Starfield1 describes continuity of care as 1 of the 4 pillars of primary care, along with first-contact, comprehensiveness, and concern for the entire patient.2 Continuity of care has been described in the literature in a number of ways including managerial, relational, and informational.3 In particular, relational continuity of care can be defined as “an ongoing therapeutic relationship between a patient and a provider.”4 One core approach to measuring and defining relational continuity is by identifying whether a patient has a “regular” or “personal” physician.
Many studies have demonstrated that the continuity provided by having a regular or personal physician can lead to improvements in a wide range of clinical quality measures. Researchers have found a strong relationship between having a personal physician and receiving preventive care such as vaccinations, cancer screening and/or diagnosis, and cholesterol checks.5–13 Other quality measures positively associated with having a personal physician include receipt of eye examinations for diabetics,8 beta-blocker use among patients who had a myocardial infarction,8 follow-up after mental health hospitalization,8 emergency department utilization,14 and self-reported health status.15 Although only 2 of these studies focused on Medicare or otherwise older populations, the health benefits of this higher quality care identified in the studies may be especially important for the Medicare population. Lower quality of care and less receipt of preventive care may have immediate or severe consequences for those Medicare beneficiaries who are older and/or more vulnerable. However, relatively little is known about the demographic or clinical characteristics of older adults who lack a personal physician.16,17 Furthermore, very little is known about how patient-reported care experiences differ for people that do and do not have a personal physician.13,18
Using data from a nationally representative sample of Medicare beneficiaries, this study aims to describe differences in patient experience among beneficiaries who do and do not have a personal physician, as well as beneficiary characteristics associated with not having a personal physician.
Findings from this study potentially have important implications for health care policy and practice. By providing coverage for health care services, Medicare removes a large barrier to access to care. If we find that a nontrivial proportion of Medicare beneficiaries do not have access to a personal physician, understanding the access to care barriers among this insured population may be relevant to other insured populations. Furthermore, patient-centered care is a key factor for improving the US health care system.15 To the extent that persons without a personal physician report worse care experiences, improving access to a personal physician may be an important tool to improve care quality. Also, to the extent that the lack of a personal physician varies by race/ethnicity and socioeconomic status, improving access to personal physicians may be an important approach to reducing disparities in patient experience.
In this paper, we analyze the relationship between 3 key concepts: (1) continuity, (2) patient experience, and (3) patient characteristics. Consistent with Haggerty et al3 we define continuity of care as “the relationship between a single practitioner and a patient that extends beyond specific episodes of illness or disease.” Continuity is a key aspect of an effective primary care system. We differentiate between informational, management, and relational continuity. The presence of relational continuity, which we define as an ongoing therapeutic relationship with a provider, can be measured and assessed by the extent to which an individual has identified a personal physician.
In this study, patient experience is conceptualized as both the patients’ overall rating of their health care as well as the extent to which patients receive needed care. Continuity of care, afforded by having a personal physician, is marked by having a long-term relationship with a physician that becomes a trusted advocate for patients’ needs. Therefore, a patient may feel more bonded to the physician and be more likely to report a positive experience with that provider. Furthermore, if a patient has a personal physician, they are likely to be known by the physician and be a part of their panel. This relationship may help a patient be seen in a timely manner, improving patient experience.
We used data from a nationally representative sample of Medicare beneficiaries who participated in the 2012 Medicare Consumer Assessment of Healthcare Providers and Systems (MCAHPS) survey [both Medicare Advantage and fee-for-service (FFS)].19 Respondents were included in the present analysis if they lived in the United States, were 65 years of age or older, and answered the personal physician item on the survey (A personal doctor is the one you would see if you need a check-up, want advice about a health problem, or get sick or hurt. Do you have a personal doctor?). The survey response rate for the population studied was 48.1%, resulting in a final sample size of 272,463 beneficiaries for this analysis.
We constructed a binary measure assessing whether a beneficiary had a personal physician. Beneficiaries reported their race/ethnicity, sex, education, self-rated general and mental health, and use of proxy assistance. Age, urbanicity (using Beale Codes20), and the survey language used were available administratively. These beneficiary characteristics were measured categorically; their specifications are shown in Table 1. We used 4 patient experience measures from the MCAHPS survey as dependent variables: 3 composite measures constructed from multiple items (getting needed care, getting care quickly, and getting needed drugs) and a single global rating (the respondent’s overall rating of their health care). We chose these 4 measures that were available for respondents with and without a personal physician because we hypothesized that they were the most likely to be affected by having a personal physician. The patient experience measures were all transformed to a 0–100 scale for ease of comparison.
We first described the characteristics of the study sample based on coverage type, age, sex, education, race/ethnicity, self-rated general and mental health, use of proxy assistance, Medicaid eligibility, survey language, and urbanicity. We then tested for bivariate associations between these beneficiary characteristics and whether the beneficiary had a personal physician using weighted regressions.
To estimate adjusted relationships between each of the respondent characteristics and having a personal physician, we compared the rate of not having a personal physician across demographic subgroups using multivariate logistic regression. Specifically, we included as predictors age, race/ethnicity, sex, education, urbanicity, self-rated general and mental health status, Medicaid eligibility, survey language, and proxy assistance in a main effects-only model. In a second model, we also included an interaction term of age and education, and interactions of general health status with education and race/ethnicity. For the purposes of the interaction terms only, age, general health status, and education were coded linearly (as 1 degree of freedom).
Finally, we compared the experience for beneficiaries with and without a personal physician. To do this, we used doubly robust propensity-score–weighted regression models.21 We ran 4 separate models, 1 for each patient experience measure. We fit the doubly robust regressions by estimating a propensity-score model, deriving propensity-score weights, and then fitting a covariate-adjusted propensity-score–weighted model to the dependent variable of interest. In this case, we used gradient-boosted regression via the twang package in R to obtain the propensity-score weights.22 The doubly robust models incorporate the benefits of both propensity-score weighting and regression estimators,23 providing additional protection against model misspecification.
Propensity-score weights for beneficiaries with a personal physician were derived to match the weighted distribution of their characteristics to the unweighted distribution of characteristics in the group without a personal physician; we used the established criterion of all standardized differences being <0.2 in absolute value to confirm balance.24 The propensity-score weights were multiplied by the survey weights to derive final analytic weights. We fit weighted linear regression models predicting patient experience measures from an indicator of not having a personal physician and all beneficiary characteristics and interaction terms in the multivariate model that predicted not having a personal physician. We also fit main effects patient experience models that omitted the interaction terms so that the coefficients for characteristics such as race/ethnicity could be used as a reference.
Table 1 presents the descriptive statistics for our study population and compares them across respondents that do and do not have a personal physician. We found that 4.9% of beneficiaries did not have a personal physician (95% confidence interval, 4.8–5.1). Moreover, there were several subpopulations with relatively large proportions of beneficiaries without a personal physician, including 9.3% of beneficiaries with less than a high school education, 10.5% of Hispanic beneficiaries, 16.3% of American Indian/Alaska Native, 9.1% of dual eligibles, and 18.3% of those responding in Spanish. Some cross-tabulated subpopulations have particularly high proportions without personal physicians; for Medicaid-eligible men and Hispanic men the proportion exceeds 13% (data not shown).
In bivariate analysis, we found that those without a personal physician were more likely to be enrolled in FFS Medicare without a prescription drug plan (“FFS-Only” coverage) and less likely to be enrolled in Medicare Advantage relative to those with a personal physician. They were also more often male, younger, of lower educational attainment, a racial/ethnic minority, and living in a more rural area. Beneficiaries without a personal physician were also more likely to be in excellent general health but less likely to be in excellent or very good mental health. They were also slightly more likely to use any proxy assistance and complete the survey in Spanish. Finally, they were much more likely to be dually eligible for Medicaid than those with a personal physician.
Table 2 shows the results of the main effects multivariate regression model that predicted not having a personal physician. All the joint tests for multiple levels of a single conceptual variable (eg, race/ethnicity) were significant, suggesting that each of the variables was associated with not having a personal physician after controlling for the other covariates. Specifically, multivariate results show that men, racial/ethnic minorities, those who completed the survey in Spanish, those with less education, those living in the most rural locations, and those with FFS-Only coverage were more likely to lack a personal physician. Lacking a personal physician was more common for those in better general health and those in worse mental health. Older beneficiaries were less likely to lack a personal physician. In the multivariate model that added interaction terms, the joint test for the interaction terms was not significant (P=0.08; results not shown).
Table 3 summarizes the results of the doubly robust estimation for patient experience measures and Table 4 shows coefficients for all predictors in the models. Propensity-score weighting resulted in well-balanced samples across those beneficiaries with and without a personal physician (Supplemental Digital Content 1, http://links.lww.com/MLR/B533). We found that those without a personal physician provided much lower ratings of their overall health care (β=−7.96), indicating that beneficiaries who do not have a personal physician rate care 8 points lower than otherwise similar beneficiaries with a personal physician. Beneficiaries without a personal physician also reported much worse experiences on composite measures of getting needed care (β=−7.13), getting care quickly (β=−8.10), and getting needed drugs (β=−11.85), (P<0.001 for all results in this paragraph). These coefficients are significantly larger than the effect sizes for any other covariate, including race/ethnicity, socioeconomic status, sex, and age. These comparison results from the main effects patient experience models (without interaction terms) are shown in Table 5.
This study examines the care experience of Medicare beneficiaries with and without a personal physician. We find that 4.9% (95% confidence interval, 4.8–5.1) of Medicare beneficiaries do not have a personal physician. It is unsurprising that only a small proportion of Medicare beneficiaries do not have a personal physician given that all have health insurance coverage. Because insurance coverage is a positive determinant of access to care, one would expect Medicare beneficiaries to have access to a provider. However, the 5% of the Medicare population without a personal physician is more than 2 million people, a significant population.
Lacking a personal physician is consequential. Even after controlling for many key observable beneficiary characteristics, beneficiaries without a personal physician report worse care experiences, rating their overall quality of care substantially lower than those with a personal physician. In addition, those without a personal physician reported lower ratings of patient experiences on measures most closely related to access, such as getting needed care and drugs and getting care quickly. These differences were quite large, ranging from 7 to 12 percentage points. Previous studies have suggested that differences of 3 percentage points in CAHPS scores are meaningfully large.25,26 Furthermore, we found that lacking a personal physician was a stronger predictor of patient experience than any of the other covariates including race/ethnicity, socioeconomic status, and sex. Only 2 previous studies have looked at this issue and were limited by small sample sizes or focused on non-US populations.13,18 Our study provides one of the first large-scale investigations of this issue among seniors, a population that is older and less healthy than the overall adult population and therefore may be especially affected by poor access to care. Our findings provide further evidence of the relatively poorer care experiences among those who do not have a personal physician.
Our findings have important implications for disparities. We found that beneficiaries without a personal physician differ markedly from other beneficiaries. Most notably, beneficiaries without a personal physician are more likely to be male, younger, less educated, lower-income (Medicaid-eligible), a racial/ethnic minority, in poorer mental health (but better overall health), and covered under FFS Medicare without a Prescription Drug Plan. For some subpopulations, the proportion of beneficiaries without a personal physician is much higher, as high as 10%–16%.
Our work confirms previous studies describing racial/ethnic disparities in having a usual source of care or a personal (ie, regular) physician.16,17 Our study also finds that there are other important beneficiary characteristics that are significantly associated with not having a regular provider. Less educated beneficiaries are less likely to have a personal physician, which may suggest a reduced capacity to navigate the health care system compared with more educated beneficiaries.27,28 Those in worse general health are less likely to lack a personal physician, as they have greater need for care and more experience in the health care system, but those in poor mental health are at particular risk. Those in the most rural locations may have poorer access to physicians in general.29
Our study is not without limitations. First, caution must be used in making causal attributions to our estimates. However, we do control for many observable beneficiary characteristics that are likely to be associated with patient experience. Furthermore, we use a robust estimation procedure that protects against several forms of misspecification. Second, the survey’s definition of a personal doctor does not allow for having a nurse practitioner or other provider as a primary care provider. Cognitive testing of patient experience measures indicated that many patients are confused when asked about “providers” rather than “doctors” or “physicians.”30 Also, when asked about their personal physician patients generally also included providers who were not physicians.30 Nonetheless, this is a potential limitation, and it may be the case that some who say they do not have a personal doctor instead have regular access to another type of provider. We believe that our findings, that (1) patients report having no personal doctor only 5% of the time, and (2) those people have negative experiences, are consistent with the interpretation that respondents are often or even usually including personal providers who are not physicians, as applicable. Third, our analysis is limited to the Medicare population older than 64, so may not generalize to younger populations. Nevertheless, our results may have important implications for other adults as they transition from being uninsured to insured. Finally, our response rates are moderate, but they are similar to other similar surveys of patient experience.31–33 Furthermore, research suggests that response rates are a poor indicator of nonresponse bias in surveys that, like this survey, adhere to recommended survey process standards.34
In conclusion, we find that a small proportion, but large number of Medicare beneficiaries older than 64 have no personal physician; these beneficiaries are different from patients that have a personal physician with respect to age, race/ethnicity, sex, education, language preference, and health status. Furthermore, those without a personal physician are likely to report substantially worse patient experiences and much less recent use of outpatient care. Taken together, lack of a personal physician may be 1 determinant of disparities in patient experience and other aspects of health and health care by race/ethnicity and socioeconomic status, even when health care coverage is assured. Medicare would be well-served to better understand who does and does not have a personal physician and take actions to help connect beneficiaries to providers. More research is needed to understand policy initiatives that can help Medicare beneficiaries and others to connect to a regular health care provider.
The authors would like to thank Biayna Darabidian, BA for preparation of the manuscript.
1. Starfield B. Primary Care: Balancing Health Needs, Services, and Technology. Oxford, UK: Oxford University Press; 1998.
2. Bodenheimer T, Pham HH. Primary care: current problems and proposed solutions. Health Aff (Millwood). 2010;29:799–805.
3. Haggerty JL, Reid RJ, Freeman GK, et al. Continuity
of care: a multidisciplinary review. BMJ. 2003;327:1219–1221.
4. Kristjansson E, Hogg W, Dahrouge S, et al. Predictors of relational continuity
in primary care: patient, provider and practice factors. BMC Fam Pract. 2013;14:72.
5. Lu PJ, Euler GL, Callahan DB. Influenza vaccination among adults with asthma: findings from the 2007 BRFSS survey. Am J Prev Med. 2009;37:109–115.
6. Allen SM, Wieland S, Griffin J, et al. Continuity
in provider and site of care and preventive services receipt in an adult Medicaid population with physical disabilities. Disabil Health J. 2009;2:180–187.
7. Shenson D, Bolen J, Seeff L, et al. Are older adults up-to-date with cancer screening and vaccinations? Prev Chronic Dis. 2005;2:3.
8. Schneider EC, Zaslavsky AM, Landon BE, et al. National quality monitoring of Medicare health plans: the relationship between enrollees' reports and the quality of clinical care. Med Care. 2001;39:1313–1325.
9. Carlos RC, Fendrick AM, Patterson SK, et al. Associations in breast and colon cancer screening behavior in women. Acad Radiol. 2005;12:451–458.
10. Carlos RC, Underwood W III, Fendrick AM, et al. Behavioral associations between prostate and colon cancer screening. J Am Coll Surg. 2005;200:216–223.
11. Martínez-Donate AP, Cala LM, Zhang X, et al. Prevalence and correlates of breast and cervical cancer screening among a Midwest community sample of low-acculturated Latinas. J Health Care Poor Underserved. 2013;24:1717–1738.
12. Siminoff LA, Rogers HL, Thomson MD, et al. Doctor, what’s wrong with me? Factors that delay the diagnosis of colorectal cancer. Patient Educ Couns. 2011;84:352–358.
13. Beal A, Hernandez S. Patient reports of the quality of care in community health centers: the importance of having a regular provider. J Health Care Poor Underserved. 2010;21:591–605.
14. McCusker J, Tousignant P, Da Silva RB, et al. Factors predicting patient use of the emergency department: a retrospective cohort study. Can Med Assoc J. 2012;184:E307–E316.
15. Institute of Medicine: Committee on Quality of Health Care in America. Crossing the quality chasm: a new health system for the 21st century. Washington, DC; 2001.
16. Carpenter WR, Godley PA, Clark JA, et al. Racial differences in trust and regular source of patient care and the implications for prostate cancer screening use. Cancer. 2009;115:5048–5059.
17. Mahmoudi E, Jensen GA. Exploring disparities
in access to physician services among older adults: 2000-2007. J Gerontol B Psychol Sci Soc Sci. 2013;68:128–138.
18. Kalda R, Polluste K, Lember M. Patient satisfaction with care is associated with personal choice of physician. Health Policy. 2003;64:55–62.
19. Agency for Healthcare Research and Quality. CAHPS Clinician and Group Survey and Instructions. Rockville, MD: Agency for Healthcare Research and Quality; 2015.
20. National Cancer Institute. Rural-Urban Continuum Codes. 2014. Available at: http://seer.cancer.gov/seerstat/variables/ses/beale_codes/yr1969_2002/
. Accessed October 23, 2014.
21. Robins JM, Hernan MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology. 2000;11:550–560.
22. Ridgeway G, McCaffrey D, Morral A, et al. twang: Toolkit ftwang: toolkit for weighting and analysis of nonequivalent groups. R package version 1.3–20. 2013.
23. Bang H, Robins JM. Doubly robust estimation in missing data and causal inference models. Biometrics. 2005;61:962–973.
24. Cochran WG. The effectiveness of adjustment by subclassification in removing bias in observational studies. Biometrics. 1968;24:295–313.
25. Paddison CA, Elliott MN, Haviland AM, et al. Experiences of care among Medicare beneficiaries with ESRD: Medicare Consumer Assessment of Healthcare Providers and Systems (CAHPS) survey results. Am J Kidney Dis. 2013;61:440–449.
26. Lied TR, Sheingold SH, Landon BE, et al. Beneficiary reported experience and voluntary disenrollment in Medicare managed care. Health Care Financ Rev. 2003;25:55–66.
27. Martin LT, Ruder T, Escarce JJ, et al. Developing predictive models of health literacy. J Gen Intern Med. 2009;24:1211–1216.
28. Elliott MN, Haviland AM, Orr N, et al. How do the experiences of Medicare beneficiary subgroups differ between managed care and original Medicare? Health Serv Res. 2011;46:1039–1058.
29. Chan L, Hart LG, Goodman DC. Geographic access to health care for rural Medicare beneficiaries. J Rural Health. 2006;22:140–146.
30. Harris-Kojetin LD, Fowler FJ Jr, Brown JA, et al. The use of cognitive testing to develop and evaluate CAHPS™ 1.0 core survey items. Med Care. 1999;37:MS10–MS21.
31. Roland M, Elliott MN, Lyratzopoulos G, et al. Reliability of patient responses in pay for performance schemes: analysis of national General Practitioner Patient Survey data in England; 2009.
32. Jha AK, Orav EJ, Zheng J, et al. Patients’ perception of hospital care in the United States. N Engl J Med. 2008;359:1921–1931.
33. Elliott MN, Kanouse DE, Burkhart Q, et al. Sexual minorities in England have poorer health and worse health care experiences: a national survey. J Gen Intern Med. 2015;30:9–16.
34. Groves RM, Peytcheva E. The impact of nonresponse rates on nonresponse bias a meta-analysis. Public Opin Q. 2008;72:167–189.