Factors Associated With Seeking Physician Care by Medicare Beneficiaries Who Receive All Their Primary Care From Nurse Practitioners

Abstract

Background: A shortage of primary care physicians has led to the alternative strategy of nurse practitioners (NPs) as primary care providers for the growing elderly population. Many states have implemented policies that allow NPs to practice independently with no physician oversight. Little is known about the continuity of primary care provided by NPs. Objective: To examine rate and correlates of switching from exclusive NP primary care to receiving some or all primary care from physicians. Design: A retrospective cohort study. Participants: Medicare beneficiaries (n = 38 618) with diabetes, congestive heart failure, or chronic obstructive pulmonary disease who received all their primary care from NPs in 2007. Main Measures: Multivariable logistic regression model was used to assess patient and disease characteristics associated with switching from sole NP primary care in 2007 to receiving some or all primary care from physicians between 2008 and 2010. Results: Of elderly patients receiving all their primary care from NPs in 2007, 53.8% switched to receiving some or all primary care from physicians in 2008-2010. The switching patients had less comorbidity before the switch and were more likely to reside in metropolitan areas, ZIP code areas with high education or states with the most restriction on NP scope of practice. In multivariable analyses, significant predictors of switching included one of the following within 30 days before the switch: emergency room visits (odds ratio [OR] = 1.55, 95% confidence interval [CI] = 1.44-1.68), hospitalization (OR = 1.13, 95% CI = 1.02-1.25), new diagnosis of heart attacks (OR = 5.52, 95% CI = 4.33-7.02), pneumonia (OR = 4.84, 95% CI = 3.71-6.32), atrial fibrillation (OR = 3.99, 95% CI = 2.93-5.44), stroke (OR = 2.94, 95% CI = 2.31-3.74), or cancer (OR = 2.65, 95% CI = 1.94-3.63). Conclusions: About half of Medicare patients under exclusive NP primary care switched to physicians for some or all primary care over a 3-year period. Future study is needed to understand the reasons for switching.

Introduction

Having one physician as a usual source of primary care is key to effective primary care of older patients with multiple coexisting diseases that require care coordination across multiple health care settings.^1-5 In a study of adult primary care network of 181 primary care physicians (PCPs), patients who did not identify one physician as their usual source of care were less likely to receive guideline-recommended diabetes care and cancer screenings.³ Older patients who see different PCPs at each visit are more likely to experience medical errors and potentially avoidable acute care visits, and have high health care cost.^2,6-10 To reduce such care fragmentation in older patients, the Affordable Care Act (ACA) authorized implementation of patient-centered medical home delivery models and other alternative health care models.^11-13 Success for these models depends on long-term continuity of primary care and a stable patient-clinician relationship.

A major challenge to long-term continuity of primary care is the growing shortage of PCPs; nurse practitioners (NPs) are increasingly being used to address this shortage.^11,14,15-22 Between 1998 and 2010, the proportion of Medicare recipients billed for outpatient care by NPs increased 9.5%, with the greatest increase in states with no restriction on NP practice.¹⁵ Some states such as Arizona and Maryland have policies that allow NPs to practice independently with no physician oversight.¹⁶ In addressing the shortage of PCPs, the federal government under ACA provided $50 million per year (from 2012 to 2015) to expand the training of NPs in primary care.¹⁷ Several studies have examined continuity of care among PCPs,^3-5,23 but there have been no population-based studies of long-term continuity of primary care among older patients receiving all their primary care from NPs. Understanding factors associated with longitudinal continuity of primary care in this population is critically important in improving quality and decreasing cost of care. The present study examined rates and correlates of switching from exclusive NP primary care to receiving care from PCPs between 2008 and 2010 among Medicare beneficiaries with diabetes, congestive heart failure (CHF), or chronic obstructive pulmonary disease (COPD) who received all their primary care from NPs in 2007.

Methods

Establishment of the Cohort Study

We identified all Medicare beneficiaries with diabetes, CHF, or COPD in 2006-2007 from the CMS Chronic Disease Data Warehouse (CCW). To select the NP cohort, we identified all patients aged 66 years or older who received all of their primary care from NPs in 2007. We identified the patients through bill records for 2 or more outpatient evaluation and management (E&M) services by NPs and with no outpatient E&M services from MDs (general practitioner, family physician, general internist, or geriatrician). We excluded patients whose original entitlements were based on disability or end-stage renal disease. Also excluded were patients who stayed in a nursing home in 2007 and those without continuous enrollment in parts A and B or covered in a health maintenance organization (HMO) in 2006 and 2007.

Study Outcome

Patients were followed up from 2008 to identify discontinuity of primary care from NPs. Discontinuity of care was defined as a switch from exclusive NP primary care to receiving any care from a PCP any time between 2008 and 2010. Patients were censored at death, loss of coverage or the end of the study (December 31, 2010).

Measures

The demographic information on patients’ age, gender, and race/ethnicity was obtained from Medicare enrollment files. We used a Medicaid indication in the enrollment file as a proxy for low socioeconomic status. Education for ZIP code areas was obtained from the 2010 Census data and categorized by quartiles. The Elixhauser comorbidity measures were generated from inpatient and outpatient claims in 2007 and the comorbidity sum for patients were the summation of all Elixhauser comorbidity measures excluding CHF, CPOD, and diabetes with or without chronic complications. Number of provider visits; whether patients saw a cardiology, endocrinology or pulmonary specialist; and number of hospitalizations in 2007 were generated. The size of residential area was categorized using Rural-Urban Continuum Codes that distinguish metropolitan counties by size and nonmetropolitan counties by degree of urbanization and proximity to metropolitan areas. State regulations of NP practice were classified into 5 categories from least to most restricted.¹⁵

To further explore potential factors associated with discontinuity of primary care, we examined rates of acute hospitalization or an emergency room (ER) visit in the 30 days prior to switching to an MD primary care. We also assessed whether patients had a new diagnosis of major health events in the 30 days prior to care switch. The incident events were cancer (International Classification of Diseases, Ninth Revision [ICD-9]: 140-239), stroke (ICD-9: 430-437), heart attack (ICD-9: 410-414), hip fracture (ICD-9: 820), atrial fibrillation (ICD-9: 427.3) and pneumonia (ICD-9: 480-486). The incident event was defined by a primary diagnosis in either inpatient or outpatient claims for the event and no such event in the previous 12 months. To ensure that the follow-up time was comparable to the 30-day look-back period between patients who switched to physician care (the switch group) and those who stayed with NP care (stay group), we randomly assigned the follow-up time to those in the stay group based on the distribution of follow-up time from the switch group.

Statistical Analyses

Using t test for continuous variables and chi-square test for categorical variables, we compared characteristics of patients who continuously received all primary care from NPs to those who switched to receiving some or all primary care from physicians. To examine potential factors associated with switching from sole NP primary care to physician primary care, we used a multivariable logistic regression model to evaluate the independent relationships between patient characteristics and the odds of switching to MDs for primary care. All tests of statistical significance were 2-sided (P < .05). Analyses were performed with SAS version 9.3 (SAS Institute, Cary, NC).

Results

The study cohort included 38 618 Medicare beneficiaries with CHF, COPD, or diabetes who received primary care from NPs in 2007. Table 1 shows the characteristics of the beneficiaries grouped according to whether they switch care from sole NP to any MD (switch group) or stayed with NP care (stay group) any time between 2008 and 2010. Of the patients receiving NP care, 53.8% switched to MDs for their primary care. Patients in the switch group were more likely to be female, younger, reside in metro areas, reside in ZIP codes with higher education, and reside in states with most restrictions on the scope of NP practice. They were also more likely to have diabetes or liver disease, and have endocrinology specialty consult. They were less likely to be black and impoverished. The switch group had fewer comorbidities and prior hospitalizations than the stay group, but more NP provider visits in the previous year. The switch group participants were less likely to have CHF, COPD, pulmonary circulation disease, peripheral vascular disease, neurological disorder, renal failure, AIDS, anemia, peptic ulcer, weight loss, and cancer.

Table 1.

Characteristics of Medicare Patients Who Received All Their Primary Care From Nurse Practitioners (NPs) in 2007 Stratified According to Whether the Patients Received Some or All of Their Care From Primary Care Physicians or Continue With NP Primary Care Between 2008 and 2010.

Variables	n		Switch to Physicians (%)	Stay With NPs (%)	P
N	38 618		20 766	17 852
%			53.8	46.2
Age in years
66-74	15 477		54.8	45.2	<.0001
75-84	17 190		54.4	45.6
85+	5951		49.3	50.7
Gender
Female	24 144		55.3	46.7	<.0001
Male	14 474		51.2	48.8
Race
Whites	33 399		53.6	46.4	<.0001
Blacks	2793		51.9	48.1
Hispanics	1414		54.5	45.5
Other	1012		62.5	37.5
Medicaid eligibility
No	31 577		54.6	45.4	<.0001
Yes	7 041		50.2	49.8
Proportion of education >12 years
Q1 (lowest)	9539		51.9	48.1	<.0001
Q2	9524		53.0	47.0
Q3	9345		53.4	46.6
Q4 (highest)	9587		56.6	43.4
Rural/urban
Metro	23 059		56.8	43.2	<.0001
Nonmetro urban	13 148		49.8	50.2
Rural	2411		46.1	53.9
State regulation
1 (least restrictive)	6484		53.8	46.2	<.0001
2	6337		53.0	47.0
3	10 418		52.4	47.6
4	8718		54.0	46.0
5 (most restrictive)	6661		56.3	43.7
Prior hospitalization
0	28 007		54.6	45.4	<.0001
1	6764		52.6	47.4
2	2348		51.7	48.3
3+	1499		46.7	53.3
Previous provider visit			9.9 ± 7.2	8.9 ± 7.2	<.0001
Specialty visit
Cardiology	Yes	13 604	53.9	46.1	.75
	No	25 014	53.7	46.3
Pulmonary	Yes	4346	53.4	46.6	.58
	No	34 272	53.8	46.2
Endocrinology	Yes	2658	58.9	41.1	<.0001
	No	35 960	53.4	46.4
Conditions
Diabetes	Yes	26 055	54.6	45.4	<.0001
	No	12 563	52.1	47.9
Congestive heart failure	Yes	12 860	51.2	48.8	<.0001
	No	25 758	55.1	44.9
COPD	Yes	7600	51.4	48.6	<.0001
	No	31 018	54.4	45.6
Number of conditions
1	31 542		54.6	45.4	<.0001
2	6255		50.7	49.3
3	821		46.5	53.5
Comorbidity sum
0	7130		54.7	45.3	<.0001
1	14 551		54.5	45.5
2	8340		54.4	45.6
3+	8597		51.2	48.8
Comorbidity
Valvular disease	Yes	3216	52.7	47.3	.19
	No	35 402	53.9	46.1
Pulmonary circulation disease	Yes	828	45.1	54.9	<.0001
	No	37 790	54.0	46.0
Hypertension	Yes	27 103	53.9	46.1	.61
	No	11 515	53.6	46.4
Peripheral vascular disease	Yes	3735	50.2	49.8	<.0001
	No	34 883	54.1	45.9
Paralysis	Yes	281	50.2	49.8	.23
	No	38 337	53.8	46.2
Neurological disorders	Yes	1264	44.9	55.1	<.0001
	No	37 354	54.1	45.9
Hypothyroidism	Yes	4694	54.8	45.2	.12
	No	33 924	53.6	46.4
Renal failure	Yes	3086	51.7	48.3	.02
	No	35 532	53.9	46.1
Liver disease	Yes	232	60.8	39.2	.03
	No	38 386	53.7	46.3
Peptic ulcer	Yes	21	33.3	66.7	.06
	No	38 597	53.8	46.2
Acquired immune deficiency syndrome	Yes	15	26.7	73.3	.04
	No	38 603	53.8	46.2
Rheumatoid arthritis	Yes	1005	55.0	45.0	.42
	No	37 613	53.7	46.3
Coagulopthy	Yes	920	52.5	47.5	.43
	No	37 698	53.8	46.2
Obesity	Yes	1216	54.0	46.0	.86
	No	37 402	53.8	46.2
Weight loss	Yes	609	48.4	51.6	.01
	No	38 009	53.9	46.1
Fluid and electrolyte disorders	Yes	3414	50.5	49.5	<.0001
	No	35 204	54.1	45.9
Chronic blood loss anemia	Yes	438	48.6	51.4	.03
	No	38 180	53.8	46.2
Lymphoma	Yes	387	52.2	47.8	.53
	No	38 231	53.8	46.2
Metastatic cancer	Yes	401	41.6	58.4	<.0001
	No	38 217	53.9	46.1
Solid tumor without metastasis	Yes	3250	50.3	49.7	<.0001
	No	35 368	54.1	45.9
Deficiency anemias	Yes	5458	50.9	49.1	<.0001
	No	33 160	54.2	45.8
Alcohol abuse	Yes	178	57.9	42.1	.24
	No	38 440	53.7	46.3
Drug abuse	Yes	39	48.7	51.3	.53
	No	38 579	53.8	46.2
Psychoses	Yes	810	52.8	47.2	.59
	No	37 808	53.8	46.2
Depression	Yes	1720	52.7	47.3	.38
	No	36 898	53.8	46.2

Table 2 shows the rates of acute care events and new diagnoses in the 30 days before switch in the switch group compared with rates in the stay group. Patients in the switch group were more likely to have experienced hospitalization (8% vs 4.5%) and ER visits (13.8% vs 6.5%) in the 30 days before they switched to MD care. In the 30 days before they switched, they were also more likely to have been newly diagnosed with cancer, stroke, heart attack, atrial fibrillation, and pneumonia.

Table 2.

Hospitalization and Emergency Room (ER) Visits and New Diagnoses in Medicare Patients 30 Days Before the Patients Switched From Sole Nurse Practitioner (NP) Primary Care to Receiving Any Care From a Primary Care Medical Doctor (MD), 2008-2010.

Variables	Switch to MD	Stay With NPa	P
N	20 766	17 852
Length of follow-up months (mean ± SD)	12.3 ± 9.9	10.8 ± 9.8
Hospitalization and ER (%)
Hospitalization	8.0	4.5	<.0001
Preventable hospitalization	2.4	1.1	<.0001
Nonpreventable hospitalization	5.9	3.5	<.0001
ER visits	13.8	6.5	<.0001
Specific disease (new case) (%)
Principal diagnosis
Cancer (excluding skin cancer)	0.86	0.30	<.0001
Stroke	1.80	0.50	<.0001
Heart attack	2.56	0.45	<.0001
Hip fracture	0.13	0.11	.42
Atrial fibrillation	1.35	0.29	<.0001
Pneumonia	2.12	0.37	<.0001
Pneumonia (excluding 481-484.6)	2.08	0.36	<.0001
Any position on diagnosis
Cancer (excluding skin cancer)	1.24	0.38	<.0001
Stroke	2.53	0.66	<.0001
Heart attack	3.38	0.69	<.0001
Hip fracture	0.16	0.12	.28
Atrial fibrillation	1.71	0.41	<.0001
Pneumonia	2.60	0.56	<.0001

^aTo ensure that the follow-up time was comparable to the 30-day look-back period between patients who switched to physician care (the switch group) and those who stayed with NP care (stay group), we randomly assigned the follow-up time to those in the stay group based on the distribution of follow-up time from the switch group.

Table 3 presents the associations between patient characteristics and odds of switching from NP primary care to MD primary care. In the multivariable analyses, significant predictors of switching from NP care to MD primary care include female gender, not being on Medicaid, high education, residence in a metro area or in a state with most restrictions on NP practice, fewer prior hospitalization and comorbidities, and frequent clinic visits to the NP. Other significant predictors of switching from NP to MD care included ER visits (odds ratio [OR] = 1.55, 95% confidence interval (CI) = 1.44-1.68) or hospitalization (OR = 1.13, CI = 1.02-1.25) in the 30 days prior to switch. Odds of switching from NP to MD also increased significantly after patients were newly diagnosed with heart attacks (OR = 5.52, 95% CI = 4.33-7.02), pneumonia (OR = 4.84, 95% CI = 3.71-6.32), atrial fibrillation (OR = 3.99, 95% CI = 2.93-5.44), stroke (OR = 2.94, 95% CI = 2.31-3.74), or cancer (OR = 2.65, 95% CI = 1.94-3.63). About 8% of variation in switching care from NP to MD was explained by the characteristics listed in Table 3.

Table 3.

Adjusted Odds Ratio (OR) From a Multivariable Analysis Estimating the Odds of Switching From Receiving All Primary Care From Nurse Practitioners in 2007 to Receiving Some or All Care From Primary Care Physicians Between 2008 and 2010.

	OR	95% CI
		Lower Limit	Upper Limit
Age in years
66-74	Reference
75-84	0.95	0.91	1.00
85+	0.75	0.71	0.80
Female	1.24	1.18	1.29
Race
White	Reference
Black	0.99	0.91	1.08
Hispanic	1.08	0.97	1.22
Other	1.61	1.40	1.84
Medicaid eligibility	0.83	0.78	0.88
Proportion of education >12 years
Q1 (lowest)	Reference
Q2	1.02	0.96	1.08
Q3	1.02	0.96	1.09
Q4 (highest)	1.13	1.06	1.20
Rural/urban
Metro	Reference
Nonmetro urban	0.77	0.74	0.81
Rural	0.67	0.62	0.73
State regulation
1 (least restrictive)	Reference
2	1.00	0.93	1.07
3	1.06	1.00	1.14
4	1.10	1.02	1.18
5 (most restrictive)	1.19	1.11	1.28
Prior hospitalization
0	Reference
1	0.97	0.91	1.03
2	0.94	0.86	1.04
3+	0.77	0.69	0.87
Previous provider visit	1.03	1.02	1.03
Specialty visit
Cardiology	0.95	0.91	0.99
Pulmonary	0.87	0.81	0.93
Endocrinology	1.08	0.99	1.17
No. of conditions
1	Reference
2	0.85	0.80	0.90
3	0.75	0.64	0.87
Comorbidity sum
0	Reference
1	0.94	0.89	1.00
2	0.91	0.85	0.97
3+	0.77	0.71	0.84
Hospitalization*	1.13	1.02	1.25
Emergency room visit*	1.55	1.44	1.68
Principal diagnosis*
Cancer	2.65	1.94	3.63
Stroke	2.94	2.31	3.74
Heart attack	5.52	4.33	7.02
Hip fracture	0.81	0.43	1.52
Atrial fibrillation	3.99	2.93	5.44
Pneumonia	4.84	3.71	6.32

^*In the 30 days prior to the switch.

Discussion

About half of Medicare beneficiaries with NPs as their sole primary care provider in 2007 switched to receive some or all of their primary care from physicians between 2008 and 2010. The switch group patients were more likely to have been hospitalized or visited ER within 30 days prior to the switching, relative to the stay group patients. The switch group patients were also more likely than those in the stay group to have been newly diagnosed with cancer, stroke, heart attack, atrial fibrillation, and pneumonia in the 30 days before they switched. More Medicaid-eligible patients stayed with their NP primary care providers compared with non-Medicaid eligible patients; this may reflect the reluctance of some physicians to take on Medicaid patients for primary care.²⁴

Our finding of higher NP-to-MD switching in states with the most restrictions of NP practice is consistent with prior findings of lower likelihood of NP primary care when NP practice is restricted.¹⁵ The restrictions ranged from limitations on admitting patients to hospital, making referrals, ordering diagnostic tests to prescribing medications.¹⁵ Our study showed that NP-to-MD switching occurred more frequently after a patient experienced a hospitalization or ER visit. These acute care experiences, especially in the context of high NP practice restriction, may alter the patient perception of the ability of the NP provider to facilitate future tests, medications or hospital/ER transfers. A future qualitative study is needed to examine relationship between degree of NP practice restriction and domains of patient-perceived continuity of care such as relational and informational continuity (interpersonal trust and knowledge) that are not captured in quantitative study like ours.^23,25,26-31

Some of the patients who switched might still be cared for by NPs under a team model of care where NPs and MDs work together to care for older patients with multiple chronic diseases.^27-34 The NP-MD team model may best serve the needs of the switch group patients whose health conditions have become more complex following their recent hospitalizations or new diagnoses. Evidence showed that patients with multiple comorbidities experienced better outcomes at lower cost under a collaborative model of NPs and MDs.^32-36

One potential explanation for the switching after hospitalization or ER visits may be related to physician perceptions of quality of NP clinical skills. These perceptions may implicitly or explicitly encourage these hospitalized patients to switch to primary care physicians.^37-39 The number of NPs generally is increasing,¹⁵ so a decreasing population of NPs does not explain the high switch rate. Patients who switch after ER visits or hospitalizations may also switch because they feel they need to seek physician involvement after their acute care visits or new diagnoses, given that many of these visits are to emergency specialties.

There are several limitations to our study. Because we studied only fee-for-service Medicare patients, our findings may not be generalizable to younger populations or patients with commercial insurance. We may also not have captured all patients who received all their primary care from NPs. This possibility could arise in a group medical practice when the physician rather than the NP submits the higher billing charge under the MD name for care given by the NP. We also do not have data on quality of communication and patient trust, factors that are critical to longevity of the clinician-patient relationship.^23,25,26,40 Qualitative interviewing of NPs, physicians and patients may help clarify the perceived added value of physician primary care versus NP primary care versus team care, implicit and explicit biases in the health care system, and organizational and system barriers impeding continuity of care. ^27-31,41-46

Author Biographies

Michelle Y. Raji is a rising junior at Harvard University, Cambridge, Massachusetts where she studies Global Health, Health Policy, and English. She was a 2015 summer research intern in the Department of Preventive Medicine and Community Health at the University of Texas Medical Branch (UTMB), Galveston.

Nai-Wei Chen, PhD, is a biostatistician at the UTMB Department of Preventive Medicine and Community Health. His areas of expertise include analysis of Medicare and other large claims data, generalized linear mixed models, categorical data analysis, and longitudinal studies.

Mukaila Raji, MD MS, FACP, is Edgar Gnitzinger Distinguished professor of Aging and director of the UTMB Division of Geriatric Medicine. He is a board-certified geriatrician with over 15 years’ experience in comparative drug effectiveness and toxicity research, cognitive aging and disability, and health policy and outcomes.

Yong-Fang Kuo, PhD, is professor of Preventive Medicine and Community Health, and director of the UTMB Office of Biostatistics. She has over 15 years’ experience in health policy and outcomes, comparative effectiveness and toxicity research, hierarchical generalized linear mix modelling, and instrumental variable analysis.