Identifying disparities in patient-centered care experiences between non-Latino white and black men: results from the 2008-2016 Medical Expenditure Panel Survey

Background Patient-centered healthcare in the context of a medical home (PCMH) is an important pathway to reducing healthcare inequities. To date, no work has examined the prevalence of care experiences associated with PCMH among non-elderly Black males. Methods We analyzed data, on 22 indicators representative of six healthcare domains associated with PCMH experiences, from non-Latino White (NLW) and Black males aged 18–64 from the 2008–2016 Medical Expenditure Panel Survey (n = 47,405). We used generalized linear models to test whether Behavioral Model factors attenuate any differences in access to these domains between NLW and Black males, and decomposition techniques to examine the contribution of these factors to reported differences. Results Black males reported 1) lower access to personal primary care providers, 2) poorer quality communication with providers, and 3) lower levels of care comprehensiveness (all p < 0.05). Differences between groups were attenuated but not eliminated by accounting for the Behavioral Model factors particularly through enabling and predisposing factors. Group health characteristics were not a primary driver of racial differences in care experiences across all the considered domains. Conclusions Black men, in the U.S, continue to face barriers to accessing high quality, patient-centered care, specifically as it relates to accessing specialty care, medical tests, and patient-provider communication.

four common core beliefs: holistic, individualized, respectful, and empowering [4] . Research on PCC has focused on the philosophy of care as well as the implementation across healthcare sectors.

Black Americans and PCC
The interaction between patients and their healthcare providers, namely physicians, is a powerful indicator of patient outcomes, including satisfaction, adherence to treatment recommendations, and likelihood of returning for follow-up care [2,4,5]. Black Americans rate interpersonal care received by physicians more negatively than their White counterparts [6] and studies have found that Blacks receive poorer interpersonal communication via lower levels of "rapport building and use of affective tone", heightened levels of physician verbal dominance, less patient-centeredness, and shorter medical visits [7,8]. Furthermore, Blacks and other racial/ ethnic minorities are less likely to receive referrals for medical specialists, to have a consistent usual source of care, to have comprehensive health insurance or to trust the healthcare system; and they experience greater difficulties in navigating the healthcare system [9,10]. Studies measuring Black patients' trust of physicians also find that high levels of mistrust are associated with less medication adherence and physician communication [11]. However, research on how Black men in particular, are served by patient-centered models of care, given the burden of health disparities they face, is insufficient. For example, Black men have one of the shortest life expectancies in the U.S and disproportionately high prevalence rates for several life-threatening or disabling diseases such as kidney failure, heart disease, cancer and hypertension [12,13] when compared to their White counterparts. These disparate outcomes have been partially linked to individual behaviors amenable to change, and factors such as low socio-economic status, masculine gender norms that discourage medical help seeking, and limited health literacy [14,15]. However, poor health care access and quality remain critical to perpetuating health disparities for this population, and improved health communication and longer medical visits are associated with higher trust in physicians and relatedly, greater reported quality of healthcare [7].

PCC and patient-centered medical homes
Patient-Centered Medical Homes (PCMH) arose in response to the need to implement patient centeredness within a constant care setting [16]. Patient-centered medical homes physically co-locate healthcare services and reduce transportation barriers that might normally make it difficult for patients to access care across multiple outpatient locations [3,16]. PCMH make use of team-based medical services whereby the patient's primary care physician leads a team of other health care providers focused on their specific needs; care is coordinated throughout the healthcare system and community, consequently improving access to different aspects of care. Ideally, patients receive same day appointments and more direct access to physicians and specialists [17,18] . This model looks to improve healthcare access and quality while minimizing healthcare disparities. The current study considers how healthcare indicators of a PCMH model is experienced by Black men in relation to non-Latino White (NLW) men in a nationally representative data set.

Sample
We used data from the 2008-2016 Medical Expenditures Panel Survey (MEPS). The MEPS is a comprehensive survey on health services use and expenditures for the non-institutionalized, civilian United States population. In addition to the collected data on use and expenditures, the MEPS include detailed information on participant demographics, insurance, and health status and conditions. The MEPS analytic subpopulation utilized for the current study were eligible adult males 18-64 years of age who self-reported their race as non-Latino White or Black (n = 47,799). We excluded from our analyses participants with missing data on any of our model covariates (n = 394) for a final unweighted analytic sample of n = 47,405; the average weighted yearly equivalent of 59.7 and 11.1 million adult NLW and Black males, respectively. The MEPS has a complex survey design including stratification, clustering, and probability weighting. All analyses accounted for the complex design of the MEPS to ensure appropriate generalization and inferences to the intended target population. Specifically, we used a Taylor Series Linearization approach to variance estimation as implemented through the survey functionalities in the Stata v.15 software. All data are publicly available from the Agency for Healthcare Research and Quality (AHRQ). Study protocol qualified for "Exemption" according to the Department of Health and Human Services Code of Federal Regulations.

Outcomes
MEPS participants were surveyed in detail about their healthcare experiences. We focused on a set of questions that relate to 6 domains covering the availability of: 1) a personal primary care provider; 2) enhanced access to care; 3) quality patient-provider communication 4) patient-centered care; 5) coordination by usual source of care provider; and 6) comprehensive care. Each domain was assessed using multiple questions. For example, to investigate enhanced access to care we examined four questions that probed whether 1) the participant had easy phone access to provider during regular business hours; 2) the provider had office hours on nights or weekends; and 3) how easy it was to contact the provider during off hours if needed. Our approach to approximating domains of PCMH care have been used previously in MEPS data [19][20][21][22][23]. In line with prior work, all questions, when not dichotomous, were dichotomized to group "never" and "sometimes" responses into one category and "usually" and always as a second category. With the exception of the usual source of care provider indicator, all considered outcomes were modeled conditional on reporting a usual source of care. A detailed list of the questions, skip patterns, and the original scaling, used to assess each of the six domains listed above is provided in Additional file 1: Figure S1.

Primary exposure
Self-reported racial grouping was the primary predictor of interest. Our focus was on comparing male participants reporting their race as White and ethnicity as non-Latino ethnicity to those reporting their racial background as Black.

Covariates
We followed a modified behavioral model approach [24,25] to model care experiences as a function of: 1) predisposing factors: race (our primary outcome), age (in years), marital status (0 = not currently married; 1 = married), and a 4-category education indicator (1 = Less than high school (HS), 2 = HS or equivalent, 3 = Some college, and 4) College or more) 2) enabling factors: a 5-category poverty status, based on Census classification as coded by the AHRQ, indicator (1 = Poor, 2 = Near-Poor, 3 = Low-Income, 4 = Middle-Income, 5 = High-Income), and insurance (1 = Private, 2 = Public, 3 = Uninsured); and 3) need factors: which included two measures of self-reported physical and mental health each measured as a 5-cateogry indicator (1 = Excellent, 2 = Very good, 3 = Good, 4 = Fair, and 5 = Poor). We also controlled for two additional variables that confound access: region of residence and survey year to account for yearly fluctuations in factors that are not directly accounted for by our model.

Analytic plan
Our analyses proceeded in four steps. First, we provide descriptive statistics to characterize our sample overall and by racial group by the covariates of interest (Table 1). Second, we model the binary outcomes for each domain using survey logistic regression. For each outcome we follow a 3-step process beginning with crude (bivariate) associations (outcome as function of racial grouping), and incrementally adjusting for 1) age and 2) full set of covariates. We present a summary of the estimated models to describe differences between Blacks and NLWs in Table 2. We plot the odds ratios and 95% confidence intervals resulting from these models in Fig. 1a and provide a detailed representation of the estimated models in the Additional file 1: Tables S1-S6. Third, to facilitate the interpretation of our estimates we calculate and plot contrasts of the estimated average marginal probabilities resulting from these incremental adjustments for each outcome in each domain and their estimated 95% confidence intervals in (Fig. 1b). Fourth, to implement the institute of medicine definition of disparity, we use modified Oaxaca techniques for binary outcomes to decompose the differences in outcomes prevalence into their explained and unexplained portions. By using decomposition techniques, we can quantify the degree to which distributional differences in sets of covariates (e.g. enabling factors) across the two race groups of interest contribute to the observed (explained) differences in the prevalence of each healthcare experience outcome considered. Accessible detailed discussions of the Blinder-Oaxaca technique and its binary outcome modification have been published elsewhere [26] . We graph the decompositions for each outcome where significant differences between Blacks and NLWs are detected in Fig. 2 and present a summary of the differences and estimated explained and unexplained differences in Table 3. Detailed presentations of these decompositions are included in Additional file 1: Table S7.

Results
Descriptive characteristics are detailed in Table 1. Black males were on average younger than NLWs, less likely to report being married (35.7% vs 56.6%), and half as likely to report having a college education or more (15.9% vs 32.4%). Black males were also twice more likely to report being poor or near poor, half as likely to be classified as high-income, 2.3 times more likely to report being publicly insured and 1.8 times more likely to report being uninsured relative to NLW male counterparts. While differences in reported physical and mental health were statistically significantly different, differences were not substantively large.

Group differences
Interesting differences in healthcare experiences were revealed in the current study ( Table 2). The crude, age adjusted, and fully adjusted odds ratios and their 95% confidence intervals are presented in Additional file 1: Tables S1-S6. To facilitate visualization of our results, the crude and fully adjusted odds ratios and their 95% confidence intervals are presented in Fig. 1, and the estimated differences in marginal probabilities for each outcome between Black and NLW males and their 95 confidence intervals. The crude estimates revealed that Black and NLW males reported similar experiences on measures of enhanced access to care, care coordination, and, with the exception of differences in experiences related to decisions related to help on treatment options, on patient-centeredness of care. Black and NLW males differed on reports of access to personal primary care, quality patient-provider communication, and comprehensiveness of care. Substantial proportions of the reported differences were unexplained by the model covariates. The absolute and crude differences on the underlying measures with statistically significant differences are reported in Table 3.
Explained differences in primary care access (Fig. 2) Adjusting the models for the covariates explained 60.9, and 47.9% of the differences in, respectively, reported  Figure S1.
access to a usual provider of care, and the provider not being a specialist. Model covariates, however, explained a very small proportion (8.7%) of the difference in reporting a personal provider. The detailed decomposition indicated that the majority of the explained disparity in access to a usual provider was due to characteristic differences, between NLWs and Blacks, in predisposing (17.8%) and enabling (34.4%) factors. These factors also, respectively, explained 37.9 and 12.0% of the differences in reporting a non-specialist provider and the group differences on this indicator became statistically insignificant.

Explained differences in provider quality communication
Adjusting the models for the covariates explained 46.8, and 69.5% of the differences in indicators of quality patient-provider communication; spends enough time with the patient, and explain issues in an easy to understand manner, respectively. For both measures the majority of explained differences stemmed from advantageous predisposing (11.9 and 21.8%, respectively) and enabling characteristics (23.2, and 27.8% respectively) among NLW males. Despite the above reported attenuation in differences after controlling for covariates the differences between NLW and Black males remained statistically significant.

Explained differences in comprehensiveness of care
Finally, there were marked differences in timely access to comprehensive care. Adjusting the models for the covariates explained 35.3, 34.5, 40.1, and 18.9% of the differences in indicators of care comprehensiveness; immediate, specialist, tests or treatment, and routine care, respectively. For all these measures the majority of explained differences stemmed from advantageous enabling factors (30.5, 22.2, 26.2, and 9.8%, respectively) and, to a lesser extent, predisposing characteristics among NLW males. The differences between NLW and Black males remained statistically significant despite the above reported attenuations resulting from covariates adjustment.

Discussion
In the current study, we examined racial differences in men's experiences of patient-centered care broadly, and with defining features of the patient-centered medical home model in particular. Few studies have extracted a nuanced understanding of patient-centered experiences for Black men, particularly with regard to care coordination and provider interactions. Our analysis revealed distinct differences in health care access, patientprovider interactions, and comprehensiveness of care and care coordination between NLW and Black males. Overall, differences between groups were attenuated but  For detailed information about survey questions for indicators and domains and logical skips leading to these differential Ns see detailed information provided in Additional file 1: Figure S1. Absolute difference is calculated based on the expected average crude probabilities as: 100*(Pr(non-Latino Whites) -Pr(Blacks)). The relative difference is calculated as (1-(Pr(Blacks)/Pr(non-Latino Whites))*100. Only measures with absolute difference ≥ 2% are presented in the Table   a An "Explained" % that exceeds 100 indicates that Blacks would have an advantage on that indicator after controlling for the factors included in the model not eliminated by accounting for the Behavioral Model factors. Notably, and reflecting the IOM definition of disparities, group health characteristics were not a primary driver of racial differences in care experiences across all these domains.

Access to primary care providers
Black men were less likely than their NLW counterparts to have access to a usual or specific health care provider that was not a specialist. This is particularly concerning given that the vast majority of routine preventive health care and management of common chronic illnesses such as diabetes, hypertension, and mental health occur in a primary care context. Further, the Black men in this study were more likely to be of lower socioeconomic status (SES) as marked by less education, lower income, and being publicly insured. Recent studies have demonstrated that when primary care practices implement patient-centered medical home models, they are uniquely effective in raising preventive services utilization rates among low SES patients, including screening for colorectal cancer which disproportionately impacts Black men [27] . Black men's utilization of preventive health services has been well studied but research on Black men's general lack of connectedness to the primary care context is still insufficient. Specifically, having a relationship with a single consistent health care provider or team who can track health trajectories, symptoms, and manage care over time plays a significant and direct role in reducing disparities. This is also important because patients connected to a specific physician are more likely to receive care consistent with clinical guidelines [28]; and continuity of care is also strongly associated with fewer emergency department visits, inpatient hospitalizations, fewer complications from treatment, and lower health care costs for heart disease, chronic obstructive pulmonary disease, and diabetes [29].

Quality of patient-provider communication
In this study, Black men who reported a usual source of care were less likely to perceive that their health provider spends enough time with them, and to explain issues in a way that they can understand. However, these differences were not very large. These findings are consistent with research showing that as a result of implicit bias, some physicians are less patient-centered in their communication style and content with their Black than with their White patients [8,30] . Despite the small effects, the differences in communication quality combined with a higher perception among Black men that providers are less helpful in making treatment decisions have the potential to perpetuate historical mistrust, creating a synergistic dissuading effect on establishing a personal and continuous source of care and engaging in routine interactions with the health care system. High quality patient-provider communication empowers patients to more fully participate in their own care decision-making [31]. Empowering patients is also one of the central tenets of patient-centered care and the foundation upon which patient-centered medical homes models are built.

Comprehensiveness of care
Finally, Black men in the current study experienced poorer outcomes than their NLW counterparts related to their access to both immediate and routine health care, access to medical tests and specialty care. Unmet subspecialty care needs contribute to the disparities experienced by medically underserved populations such as Black men. Generally, research on care comprehensiveness needs, barriers, adherence to care management plans and related decision-making by or on behalf of Black men is virtually nonexistent. Patient navigation interventions have shown particular promise for improving the coordination and completion of cancer screening for Black men specifically [32]. However, evidence is lacking on similar efforts to support the prevention or treatment of other chronic health conditions with high or rising prevalence in this population. Black men would undoubtedly benefit from greater support and advocacy while navigating the health care system. Specifically, systemic care coordination efforts should take into consideration unique needs related to age, health literacy, socioeconomic status, insurance gaps, and other layers that account for the impact of environmental context and identity on health outcomes.

Limitations
We point to three major limitations to this work. First, all measures used in our analyses are based on self-reported survey data. Self-reports are subject to recall and other biases which could have affected our estimates. Sensitivity studies by AHRQ indicate good response validity for MEPS data [33]. Second, our measures do not cover the entire spectrum of the operational definitions of a patient-centered care or the medical home and its domains. We used indicators as determined and published in previous work on patient centered care in MEPS. More expansive definitions, using other data sources, can potentially uncover patterns and differences not attended to in this work. Finally, we used a modified version of the Behavioral Model given available data in the MEPS. As such, several measures (e.g. contextual and health system level factors) were omitted, which explains the high levels of unexplained differences reported in our study. Any inferences about the contribution of these critical factors are omitted from our discussion and would require further research.

Conclusions
Consensus is growing that patient-centered medical home models of delivering care are on balance, positive for patient outcomes and potentially reduce health disparities. They have been increasingly integrated across healthcare settings as the "gold-standard" of care. Our findings show that Black men, particularly those who are socioeconomically disadvantaged, are not equitably experiencing the core standards of this model and point to specific areas where Black men are not being adequately served by the health care system. Our findings suggest that health care team members, including community advocates, may need to better support Black men in connecting to a primary care physician, and once connected, in coordinating their comprehensive care and communicating their questions and concerns to optimize their medical care and patient experience.
Additional file 1: Figure S1. Detailed list of the questions, and the original scaling, used to assess each of the six domains of health care experiences.