Medicare beneficiaries with more comprehensive primary care physicians report better primary care

Abstract

Objective

To examine whether primary care physician (PCP) comprehensiveness is associated with Medicare beneficiaries' overall rating of care from their PCP and staff.

Data Sources

We linked Medicare claims with survey data from Medicare beneficiaries attributed to Comprehensive Primary Care Plus (CPC+) physicians and practices.

Study Design

We performed regression analyses of the associations between two claims‐based measures of PCP comprehensiveness in 2017 and beneficiaries' rating of care from their PCP and practice staff in 2018.

Data Collection/Extraction Methods

The analytic sample included 6228 beneficiaries cared for by 3898 PCPs. Regressions controlled for beneficiary, physician, practice, and market characteristics.

Principal Findings

Beneficiaries with more comprehensive PCPs rated care from their PCP and practice staff higher than did those with less comprehensive PCPs. For each comprehensiveness measure, beneficiaries whose PCP was in the 75th percentile were more likely than beneficiaries whose PCP was in the 25th percentile to rate their care highly (2 percentage point difference, p = 0.02).

Conclusions

Medicare beneficiaries with more comprehensive PCPs rate overall care from their PCPs and staff higher than those with less comprehensive PCPs.

What is known on this topic

• Greater primary care comprehensiveness has been associated with reductions in patient hospitalizations, emergency department visits, total expenditures, and duplicative service use—and with improvements in overall health and equity.

• Less is known about whether primary care physician (PCP) comprehensiveness, measured using claims, is associated with patient‐reported outcomes, namely patients' rating of care from their PCP and practice staff.

• Claims‐based measures of PCP comprehensiveness avoid potential bias in patient reports, which can reflect patient fondness for a PCP or different expectations of what the PCP can deliver.

What this study adds

• We explore whether PCP comprehensiveness (measured via claims data) is associated with Medicare fee‐for‐service beneficiaries' ratings of overall care from their PCP and practice staff.

• Medicare beneficiaries with more comprehensive PCPs rate overall care received from their PCP and staff at the practice more highly than those with less comprehensive PCPs.

1. INTRODUCTION

Primary care is essential to a well‐organized, efficient, high‐quality, and equitable health care delivery system—and comprehensiveness is a major feature of primary care. ¹ , ² , ³ , ⁴ , ⁵ Comprehensiveness is defined as how much a primary care provider (clinician, practice, or team) recognizes and meets most of a patient's physical and common behavioral health care needs. ¹ , ² , ³ , ⁴ , ⁵ A comprehensive primary care physician (PCP) works with patients to manage the depth and breadth of their conditions, and to recognize and manage those new problems that are within their training competencies. ¹ , ⁶ , ⁷ , ⁸ Without a comprehensive PCP, patients experience fragmented care from different specialists, resulting in higher health care costs, more diagnostic tests and interventions, and more medications. ¹ , ⁹ , ¹⁰ , ¹¹

PCP and practice comprehensiveness are associated with reductions in patient hospitalizations, ED visits, total expenditures, ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² and duplicative service use—and with improvements in overall health and equity. ¹³ , ¹⁴ , ¹⁵ Less is known about whether PCP comprehensiveness is associated with patient‐reported outcomes, specifically, patients' overall rating of care from their PCP and practice staff. Using patient self‐reports to assess both PCP comprehensiveness and overall primary care experience may be biased because patients may give clinicians high ratings on comprehensiveness, as they do with factors such as trust, partly because the clinician does things the patient values or they have a longstanding relationship. ¹⁶ , ¹⁷ Patients' ratings may also be influenced by their beliefs about the relative merits of PCPs versus specialists and the prevailing views in their community, ¹⁸ , ¹⁹ rather than whether the clinicians provide the professionally appropriate level of comprehensiveness. ²⁰ Furthermore, no patient or group of patients in a practice experience the full range of symptoms, problems, and conditions needed to assess their PCP's skills related to comprehensiveness.

Thus, using measures of PCP comprehensiveness from an objective source (i.e., claims data) may help avoid this bias. ⁸ To this end, we use data from the evaluation of the Comprehensive Primary Care Plus (CPC+) model, ²¹ the largest national effort to date to support and strengthen primary care. CPC+ data include claims, a survey of Medicare fee‐for‐service (FFS) beneficiaries, and information on a wide range of beneficiary, physician, practice, and market characteristics. In addition, despite the demonstrated benefits of PCP comprehensiveness to Medicare beneficiaries' care, ⁶ , ⁷ , ⁸ , ⁹ it is unclear whether beneficiaries will value their PCPs providing more comprehensive care. Patients often equate access to more care with higher‐quality care, ²² and recent evidence suggests that Medicare FFS beneficiaries' appetite for accessing specialized care has increased over time. ²³ , ²⁴ For both of these reasons, we explore whether PCP comprehensiveness is associated with Medicare beneficiaries' rating of overall care from their PCP and staff.

2. METHODS

2.1. Overview

This work was conducted in the context of the CPC+ model ²¹ launched in 2017 by the Centers for Medicare & Medicaid Services (CMS). To participate in CPC+, a practice site (the single physical location where patients are served) needed to treat at least 125 Medicare FFS beneficiaries and have at least 1 primary care practitioner. We constructed two physician‐level comprehensiveness measures for PCPs in CPC+ practices in 2017, using Medicare FFS claims of all beneficiaries they had seen, as described in prior publications. ⁶ , ⁹ We linked beneficiaries in the CPC+ Medicare FFS Beneficiary Survey to physicians to whom they were attributed, based on National Provider Identifiers (NPIs) in Medicare claims. We attributed beneficiaries to the PCP who provided the largest share of their care (described in the Appendix S1). We then calculated the association of the physician‐level comprehensiveness measures in 2017 with beneficiary‐level survey responses in 2018, controlling for various beneficiary‐, physician‐, practice‐, and market‐level characteristics.

2.2. Data sources

The CPC+ Beneficiary Survey was administered to a random sample of Medicare FFS beneficiaries attributed to CPC+ practices in 2018. The 15‐ to 20‐min survey asked about beneficiaries' experiences with their PCPs and staff. The survey was sent by US mail, with no respondent incentive payment. Most survey items came from the standard Clinician and Group—Consumer Assessment of Health care Providers and Systems (CAHPS) and we followed the CAHPS fielding procedures. ²⁵ , ²⁶ The response rate (the number of eligible and complete survey responses divided by the eligible sample) was 42%. ²⁷ We weighted the analysis to account for differences between respondents and nonrespondents.

Of the 7915 beneficiaries in CPC+ practices who responded to the 2018 CPC+ Beneficiary Survey, we excluded beneficiaries if they were not attributed to a PCP in our sample (12.8%), their attributed PCP did not have comprehensiveness data (0.7%), or they were missing model covariates (3.7%). In addition, 4.8% of respondents did not answer the survey question used in this analysis. Our final sample included 6228 beneficiaries cared for by 3898 PCPs within 2111 CPC+ practices. Beneficiaries in our final sample did not differ from all respondents in personal, physician, practice or market characteristics nor in their mean ratings of the care received from their PCP and staff at their primary care practice.

We used Medicare FFS claims from the CMS Virtual Research Data Center's Research Identifiable Files to construct our measures of physician comprehensiveness and the Medicare Enrollment Database and beneficiary summary files for beneficiary characteristics. Other beneficiary demographic data came from the CPC+ Beneficiary Survey. Data for our physician‐, practice‐ and market‐level control variables came from various sources including CPC+ practices' applications to CMS, the HRSA Area Health Resources file, SK&A (an IQVIA database), the Medicare Data on Provider Practice and Specialty (MD‐PPAS), and the National Committee for Quality Assurance (see Table S3).

2.3. Measures of primary care physician comprehensiveness

We identified PCPs in CPC+ practices in 2017 using the following primary MD‐PPAS specialty codes: 11 (internal medicine), 08 (family practice), 38 (geriatric medicine), or 01 (general practice). We excluded hospitalists. We excluded nurse practitioners (NPs) and physician assistants (PAs), because our sample had a low prevalence of NPs and PAs serving as a patient's usual practitioner, and it was not feasible to discern all services NPs/PAs independently provided because they commonly bill “incident to” services under a physician's NPI.

We analyzed two previously validated claims‐based measures of PCP comprehensiveness: (1) involvement in patient conditions and (2) new problem management. For each measure, we calculated a physician‐specific score using the Medicare FFS claims for all beneficiaries the physician treated in 2017. Details on each measure have been published ⁶ , ⁹ and are in the Appendix S1. Briefly, the involvement in patient conditions measure defines comprehensive physicians as those involved in caring for the broad range of their patients' health conditions reflected in visit‐related International Classification of Disease (ICD‐10) codes in a calendar year. To calculate this measure, we examined all of the beneficiaries' evaluation and management (E&M) claims to both primary care and specialist physicians. For each physician, we calculated the percentage of patients seen that calendar year for whom the physician was most involved with the patient's conditions. “Most comprehensive” for this measure means the physician saw the patient for the plurality of their unique diagnosis codes. This percentage is represented as a score ranging from 0 to 1.

The new problem management measure assesses the extent to which a PCP manages a patient's new condition instead of referring them to (or the patient seeking) another provider. Comprehensive PCPs should be able to address their patients' common health problems, ¹ defined as the 20 most common reasons for primary care visits among people age 65 and over. We identified patients with an office‐based E&M claim for one of those conditions but no E&M claims for that problem in the prior 24 months having a “new problem.” For each patient–new problem combination, we calculated a patient score as the proportion of visits to the first (index) physician out of all office‐based E&M visits for that problem within 1 year of the first visit. For each index physician, we calculated a physician score as the average of their patients' scores for each new problem. For each condition, we calculated an “expected score” for each condition as the average physician score across all physicians who treated the condition. Next, we calculated each physician's observed score (and the numerator for the measure) as the average patient score across all new problems for each index physician. Finally, we substituted each condition's “expected score” for each patient score and averaged this for each physician as the denominator for this measure. The final physician score was the physician's observed score divided by their average expected score. A score above 1 indicates the physician was more comprehensive than the average physician who saw the same conditions, whereas a score below 1 indicates the physician was less comprehensive than other physicians who saw the same conditions. These two measures varied substantially across our PCP sample (Table 1). Additional details on each measure's interpretation are in the Appendix S1.

TABLE 1.

Mean and percentile distribution of the two measures of PCP comprehensiveness in the analytic sample

Comprehensiveness Measures	Mean	S.D	Min	5th percentile	25th percentile	50th percentile	75th percentile	95th percentile	Max
Involvement in patient conditions	0.72	0.12	0.15	0.51	0.66	0.74	0.81	0.88	1.00
New problem management	1.01	0.06	0.56	0.91	0.99	1.02	1.05	1.08	1.28

Abbreviations: PCP, primary care physician; SD, standard deviation.

2.4. Beneficiary experience outcome

We calculated patients' ratings of their PCP and staff among beneficiaries who were attributed to the physicians and who participated in the CPC+ Beneficiary Survey in 2018.

The outcome measure is beneficiaries' overall rating of their PCP and staff at the primary care practice. This CAHPS item asks: Using any number from 0 to 10, where 0 is the worst care possible and 10 is the best care possible, what number would you use to rate the care you have received from the primary care doctors and their staff from this doctor's office? We categorized the responses into two levels: 0–8 (15% of beneficiaries) and 9–10 (85%) based on the CAHPS creators' guidance. ²⁵ , ²⁶ Results were similar from a sensitivity test classifying the PCP/staff rating as 0–8 versus 9 versus 10.

2.5. Control variables

Beneficiary‐level control variables were measured in 2016 before CPC+ began and included: beneficiaries' Hierarchical Condition Category (HCC) score and individual HCC condition‐specific indicators, age, sex, reason for Medicare eligibility, and dual eligibility for Medicare and Medicaid. We included race, ethnicity, education, and self‐rated health status from the 2018 beneficiary survey. We also adjusted for the extent to which a beneficiary had fragmented care ²⁸ , ²⁹ across all the providers (primary care and specialists) to ensure that any observed associations between the comprehensiveness of care and outcomes were not driven by variation in patients' inherent propensities to seek care from providers outside the primary care practice⁹ (details in the Appendix S1). We did not control for other aspects of patients' experiences from CAHPS in our models for reasons explained in the Appendix S1.

Physician‐level control variables included sex, age, specialty, and whether they were full‐ or part‐time.

Our models controlled for primary care practice characteristics including: CPC+ track, number of primary care clinicians (physicians, NPs, and PAs) in the practice, whether or not the practice had at least one NP or PA, ownership type (independent versus system‐ or hospital‐ owned), prior recognition as a medical home, primary care or multispecialty practice, Meaningful Use criteria met by 1+ practice physician, and whether the practice was enrolled in the Medicare Shared Savings Program. These variables account for resource availability, health care technology use, and a practice's ability to provide advanced primary care, which can affect comprehensiveness and patient outcomes.

The market characteristics of the practice included whether the practice was in a county health professional shortage area, median household income, Medicare Advantage penetration rate, percentage of the population living in poverty, percentage of adults with 4 years of college, urbanicity, number of hospital beds, Hospital Referral Region price index, and geographic region, to account for differences in physician practice patterns that can lead to geographic variation in comprehensiveness and outcomes.

2.6. Analysis and models

In CPC+, practice sites could apply to participate in one of two “tracks” offering different options to meet primary care practices' diverse needs. ²¹ PCPs' comprehensiveness scores and beneficiaries' characteristics and care ratings were comparable across the two tracks. Therefore, we combined beneficiaries in both CPC+ tracks.

To ensure PCPs' comprehensiveness was in place before measuring patient outcomes, we assessed physician‐level comprehensiveness in 2017 and beneficiary‐level survey responses in 2018. We tested for associations between each physician‐level comprehensiveness measure and the beneficiary‐level binary survey outcome using logistic regression. Each regression model controlled for the variables described above and accounted for clustering of patient outcomes within practices (to account for beneficiaries seeing the same physician and those seeing different physicians within the same practice). The unit of observation was the beneficiary; observations were weighted to account for survey nonresponse. Analyses were conducted using STATA 16.

We examine associations between physician comprehensiveness measures and beneficiary‐level outcomes, we report both the magnitude and percentage difference in the adjusted mean outcomes for an increase in the comprehensiveness score from the 25th to 75th percentile among all PCPs in the analysis.

3. RESULTS

Characteristics of Medicare beneficiaries included in this analysis are presented in Table 2. The beneficiary, physician, practice, and market characteristics were similar across tracks except for prior participation in a medical home initiative, which we controlled for in our regression models.

TABLE 2.

Selected beneficiary, primary care physician, practice, and market characteristics

Number of beneficiaries	N = 6228
	%
Beneficiary characteristics
% of beneficiaries (unadjusted) in each track who gave the highest rating of care from their PCP and staff at the practice	84.68
Age
Under 65	12.48
65–74	48.60
75–84	29.00
85 and over	9.93
Male	41.82
Education
Some high school or less	8.01
High school graduate or GED	31.15
At least some college	58.36
Race and Ethnicity
Black or African American (Non‐Hispanic)	4.22
Hispanic	2.03
White (Non‐Hispanic)	87.06
Other a	6.68
Dual eligibility	10.36
Original reason for entitlement
OASI	81.51
End‐stage renal disease	0.51
Disability	17.98
HCC score (normalized)	1.03
Self‐reported health status
Poor/fair	22.09
Good	36.43
Very good/excellent	40.87
Urbanicity
Rural	8.32
Suburban	16.84
Urban	74.83
Beneficiaries' primary care physician characteristics
Male	68.88
Age
<50	39.25
Over 50	60.75
Practice type
Family practice	52.88
Internal medicine	45.63
Geriatric medicine	0.80
General practice	0.54
PCP's comprehensiveness
Involvement in patient conditions (mean)	0.73
New problem management (mean)	1.01
Beneficiaries' practice characteristics
Number of primary care practitioners
1–2	16.53
3–5	33.50
6+	49.97
Owned or managed by a health system or owned by a hospital	56.83
Medical home recognition, MAPCP or CPC Classic	68.98
Market characteristics
Medicare advantage penetration rate 2015	30.40
Median household income 2014 ($)	57,614.13

Note: When categories for a variable do not add to 100%, it is due to the small number of beneficiaries in the “do not know” response, which we did not include in the table.

Abbreviations: CMS, centers for medicare & medicaid services; CPC, comprehensive primary care Initiative, CPC+, comprehensive primary care plus model; HCC, hierarchical condition category; MAPCP, multi‐payer advanced primary care practice; OASI, old age and survivor's insurance; PCP, primary care physician.

^a “Other” refers to any survey respondent who self‐identified as “Asian,” “Native Hawaiian or Other Pacific Islander,” “American Indian or Alaskan Native,” or “Other.”

Source: CPC+ 2018 Beneficiary Survey and Medicare fee‐for‐service claims from the CMS Virtual Research Data Center's Research Identifiable Files for data on beneficiaries and physicians. Data on practice and market characteristics come from a range of sources including CPC+ practices' application data to CMS, the Area Resource file, SK&A, the Medicare Data on Provider Practice and Specialty (MD‐PPAS), the Health Resources and Services Administration, and the National Committee for Quality Assurance.

After controlling for beneficiary‐, physician‐, practice‐, and market‐ characteristics, beneficiaries cared for by more comprehensive PCPs (measured by involvement in patient conditions and new problem management) rated the care received from their PCP and practice staff higher (Table 3). A greater percentage of beneficiaries (2.1 percentage point difference, p = 0.026) gave the highest rating for care from their PCP and staff if their PCP was at the 75th percentile versus the 25th percentile for involvement in patient conditions. Similarly, a greater percentage of beneficiaries (2.4 percentage point difference, p = 0.005) gave the highest rating for care from their PCP and staff if their PCP was at the 75th percentile versus the 25th percentile for new problem management. (Table S3 presents full regression model results.)

TABLE 3.

Associations between physician comprehensiveness and beneficiaries' higher rating of overall care received from their primary care physician and staff at the practice ^a

Primary care physician comprehensiveness measures	Estimated marginal effects of PCP comprehensiveness on beneficiaries' rating of their primary care	Difference in patient's rating of their primary care as a percentage of the mean at the 25th versus 75th percentile of PCP comprehensiveness	p‐value
	Coefficient (standard error)	%
Involvement in patient conditions	0.12 (0.05)*	2.12	0.026
New problem management	0.32 (0.11)**	2.41	0.005

Note: Each row summarizes a separate regression. All regressions control for patient, practitioner, practice, and market characteristics. Standard errors were clustered at the practice level and observations are weighted by non‐response weights. Marginal effects are shown in the table with their significance.

Abbreviations: CMS, centers for medicare & medicaid services; CPC+, comprehensive primary care plus; PCP, primary care physician.

^a Estimates present the likelihood of beneficiaries reporting a higher (score of 9 or 10) versus lower score (0–8).

^* p < 0.05;

^** p < 0.01;

Source: CPC+ 2018 Beneficiary Survey and Medicare fee‐for‐service claims from the CMS Virtual Research Data Center's Research Identifiable Files for data on beneficiaries and physicians. Data on practice and market characteristics, come from a range of sources including CPC+ practices' application data to CMS, the Area Resource file, SK&A, the Medicare Data on Provider Practice and Specialty (MD‐PPAS), the Health Resources and Services Administration, and the National Committee for Quality Assurance.

4. DISCUSSION

Medicare FFS beneficiaries with more comprehensive PCPs rated their primary care physician and staff higher than those with less comprehensive PCPs. Prior studies show that PCP comprehensiveness is associated with lower Medicare FFS hospitalizations, ED use, and expenditures ⁶ , ⁹ ; this study adds an important patient‐reported outcome.

These findings suggest that more comprehensive PCPs are not perceived as lower quality when they manage a wider range of beneficiaries' conditions and address new concerns and symptoms in the primary care setting, rather than referring beneficiaries to specialists. This adds to prior work ³⁰ demonstrating that patients value being able to initiate care for new problems with their PCP and seeking the PCP's input on and integration of referrals to specialists. It is also compatible with findings from a more recent cross national study reporting that in countries where primary care practices “serve as a ‘one‐stop shop’, patients perceive better quality of care.” ³¹

Without a comprehensive PCP, Medicare beneficiaries experience more fragmented care, which can be stressful for them and caregivers trying to coordinate care and information received from multiple specialists. ¹ , ³² Fragmented care also increases health care costs, duplicative testing and interventions, hospitalizations, and medications. ¹ , ⁸ , ¹⁰ , ²⁸ , ²⁹ Although specialists play a critical role caring for particularly severe or rare conditions, comprehensive primary care is efficient and effective for more common conditions. ¹ , ³ , ⁶ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ³² Thus, it is not surprising that beneficiaries are also more satisfied with a more comprehensive PCP.

This study has limitations. First, there may be confounding in that more comprehensive PCPs may also do other things (that we cannot necessarily control for) that make patients rate them more favorably. Second, unobserved characteristics could differ between non‐respondent and respondent patients and could affect the observed results; for example, more (or less) satisfied patients may be more likely to respond to the survey. Third, our findings can only be generalized to Medicare patients who receive care from primary care practices that resemble practices participating in CPC+. ³³ Finally, while Medicare claims allow us to capture care delivered across all settings and providers, they lack clinical nuance. Claims cannot easily identify failure to deliver needed services; for example, a highly comprehensive PCP might wait too long to refer a patient to a specialist for a new problem. Assessment of a PCP's involvement in patients conditions is potentially susceptible to differences in providers' inclusion of multiple ICD codes on a claim; however, in prior work we did not find such differences between PCPs who were or were not participating in an advanced payment model where HCC scores affected enhanced payments. ⁶ , ⁸

In sum, this study suggests that efforts to improve the patient‐reported outcomes of health care should note that Medicare beneficiaries rate their primary care physicians' and staff's care higher when their PCP is more comprehensive.

Supporting information

Data S1. Supporting Information.

O'Malley AS, Rich EC, Ghosh A, et al. Medicare beneficiaries with more comprehensive primary care physicians report better primary care. Health Serv Res. 2023;58(2):264‐270. doi: 10.1111/1475-6773.14119