eTable 1. Clinical and Patient Experience Quality Measures
eTable 2. Additional Characteristics of the Medical Expenditure Panel Survey Respondents With and Without Primary Care, 2012-2014
eTable 3. Propensity-Score Weighted Outpatient Quality and Experience for Patients With Primary Care Versus Those Without Primary Care and No Outpatient Visit, 2012-2014
eTable 4. Health Care Utilization With and Without Primary Care Without Propensity Score Weighting, 2012-2014
eTable 5. Outpatient Quality and Experience With and Without Primary Care Without Propensity Score Weighting, 2012-2014
eTable 6. Characteristics of the Medical Expenditure Panel Survey Respondents With and Without Primary Care, 2002-2004
eTable 7. Additional Characteristics of the Medical Expenditure Panel Survey Respondents With and Without Primary Care, 2002-2004
eTable 8. Health Care Utilization With and Without Primary Care, 2002-2004
eTable 9. Propensity Score Weighted Outpatient Quality and Experience With and Without Primary Care, 2002-2004
eTable 10. Outpatient Quality and Experience Without Propensity Score Weighting With and Without Primary Care, 2002-2004
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Levine DM, Landon BE, Linder JA. Quality and Experience of Outpatient Care in the United States for Adults With or Without Primary Care. JAMA Intern Med. 2019;179(3):363–372. doi:10.1001/jamainternmed.2018.6716
How do the quality and experience of outpatient care differ between adults with or without an endorsed source of primary care?
In this nationally representative survey study of 49 286 adults with and 21 133 adults without primary care, Americans with primary care received significantly more high-value care (4 of 5 composites), received slightly more low-value care (1 of 4 composites), and reported significantly better health care access and experience. These differences were stable from 2002 to 2014.
Policymakers and health system leaders seeking to improve value should consider increasing investment in primary care.
The US health care system is typically organized around hospitals and specialty care. The value of primary care remains unclear and debated.
To determine whether an association exists between receipt of primary care and high-value services, low-value services, and patient experience.
Design, Setting, and Participants
This is a nationally representative analysis of noninstitutionalized US adults 18 years or older who participated in the Medical Expenditure Panel Survey. Propensity score–weighted quality and experience of care were compared between 49 286 US adults with and 21 133 adults without primary care from 2012 to 2014. Temporal trends were also analyzed from 2002 to 2014.
Patient-reported receipt of primary care, determined by the 4 “Cs” of primary care: first-contact care that is comprehensive, continuous, and coordinated.
Main Outcomes and Measures
Thirty-nine clinical quality measures and 7 patient experience measures aggregated into 10 clinical quality composites (6 high-value and 4 low-value services), an overall patient experience rating, and 2 experience composites.
From 2002 to 2014, the mean annual survey response rate was 58% (range, 49%-65%). Between 2012 and 2014, compared with respondents without primary care (before adjustment), those with primary care were older (50 [95% CI, 50-51] vs 38 [95% CI, 38-39] years old), more often female (55% [95% CI, 54%-55%] vs 42% [95% CI, 41%-43%]), and predominately white individuals (50% [95% CI, 49%-52%] vs 43% [95% CI, 41%-45%]). After propensity score weighting, US adults with or without primary care had the same mean numbers of outpatient (6.7 vs 5.9; difference, 0.8 [95% CI, −0.2 to 1.8]; P = .11), emergency department (0.2 for both; difference, 0.0 [95% CI, −0.1 to 0.0]; P = .17), and inpatient (0.1 for both; difference, 0.0 [95% CI, 0.0-0.0]; P = .92) encounters annually, but those with primary care filled more prescriptions (mean, 14.1 vs 10.7; difference, 3.4 [95% CI, 2.0-4.7]; P < .001) and were more likely to have a routine preventive visit in the past year (mean, 72.2% vs 57.5%; difference, 14.7% [95% CI, 12.3%-17.1%]; P < .001). From 2012 to 2014, Americans with primary care received more high-value care in 4 of 5 composites. For example, 78% of those with primary care received high-value cancer screening compared with 67% without primary care (difference, 10.8% [95% CI, 8.5%-13.0%]; P < .001). Americans with or without primary care received low-value care with similar frequencies on 3 of 4 composites, although Americans with primary care received more low-value antibiotics (59% vs 48%; difference, 11.0% [95% CI, 2.8%-19.3%] P < .001). Respondents with primary care also reported significantly better health care access and experience. For example, physician communication was highly rated for a greater proportion of those with (64%) vs without (54%) primary care (difference, 10.2%; 95% CI, 7.2%-13.1%; P < .001). Differences in quality and experience between Americans with or without primary care were essentially stable between 2002 and 2014.
Conclusions and Relevance
Receipt of primary care was associated with significantly more high-value care, slightly more low-value care, and better health care experience. Policymakers and health system leaders seeking to improve value should consider increasing investments in primary care.
Primary care—defined herein as first-contact, comprehensive, coordinated, and continuous care—is considered an essential component of well-functioning health care systems.1,2 Beginning in the 1920s, following the Dawson Report, many countries made primary care the foundation of their health systems.3,4 By contrast, the US health care system is generally organized around hospitals and specialty care despite landmark reports, such as the 1966 Millis Commission Report, recommending that each person have a primary care physician.5,6 Moreover, Medicare only recently began supporting free coverage for preventive services and annual wellness visits.7
Consequently, the value of primary care remains unclear and debated in the United States.4 No definitive, large-scale randomized controlled trial has evaluated the effect of primary care on quality and patient experience, nor will such a trial likely occur. Moreover, observational analyses are challenged owing to selection effects compounded by poor identification of participants with or without primary care and poor granularity of the quality and experience that primary care delivers. Some observational studies have examined the association of primary care with quality and experience of care, but these studies generally have been ecological in nature.8 For instance, Baicker and Chandra9 demonstrated that states with more primary care clinicians had higher quality and lower costs. Starfield and colleagues10 found that the regional supply of primary care physicians was associated with lower mortality, higher birth weight, and better self-reported health. Other studies have tested the effect of health insurance on health outcomes, although insurance may be a poor surrogate for primary care.11-13
Ideally, well-functioning primary care should result in increased high-value care, reduced low-value care, and better patient experience and access to care.14,15 To date, however, individual-level empirical data on the impact of primary care are lacking. A more complete understanding of the association between receipt of primary care and the quality and experience of care, as well as how this has changed over time, could inform investments in and use of primary care. Thus, we examined whether receipt of primary care was associated with high-value care, low-value care, or patient access and experience.
We analyzed data from the Medical Expenditure Panel Survey (MEPS) from 2002 to 2014 with particular focus on 2012 to 2014. The MEPS is a nationally representative annual survey of the noninstitutionalized United States civilian population drawn from respondents to the National Health Interview Survey.16 The MEPS employs a complex survey design across 2 years that delivers English or Spanish language computer-assisted personal interviews to collect detailed data on demographic characteristics, health conditions, health status, medical services use, medications, cost, source of payments, health insurance coverage, income, employment, experience with care, and access to care. From 2002 to 2014, the annual MEPS response rates ranged from 49% to 65% (mean 58%). The Harvard Medical School Institutional Review Board determined this study not to be human subject research and thus waived the need both for review and for obtaining informed patient consent.
The MEPS supplements and validates self-reported information by contacting respondents’ clinicians (mean response rate, 86%), hospitals (mean response rate, 90%), pharmacies (mean response rate, 77%), and employers (mean response rate, 91%). Clinicians specify details regarding office visits (diagnosis, diagnostic test, cost, etc); hospitals specify admissions; pharmacies specify individual medications; and employers specify insurance plan particulars.
The MEPS also includes 2 additional mail-back surveys: the adult self-administered questionnaire and the diabetes care survey. The self-administered questionnaire includes items from the Consumer Assessment of Healthcare Providers and Systems survey, the 12-item Short Form Health Survey, and additional items measuring respondents’ attitudes about health care (annual response rate range, 89%-94%). The diabetes care survey, administered to respondents with self-reported diabetes, includes items related to diabetes care (annual response rate range, 88%-97%).
We restricted our analyses to the adult population aged 18 or older. Sample sizes ranged from 21 915 to 26 509 respondents per year.
We used a patient-centered definition of primary care that used responses to a series of questions about core aspects of primary care to determine whether the respondent was engaged in a primary care relationship. The MEPS first determines whether respondents have a “usual source of care” by asking them the name of a physician to whom “you usually go if you are sick or need advice about your health.” We considered respondents able to identify such a physician who practiced outside of the emergency department as having a “usual source of care.”
To further delineate respondents with primary care, we used 4 additional questions to replicate the 4 “Cs” of primary care: first-contact care that is comprehensive, continuous, and coordinated.10 We only included those who responded affirmatively that they would visit their usual source of care for all 4 of the following: “new health problems” (first contact); “preventive health care, such as general checkups, examinations, and immunizations” (comprehensive); “ongoing health problems” (continuous); and “referrals to other health professionals when needed” (coordinated). Adults who answered no to the usual source of care question or to any of the other 4 questions were considered to not have primary care. Among respondents with a usual source of care, 95% met the full criteria for having primary care. Respondents could have selected a health professional from any specialty as their primary care clinician as long as they met those criteria. Of the health professionals selected, 70% were general or family practice physicians, 19% were general internists, 3% were nurse practitioners or physician assistants, 1% were pediatricians, 1% were obstetrician/gynecologists, and the remainder were from all other specialties.
We developed clinical quality measures and quality composites from the MEPS as previously described (eTable 1 in the Supplement).17 We evaluated performance on 39 clinical quality measures, including 25 high-value measures and 14 low-value measures. From these measures, we constructed 6 clinically meaningful underuse composites (eg, recommended cancer screening) in which delivery of the service is likely of benefit to the respondent, and 4 overuse composites (eg, avoidance of imaging in specific clinical situations) in which delivery of the service is considered either inappropriate or of little to no benefit.
To calculate performance for each measure, we first identified those respondents who were eligible for the measure (eg, those with diabetes) and then determined whether or not they received the particular care (eg, retinal exam). To calculate composites, we divided all instances in which recommended care was delivered (for high-value measures) or avoided (for low-value measures) by the number of times respondents were eligible for care in the category, as others have previously done.18
We evaluated a global rating measure that asked about respondent experience with all health providers (range, 0 “worst health care possible” to 10 “best health care possible”). We also evaluated a doctor communication composite that asked 4 items (eg, “How often did the doctor spend enough time with you?”) and an access to care composite that included 2 items (eg, “How often did you get a medical appointment as soon as wanted?”)17; responses were coded from “never” (1) to “always” (4). To better discriminate changes over time, we dichotomized all measures such that a positive response included 8, 9, or 10 for the items scored from 0 to 10 and 4 for the items scored from 1 to 4, similar to the Healthcare Effectiveness Data and Information Set analyses.19 We calculated both experience composites by first computing the mean for each respondent and then taking the mean for all respondents.
We used propensity score weighting to address potential sources of confounding between receipt of primary care and the outcomes of interest. Potential sources of confounding included demographic factors, socioeconomic status, health and functional status, and engagement with the health system.20-22 This method resulted in a comparison between those with or without a primary care relationship, with similar levels of engagement in care who were balanced on the above stated factors.
We used survey-weighted logistic regression to create a propensity score of having primary care, adjusting for all variables given in Table 1 and whether a respondent needed assistance with activities of daily living, assistance with instrumental activities of daily living, family income as a percent of the poverty line, and 12-item Short Form Health Survey Physical and Mental component summary scores (eTable 2 in the Supplement). For those with primary care, we computed the inverse of the propensity score. For those without primary care, we computed the inverse of 1 minus the propensity score. We then multiplied these weights by the existing MEPS survey weights.24 Item nonresponse across the survey was low and after weighting resulted in a loss of 11% and 15% of respondents with or without primary care, respectively. Unless otherwise specified, we present propensity-weighted analyses. All analyses without propensity weighting are in the Supplement. Finally, to determine whether there was a “dose-response” association between primary care and quality and experience, we also compared respondents with primary care to those without primary care who also had no outpatient visit (eTable 3 in the Supplement).
In all analyses, we generated national estimates as recommended by the MEPS by using survey estimation weights, primary sampling unit clusters, and sampling strata that accounted for the complex survey design of the MEPS and for nonresponse.25,26 For our main analyses, we aggregated responses from the most recent 3 years of the survey: 2012 to 2014. To examine whether performance was improved at the end of the study period relative to that at the beginning, we compared composites in 2002 to 2004 to those in 2012 to 2014 using χ2 tests, adjusting for the complex survey design.27 Because we found few temporal changes, the results presented here focus on the most current data, from 2012 to 2014, but we give temporal differences where relevant. We performed all analyses using SAS, version 9.4 (SAS Institute Inc) and considered a 2-sided P < .05 to be significant.
Between 2012 and 2014, compared with respondents without primary care, those with primary care were older (mean, 50 [95% CI, 50-51] vs 38 [95% CI, 38-39] years old), more often female (55% [95% CI, 54%-55%] vs 42% [95% CI, 41%-43%]), predominately white individuals (50% [95% CI, 49%-52%] vs 43% [95% CI, 41%-45%]), more frequently smokers (14% [95% CI, 13%-14%] vs 20% [95% CI, 19%-21%]), more often poor (11% [95% CI, 10%-12%] vs 18% [95% CI, 17%-19%]), and had a higher chronic disease burden (23% [95% CI, 22%-23%] with ≥3 chronic diseases vs 4% [95% CI, 3%-4%]) (all P < .001) (Table 1). They also had lower rates of uninsurance (34% [95% CI, 32%-35%] vs 7% [95% CI, 7%-8%]; P < .001). These differences were stable between 2002 and 2014. After propensity score weighting, there were no significant differences in these measured attributes between respondents with or without primary care (Table 1; and eTable 2 in the Supplement).
After propensity score weighting, respondents with or without primary care used health care with a similar frequency (Table 2), including similar mean numbers of annual office visits (6.7 vs 5.9; difference, 0.8 [95% CI, −0.2 to 1.8]; P = .11), annual emergency department visits (0.2 for both; difference, 0.0 [95% CI, −0.1 to 0.0]; P = .17), and annual hospital admissions (0.1 for both; difference, 0.0 [95% CI, 0.0-0.0]; P = .92). By contrast, respondents with primary care filled more prescriptions each year (mean, 14.1 vs 10.7; difference, 3.4 [95% CI, 2.0-4.7]; P < .001) and more frequently had a routine preventive visit within the past year (mean, 72.2% vs 57.5%; difference, 14.7% [95% CI, 12.3%-17.1%]); P < .001).
Respondents with primary care received more high-value care compared with those without primary care in 4 of 5 composites (Table 3). Approximately 78% of respondents with primary care received high-value cancer screening compared with 67% without primary care (difference, 10.8% [95% CI, 8.5%-13.0%]; P < .001). The largest differences were for colorectal cancer screening (16.1% [95% CI, 12.0%-20.3%], P < .001) and mammography (14.2% [95% CI, 8.8%-19.6%], P < .001).
Respondents with primary care also received more recommended diagnostic and preventive testing (difference, 9.9% [95% CI, 8.7%-11.2%]; P < .001). For example, an adjusted 10.4% (95% CI, 6.1%-14.6%) more received an influenza vaccine, and 9.5% (95% CI, 8.3%-10.6%) more had blood pressure checked (both P < .001). For respondents with primary care and diabetes, an adjusted 7.8% (95% CI, 1.2%-14.4%) more received high-value diabetes care (P = .02). High-value counseling among respondents with primary care was also higher (difference, 6.9% [95% CI, 4.1%-9.7%]); P < .001), particularly for smoking cessation counseling (difference, 12.3% [95% CI, 6.2%-18.5%]; P < .001). By contrast, respondents with or without primary care received similar rates of high-value medical treatments, such as receipt of a β-blocker for treatment of coronary artery disease (difference −4.6%, [95% CI, −14.3 to 5.0]; P = .34).
For the relatively small number of patients with heart failure or pulmonary disease (30 and 48, respectively), respondents with primary care received less high-value care. For example, those with primary care received fewer β-blockers in heart failure (difference, −8.4% [95% CI, −10.3% to −6.6%]; P < .001) and fewer controller medications in poorly controlled asthma (difference, −15.4% [95% CI, −18.5% to −12.4%]; P < .001). Of those with primary care included in the β-blocker measure, 62% also were also seen by a cardiologist, and of those with primary care included in the asthma measure, 48% were also seen by a pulmonologist.
Respondents with or without primary care received similar low-value care in 3 of 4 composites (Table 3). Approximately half (49% [95% CI, 47%-51%]) of primary care respondents received low-value cancer screening, which was not significantly different from the 44% (95% CI, 37%-50%) without primary care (P = .12). Within this composite, only low-value prostate cancer screening differed significantly (difference, 9.8% [95% CI, 7.4%-12.2%) for those with primary care; P < .001). We observed no significant differences in receipt of low-value medical treatments (11% for both groups; difference, 0.0% [95% CI, −2.7% to 2.6%]; P = .99) or low-value imaging (approximately 10% in both groups; difference, −1.3% [95% CI, −5.1% to 2.5%]; P = .50). Respondents with primary care received more low-value antibiotics (59%) than those without primary care (48%; difference, 11.0% [95% CI, 2.8%-19.3%]; P = .01).
Despite similar levels of use of both outpatient and inpatient care, respondents with primary care had better experience than those without primary care (Table 3). For example, 79% of respondents with primary care reported an excellent global rating of their health care compared with 69% without primary care (difference, 10.4% [95% CI, 6.9%-13.8%]; P < .001). Physician communication was highly rated for a greater proportion of those with (64%) vs without (54%) primary care (difference, 10.2%; 95% CI, 7.2-13.1; P < .001), and report of access to care was also better (59% vs 52%; difference, 7.0% [95% CI, 3.8%-10.1%]; P < .001).
Over time, we observed no changes in the above findings, with only 1 exception (eTables 6-10 in the Supplement). There was a reduction in low-value cancer screening for respondents with primary care (53% [95% CI, 52%-55%] in 2002-2004 vs 49% [95% CI, 47%-51%] in 2012-2014; P < .001).
Our sensitivity analysis examining patients who were not at all engaged in care showed substantially larger differences (eTable 3 in the Supplement). Those with primary care received more high-value care but also received more low-value care. For example, approximately 78% of respondents with primary care received high-value cancer screening compared with 47% without primary care and no outpatient visit (difference, 31.6% [95% CI, 26.5%-36.7%]; P < .001).
In this large, nationally representative survey study, we quantified the potential benefit of primary care with respect to receipt of high- and low-value health services and experience with and access to care within the current health care delivery environment. We found that receipt of primary care was associated with more high-value care, somewhat more low-value care, and better respondent access and experience. Respondents without primary care, even though they were receiving a similar amount of care, missed substantial health care benefits: about 10% fewer went without high-value cancer screening, diagnostic and preventive testing, diabetes care, and counseling. Similarly, about 10% fewer respondents without primary care rated their overall care, physician communication, and access to care as excellent. These differences are noteworthy when considered in the context of mixed or flat improvements in quality during the last decade.17 To our knowledge, this is the first study to directly compare outpatient quality and experience when delivered inside or outside of a primary care relationship.
Primary care, however, was not uniformly associated with more high-value care. For instance, primary care was associated with worse care for heart failure and pulmonary disease, albeit with relatively small numbers of respondents without primary care qualifying for these measures (approximately 50 patients or fewer for both). Approximately half of patients with primary care who qualified for these measures also had visits with a relevant specialist. Prior research shows that, in general, specialists provide higher quality care in their area, but largely do not address issues outside of their specialty; thus, these findings should not be interpreted as suggesting that a specialty dominated model would be better.28 Care for patients with heart failure or pulmonary disease could potentially be improved with better primary-specialty care co-management, increased education of primary care physicians, or other interventions.
The association between primary care and low-value care presents a more mixed picture. We observed more preventive visits, which some have criticized as low-value care in some cases29 although, generally, this controversy relates to “annual” preventive visits, and most observers agree that some frequency of preventive visits is likely worthwhile. Americans with primary care had similar rates of low-value care on 3 of 4 composites and increased low-value antibiotic use. Antibiotic prescribing in the primary care setting has been an area of intense interest in the last 20 years. Related measures have been a standard part of many pay-for-performance programs. Thus, we would have expected that primary care would have been associated with less low-value antibiotic use. It is possible, as currently structured, that primary care does not sufficiently protect against low-value care, but as the United States transitions to a value-based system, efforts to decrease low-value care may be more effective.30
We also found that Americans’ use of primary care was relatively low. About one-quarter of adults reported not having primary care, yet 67% of Americans without primary care had health insurance (and the majority had private insurance). Poor primary care supply or access may be hurdles,31 or some Americans do not perceive the potential value of primary care, particularly if they are younger (the mean unadjusted age for those without primary care was 38 years, as opposed to a mean age of those with primary care of 50 years) and healthier. These findings contrast with those of other health systems throughout the world; for example, universal primary care registration is required in the United Kingdom32 and the Netherlands.33
There are 2 main sources of confounding that should be considered when interpreting the present results. First, some people actively avoid interacting with the health care system and thus have very few opportunities to receive recommended (or nonrecommended) services. Including such persons in the “no primary care” comparison group, therefore, would bias our results. Second, the presence or absence of an endorsed source of primary care also could be associated with health status or the presence of acute or chronic health conditions. In some cases, those with severe health conditions might choose to see only specialist physicians without identifying a single first-contact physician as their primary care physician. Alternatively, those who are relatively healthy simply might choose to forego having a primary care physician, instead choosing to access care as issues arise.
To guard against these 2 confounders, we used a propensity score weighting approach to balance sociodemographic and clinical characteristics. We did not include use measures in our propensity score model, yet weighting on all other characteristics resulted in near-identical levels of use (eg, similar numbers of outpatient, emergency department, and inpatient visits). Thus, our findings show the potential benefits of having an endorsed source of primary care for respondents with similar health status and conditions, all of whom are engaged with the health care system to a similar extent, rather than simply showing that some care is better than no care. We also performed a sensitivity analysis comparing respondents with primary care to those without primary care and without any outpatient visits. Not surprisingly, those with very limited engagement in the system had markedly worse quality and experience.
To our knowledge, this is the first study to directly compare outpatient quality and experience when delivered inside vs outside of a primary care relationship. Our work is consistent with, but also adds substantially to, prior studies showing that areas with more primary care clinicians had higher quality, lower costs,9 lower mortality, and better self-reported health.10 Our results are also consistent with prior research that demonstrates that adults with a usual source of care,34-37 those who are attributable by claims to a physician or group,38 or those who report highly patient-centered care39,40 are more likely to receive preventive services. In particular, several prior studies used earlier versions of the MEPS to examine similar questions. DeVoe and colleagues41 used 1996 data from the MEPS to compare receipt of preventive services for insured adults with a usual source of care and uninsured adults without regular care (akin to our sensitivity analysis) and found that the latter were substantially less likely to have received preventive services. Later studies found better reported communication42 and higher rates of blood pressure and hemoglobin A1C assessment among respondents with diabetes for those with vs without a usual source of care.43 The comparison group in most of those studies, however, was a group that was minimally engaged in care. In addition, a prior study by VanGompel40 also used the MEPS to assign a continuous 7-point “primary care attributes” score to respondents and then examined receipt of preventive services. That approach, however, did not allow for the comparison of those with or without defined primary care and did not use propensity score adjustment. Our work builds on these studies, adding multiple facets of outpatient quality and patient experience beyond preventive services, examining outcomes over more than a decade, and using more robust propensity score analyses.
Our study had limitations. First, although our definition of primary care was directly aligned with the 4 “C’s” of primary care and was patient-centered, it may differ from other definitions of primary care.8 Instead of assigning a patient to primary care by virtue of a claims algorithm, for example, our definition was derived from the patient’s perspective. We acknowledge that we may have been unable to detect the intricacies of each primary care service model, but if we were indeed missing important intricacies, our findings represent the minimum difference between those with or without primary care. Second, our study was observational; thus, we could not interpret the associations we observed as causal. Third, our use of propensity score weighting adjusted for observable factors but not unobserved confounders. Fourth, our quality measures did not address all outpatient quality. For instance, we lacked measures of intermediate outcomes, such as control of hypertension or diabetes. However, to our knowledge, the MEPS represents one of the largest nationally representative sets of consistently collected quality measures available for more than a decade.17 Fifth, propensity score weighting resulted in a small loss of data, but those respondents with more missing data were less likely to have primary care and more likely to have worse quality of care. Therefore, omitting respondents with more missing data biased our results toward the null.
Receipt of primary care characterized by first-contact continuous care that was whole-person oriented and responded to patient needs was associated with significantly more high-value care, slightly more low-value care, and better health care experience. Policymakers and health system leaders seeking to improve value should consider increasing investments in primary care.
Accepted for Publication: October 6, 2018.
Published Online: January 28, 2019. doi:10.1001/jamainternmed.2018.6716
Correction: This article was corrected on April 15, 2019, to fix a data error in the Findings paragraph of the Key Points.
Corresponding Author: David M. Levine, MD, MPH, MA, Division of General Internal Medicine and Primary Care, Harvard Medical School, Brigham and Women’s Hospital, 1620 Tremont St, Third Floor, Boston, MA 02120 (email@example.com).
Author Contributions: Drs Linder and Landon contributed equally to this article. Dr Levine had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: All authors.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Levine, Linder.
Critical revision of the manuscript for important intellectual content: Landon, Linder.
Statistical analysis: Levine, Linder.
Administrative, technical, or material support: Levine, Linder.
Supervision: Landon, Linder.
Conflict of Interest Disclosures: None reported.
Funding/Support: Dr Levine reported receiving an Institutional National Research Service Award (T32HP10251) from the National Institutes of Health and funding support from the Ryoichi Sasakawa Fellowship Fund.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, or approval of the manuscript.