; P < .001), were more commonly diagnosed as having substance use disorders (3656 [18.9%] vs 7797 [13.7%]; P < .001) and depression (4270 of 19 324 [22.1%] vs 10 212 of 57 068 [17.9%]; P < .001), and were under VA care for less time.

Table 1. Characteristics of Patients Cared for by Internal Medicine Residents and Attending Physicians in VA Teaching Primary Care Clinics in 10 Medical Centers in 2014.

Characteristic	Patients, No. (%)		P Valuea
	Resident (n = 19 324)	Attending (n = 57 068)
Age, mean (SD), y	59.7 (15.6)	63.2 (15.6)	<.001
Male	17117 (88.6)	52560 (92.1)	<.001
Race/ethnicity
White	12 432 (64.3)	42 650 (74.7)	<.001
Black	3892 (20.1)	7094 (12.4)
Hispanic	746 (3.9)	1822 (3.2)
Other or unknown	2254 (11.7)	5502 (9.6)
Years of VA care, mean (SD)	8.7 (5.5)	9.9 (5.2)	<.001
Elixhauser comorbidity score, mean (SD)	14.1 (16.1)	12.5 (15.1)	<.001
Medically complex patientsb	2606 (13.5)	6363 (11.2)	<.001
Selected comorbidities
Congestive heart failure	1235 (6.4)	3049 (5.3)	<.001
Hypertension	10 467 (54.2)	31 735 (55.6)	<.001
Chronic pulmonary disease	2952 (15.3)	9065 (15.9)	.04
Diabetes without chronic complications	3415 (17.7)	10 342 (18.1)	.16
Diabetes with chronic complications	1498 (7.8)	4472 (7.8)	.71
Hypothyroidism	1224 (6.3)	3914 (6.9)	.01
Liver disease	1307 (6.8)	2866 (5.0)	<.001
Alcohol abuse	2739 (14.2)	5881 (10.3)	<.001
Drug abuse	2009 (10.4)	4034 (7.1)	<.001
Depression	4270 (22.1)	10 212 (17.9)	<.001

Abbreviation: VA, US Department of Veterans Affairs.

^aP values for unpaired, 2-tailed t tests (continuous variables) and χ² tests (categorical variables).

^bMedically complex patients were defined as having an Elixhauser comorbidity score in the top 10% of the cohort. The Elixhauser comorbidity score is composed of 29 binary comorbidities based on International Classification of Diseases diagnosis codes found in administrative data. Each comorbidity is weighted, with higher weights indicating a greater association with death. In the present study, the score range was –4 to 140.

Diabetes quality-of-care measures were similar (Table 2), but patients of residents were more likely to have appropriate renal testing (difference, 3.2 percentage points; P = .001). Patients of residents were less likely to have controlled hypertension (difference, 2.9 percentage points; P = .02), but older patients of residents were less likely to be prescribed high-risk medications (difference, 4.2 percentage points; P < .001). Patients of residents also were slightly more likely to have at least 1 emergency department visit in 2014 (difference, 1.3 percentage points; P < .001).

Table 2. Probability of Patients Cared for by Internal Medicine Residents or Attending Physicians in VA Teaching Primary Care Clinics Meeting Specific Quality Measures or Using Health Care Services in 2014^a.

Measure	Probability (95% CI) by Physician Type		Difference	P Value
	Resident	Attending
Annual HbA1c examination	0.97 (0.96 to 0.98)	0.97 (0.96 to 0.98)	0.002 (−0.007 to 0.011)	.66
HbA1c poor controlb	0.23 (0.20 to 0.25)	0.22 (0.20 to 0.24)	0.005 (−0.015 to 0.025)	.62
Annual renal test	0.85 (0.82 to 0.87)	0.82 (0.79 to 0.84)	0.032 (0.013 to 0.050)	.001
Hypertension control	0.60 (0.57 to 0.64)	0.63 (0.60 to 0.67)	−0.029 (−0.054 to −0.003)	.02
Use of high-risk medication in adults >65 y	0.23 (0.22 to 0.25)	0.27 (0.26 to 0.29)	−0.042 (−0.053 to −0.032)	<.001
ED visit	0.35 (0.32 to 0.38)	0.34 (0.30 to 0.37)	0.013 (0.005 to 0.021)	<.001
Hospitalization	0.09 (0.08 to 0.10)	0.08 (0.08 to 0.09)	0.001 (−0.004 to 0.005)	.74
ACSC hospitalization	0.03 (0.02 to 0.03)	0.03 (0.02 to 0.03)	0.001 (−0.002 to 0.003)	.84

Abbreviations: ACSC, ambulatory care–sensitive condition; ED, emergency department; HbA_1c, hemoglobin A_1c; VA, US Department of Veterans Affairs.

^aEstimated probabilities from logistic mixed models with adjustment for age, sex, race/ethnicity, Elixhauser comorbidity score, years of VA care, and a random effect for site.

^bPoor HbA_1c control was defined as an HbA_1c concentration that was unmeasured or greater than 9% in a year (to convert HbA_1c to a proportion of total hemoglobin, multiply by 0.01).

Discussion

Although studies have shown that primary care provided by residents is similar to that provided by attending physicians, earlier studies have been limited to small samples⁴ and measures based on self-reported data.⁵ In this large, national study using electronic health record–based measures, we found that residents provide access for new patients and play an important role in caring for vulnerable patient groups with complex care needs.

Quality-of-care measures were similar between patients of resident physicians and patients of attending physicians, and absolute differences were small. Residents’ better performance on some measures could be a product of training in evidence-based medicine and the use of clinical guidelines, or precepting, and supervision from attending physicians.

Limitations of this work include the VA-specific context, analysis of only VA data, and use of only 10 sites. In addition, unmeasured differences between patients of resident and attending physicians, such as differences in frailty or behavioral health history, could cause residual confounding and bias our results against clinicians who care for more complex patients.

Current measures of primary care quality are limited and typically focus on disease-specific process measures and short-term outcomes.⁶ However, according to these measures, residents appear to be providing near-equivalent care to attending physicians. More investigation is needed to understand how residents perform more complex primary care functions, such as integrating and prioritizing patient needs and delivering coordinated, whole-person, relationship-based care.