Understanding the Role of Urology Practice Organization and Racial Composition in Prostate Cancer Treatment Disparities

PURPOSE:

Black men have a higher risk of prostate cancer diagnosis and mortality but are less likely to receive definitive treatment. The impact of structural aspects on treatment is unknown but may lead to actionable insights to mitigate disparities. We sought to examine the associations between urology practice organization and racial composition and treatment patterns for Medicare beneficiaries with incident prostate cancer.

METHODS:

Using a 20% sample of national Medicare data, we identified beneficiaries diagnosed with prostate cancer between January 2010 and December 2015 and followed them through 2016. We linked urologists to their practices with tax identification numbers. We then linked patients to practices on the basis of their primary urologist. We grouped practices into quartiles on the basis of their proportion of Black patients. We used multilevel mixed-effects models to identify treatment associations.

RESULTS:

We identified 54,443 patients with incident prostate cancer associated with 4,194 practices. Most patients were White (87%), and 9% were Black. We found wide variation in racial practice composition and practice segregation. Patients in practices with the highest proportion of Black patients had the lowest socioeconomic status (43.1%), highest comorbidity (9.9% with comorbidity score ≥ 3), and earlier age at prostate cancer diagnosis (33.5% age 66-69 years; P < .01). Black patients had lower odds of definitive therapy (adjusted odds ratio, 0.87; 95% CI, 0.81 to 0.93) and underwent less treatment than White patients in every practice context. Black patients in practices with higher proportions of Black patients had higher treatment rates than Black patients in practices with lower proportions. Black patients had lower predicted probability of treatment (66%) than White patients (69%; P < .05).

CONCLUSION:

Despite Medicare coverage, we found less definitive treatment among Black beneficiaries consistent with ongoing prostate cancer treatment disparities. Our findings are reflective of the adverse effects of practice segregation and structural racism, highlighting the need for multilevel interventions.

INTRODUCTION

There are racial disparities in prostate cancer incidence, treatment, and mortality that disproportionately affect Black men.¹ Although White men have a one in eight chance of being diagnosed with prostate cancer, Black men have a one in six chance of being diagnosed with prostate cancer, are less likely to receive definitive treatment, and are nearly twice as likely to die from the disease.² Although traditional speculation inferred racial differences in disease biology drive these disparities, research suggests that other factors may be driving the disparities.³

Indeed, the reasons underpinning disparities are complex.^4,5 However, the extent to which organization of urologic practices (eg, solo practitioner and multispecialty group) and whether the racial composition of their patient populations might affect treatment disparities is unknown. Better understanding whether practice organization and racial composition of patients within a practice (eg, mostly White v mostly Black patients) predispose to differences in prostate cancer treatment may add actionable nuance to prior access to care research. For example, practices with more Black patients may be more in tune with implications of undertreatment, be more culturally competent, more likely to create hospitable environments to mitigate mistrust, and ultimately more likely to treat Black men with prostate cancer.⁶ These practices are also most likely to be under-resourced with implications for quality of care.⁷ Similarly, multispecialty group practices might offer more treatment options, leading to greater treatment of Black men. These hypotheses are supported by work on unconscious bias and stereotyping from the Institute of Medicine and others as modifiable sources of health care disparities.^8,9

We examined the extent to which racial differences in prostate cancer treatment patterns among Medicare beneficiaries continue to exist in a contemporary cohort and were associated with urology practice organization and racial composition. Our findings may provide awareness and insights into the impact of urology practice characteristics and associated treatment disparities among Medicare beneficiaries.

METHODS

Data Sources and Study Population

We used a 20% sample of national Medicare data to conduct a retrospective cohort study of men newly diagnosed with prostate cancer between 2010 and 2015. The 20% Medicare representative sample is randomly selected and widely used for Medicare population estimates.¹⁰ Using previously validated methods, we identified patients with incident prostate cancer and followed them for at least one year to identify initial treatment within the first year after diagnosis.¹¹ We included Medicare beneficiaries eligible for Medicare Parts A and B to ensure accurate treatment ascertainment.

Practice Organization and Racial Composition

Because we were interested in whether the organization and racial composition of a urology practice influenced prostate cancer treatment trends, we first linked patients to their urologist using established methods.¹² Next, urologists were linked to practices by their tax identification numbers (TINs). We then assigned each TIN to one of four practice types: solo practice (SOLO), single-specialty group (SSG), large specialty group (LSG), and multispecialty group (MSG) using established methods.¹³ Finally, we linked each patient to a unique practice type according to their corresponding primary urologist's TIN.

Once we linked all patients to a practice, the practices were assessed for their overall proportion of Black patients, as found in Medicare Part B (carrier) records. Next, we rank ordered practices by the proportion of Black patients in their practice and divided practices into 4 quartiles on the basis of their fraction of Black patients with prostate cancer. This proportion was then linked back to the patients in the cohort and was our primary independent exposure variable.

Outcomes

Our primary outcome was definitive treatment at the practice and quartile level for patients diagnosed with incident (ie, newly diagnosed) prostate cancer. We defined definitive treatment as radical prostatectomy, radiation therapy, and radical prostatectomy combined with radiation, brachytherapy, and cryotherapy.¹⁴ We used International Classification of Diseases, Tenth Revision and Healthcare Common Procedure Coding System codes to identify patients with prostate cancer and treatments received. Patients who underwent active surveillance, watchful waiting, or primary ADT were classified as no treatment/expectant management, with rates of primary ADT proportioned from this group as an indicator of either incident locally advanced or metastatic disease or low-quality treatment of localized prostate cancer. Our secondary outcome was the treatment type by race.

Statistical Analysis

First, we identified differences in patient demographics, treatment type and rates, and practice organization by quartile using univariate statistics. Next, we examined treatment rates by race within each quartile. Finally, to account for clustering at the TIN level and because of the hierarchical nature of the data, we constructed a multilevel mixed-effects model to examine associations between treatment at the individual and practice level. Model 1 included fixed effects of age, comorbidity, socioeconomic status, and residential area (level one). Socioeconomic status was assessed at the ZIP code level and obtained using a summary score, including income, education, and occupation as described by Diez Roux et al.¹⁵ Model 2 included fixed effects of age, comorbidity, socioeconomic status, residential area (level one), and practice type and quartile (level two). The random intercept was TIN/practice. The type 3 test of fixed effects was performed to test for significance. Model comparison was performed using odds ratio estimates with 95% CIs. Least squares means and predicted probabilities were calculated for each model.

Sensitivity Analysis

We conducted sensitivity analyses to assess the effect of the race variable at the individual level and the practice (TIN) level to ascertain if these were different constructs. This was performed by constructing a multilevel mixed-effects model for quartile 4 only.

All analyses were conducted using SAS 9.4 software with the probability of a type I error set at 0.05, and all testing was two-sided. This study was deemed exempt by the University of Michigan Internal Review Board because of the deidentified nature of the data.

RESULTS

We identified 54,443 Medicare patients diagnosed with incident prostate cancer between 2010 and 2015, with 87% (n = 47,481) White and 9% (n = 4,908) Black patients assigned to 4,194 urology practices. Mean proportions of Black patients by practice quartiles are as follows: quartile 1: 1,288 practices, 1.1% Black patients (n = 150); quartile 2: 1,126 practices, 4.2% Black patients (n = 571); quartile 3: 825 practices, 8.8% Black patients (n = 1198); and quartile 4: 955 practices, 22.2% Black patients (n = 2,994). Across all urology practices, we identified an even broader range in the proportion of Black patients from 0% to 97.2% as illustrated in Figure 1. Within the highest quartile, we identified 246 practices with 25%-50% Black patients, 45 practices with 50%-75% Black patients, and 42 practices with 75%-97% Black patients. As shown in Table 1, patients with newly diagnosed prostate cancer in the practices that had a higher proportion of Black patients were slightly younger and had greater comorbidity burden and lower socioeconomic status.

FIG 1.

Variation in urology practice composition of Black men diagnosed with incident prostate cancer (African American men with prostate cancer by practice) in the Medicare program from 2010 through 2015.

TABLE 1.

Urology Practice Quartiles and Demographics According to Racial Composition

We identified a few notable differences in practice organization across the quartiles. First, SSGs cared for the majority of patients across quartiles. Second, LSGs were the least common regardless of quartile and collectively cared for approximately 10% of patients with incident prostate cancer. Third, we identified a somewhat bimodal distribution for solo practitioners. Most practices that had 75%-97% of Black patients were solo practitioners. However, solo practitioners were also common in the lower quartile practices serving the lowest proportion of Black patients.

We found that most White and Black Medicare beneficiaries with incident prostate cancer underwent definitive treatment within the first year after diagnosis. Across practice composition quartiles, approximately one third of men did not receive definitive treatment. As shown in Appendix Table A1 (online only), we identified variation in unadjusted treatment patterns and rates by race within and across quartiles. Across quartiles, Black patients with incident prostate cancer underwent definitive treatment less often than White patients. The rates of radiation therapy as primary treatment were higher for Black patients across all quartiles, with the most notable difference in the third quartile where nearly half of Black patients were treated with radiation therapy (45.2%). Black patients treated at practices in the highest quartile had the highest unadjusted rates of brachytherapy and cryotherapy. Finally, Black patients had slightly higher unadjusted rates of no treatment or expectant management across quartiles, with the highest rate of no treatment or expectant management in quartile 1. Within this treatment category, Black patients also received ADT monotherapy at slightly higher rates across quartiles, with the highest rate in quartile 1, although this did not reach statistical significance.

Table 2 shows the results of the multilevel mixed-effects logistic regression model. In model 1, which included fixed effects of age, race, comorbidities, socioeconomic status, and residential area, we found that Black patients with incident prostate cancer had lower odds of definitive treatment (adjusted odds ratio [aOR], 0.88; 95% CI, 0.82 to 0.94). Patients with medium socioeconomic status had higher odds of treatment than those with low socioeconomic status (medium v low, aOR, 1.08; 95% CI, 1.02 to 1.13), with no difference among patients with high socioeconomic status. In model 2, which included fixed effects from model 1 and fixed effects of practice type and quartile, we found that patients in MSG practices had lower odds of treatment for incident prostate cancer than patients in SOLO practices (aOR, 0.85; 95% CI, 0.79 to 0.91). Finally, patients in practices with higher proportions of Black patients (higher quartiles) had higher odds of receiving treatment than patients in practices with a lower proportion of Black patients (quartile 3 v 1, aOR, 1.11, 95% CI, 1.03 to 1.20; quartile 4 v 1, aOR, 1.11, 95% CI, 1.03 to 1.20). All fixed effects except residential area were significant for both models (P < .05).

TABLE 2.

Multilevel Mixed Effects Model for Factors Associated With Prostate Cancer Treatment

As shown in Appendix Table A2 (online only), least squares means for model 1 were 0.67 for Black patients and 0.80 for White patients (P < .01). The predicted probability of treatment was 66% for Black patients and 69% for White patients (P < .01). Least squares means for model 2, incorporating the practice type and quartile, were 0.66 for Black patients and 0.80 for White patients (P < .01). Predicted probability of treatment for model 2 remained the same at 66% for Black patients and 69% for White patients. Predicted probabilities of treatment were 70% for LSG and SOLO practices, 69% for SSG practices, and 67% for MSGs. Least squares means were highest for quartiles 3 and 4 (0.83) and quartile 2 (0.79) and lowest for quartile 1 (0.72; all P < .01). Predicted probabilities of treatment were 70% for quartiles 3 and 4, 69% for quartile 2%, and 67% for quartile 1. Our sensitivity analyses demonstrated that Black patients in quartile 4 still had 0.90 odds of treatment (95% CI, 0.81 to 0.99) compared with White patients.

DISCUSSION

In this study examining contemporary treatment patterns among Medicare beneficiaries with newly diagnosed prostate cancer across a variety of urology practice organizations and racial compositions, we found less definitive treatment among Black men in lower quartiles (ie, less diverse practices) and more definitive treatment among Black men in higher quartiles (ie, more diverse practices). We also found extraordinary differences in the proportion of Black patients served across urology practices ranging from all White to nearly all Black patient populations, indicating practice segregation. We discovered that SSGs delivered the majority of prostate cancer care regardless of racial practice composition and that solo practitioners had some of the most segregated practices. There were higher radiation and brachytherapy therapy rates and slightly lower surgery rates among Black patients compared with White patients, consistent with the previous literature.^16-18 This may be due to differences in information that patients receive regarding prostate cancer treatment rather than patient preferences.¹⁹ Black patients were more likely to be treated with cryotherapy, particularly in the highest quartile practices. This finding warrants further exploration since cryoablation is not recommended as a first-line treatment by the National Comprehensive Cancer Network, thus raising quality-of-care concerns.^20,21 After multivariable adjustment, Black patients still had lower predicted probabilities of definitive treatment. Taken together, these findings apply a unique practice-based lens to understanding and addressing racial differences in the treatment of incident prostate cancer among Medicare beneficiaries.

We observed a stark range in racial practice composition among urology practices. There were practices that care for up to 97% Black patients, and there were practices that do not take care of any Black patients with incident prostate cancer, reflective of practice segregation. Overall, definitive treatment for Black Medicare beneficiaries with incident prostate cancer was largely similar to that for White Medicare beneficiaries, although lower in every practice context, and all patients seen in the higher quartile practices had higher odds and predicted probabilities of definitive treatment. This may be due to more cultural competency and targeted communication to mitigate mistrust in practices, which treat more Black patients. Alternatively, there may be a greater recognition of disparate outcomes among Black patients, prompting justifiable use of more definitive treatment. On the other hand, practices with fewer Black patients may not be focused on mitigating the risks of undertreatment of Black men. There could be a lack of interpersonal skills or underappreciation of possible unconscious bias in their counseling patterns.^22,23 Although there could be more active surveillance among these practices, it is unlikely that Black men with incident prostate cancer would systematically have lower-risk disease, qualifying them for this management.²⁴ In addition, there were slightly higher rates of ADT monotherapy, indicating either more advanced disease at presentation or poor localized disease treatment choices, neither scenario favoring Black patients with incident prostate cancer. Taken together, our findings highlight an important contextual consideration with respect to prostate cancer treatment recommendations and decision making.

Despite Medicare coverage, we found overall less definitive treatment among Black Medicare beneficiaries and lower predicted probabilities of treatment consistent with ongoing prostate cancer treatment disparities. Although the effect of individual race is constant, the effect of quartile varies depending on the quartile. Regardless, the gaps in definitive treatment by race among Medicare beneficiaries appear to be improving. For example, several studies from the 1990s and 2000s indicate larger gaps in treatment disparities.^3,25,26 Although we demonstrated 3% less predicted probability of definitive treatment and 0.88 odds of treatment among Black patients compared with White patients, treatment differences in a previous SEER-Medicare study indicated 0.74 odds of treatment for Black compared with White Medicare patients.²⁵ However, the gaps are likely worse among all-payer studies, with a recent Massachusetts study indicating that Black patients had a 0.78 odds of definitive treatment from 2004 to 2015 compared with White patients.²⁷ Among patients with insurance, a previous SEER study found that insurance coverage was associated with reduced disparity in receipt of definitive therapy.²⁸ Although Medicare treatment trends may be heading in the right direction, our findings indicate the need for ongoing awareness and mitigating multilevel strategies to work toward equity in prostate cancer care.

Among the nearly 1,000 urology practices in the quartile that treated the greatest proportion of Black men, we found the greatest degree of low socioeconomic status, higher comorbidity, and importantly, men diagnosed at younger age compared with practices in the lower three quartiles. Taken together, factors demonstrated in our study beyond health insurance are manifested by the segregation of practices and variation in practice patterns among this group of urology practices and their patients, reflecting a broader construct of structural racism.²⁹ The pervasive nature of these social, organizational, and economic phenomena and its adverse effects on health are supported by these study findings. More broadly, our findings may be viewed as a microcosm of hospital segregation, a proxy for structural racism,²⁹ whereby longstanding racial differences exist in access to high-quality providers, specialized services, and distort practice patterns.^30-33 We expand this concept to the outpatient clinic and measure the extent of segregation in urology clinics by estimating the fraction of Black patients with prostate cancer seen at a given clinic. Although hospital integration is associated with decreasing health disparities,³⁴ what this means for less diverse specialty care practice settings is complex requiring further exploration.

This study should be considered in the context of several limitations. First, we only included Medicare beneficiaries in this study and there may be differences in treatment of Black patients from other payers, such as commercial payers or those without any insurance coverage. Despite this, practices that accept Medicare often accept commercial payers as well and it is unlikely that Medicare patients would receive different treatment from a patient with commercial insurance. Second, our data lack information about cancer severity as a driver of definitive treatment. There are several reasons why this is likely to have a limited impact on our findings. First, a majority of patients with incident prostate cancer have localized disease.³⁵ Second, at a population level, there is little reason for cancer severity to vary among practices in favor of less definitive treatment of Black men.³⁶ Third, since Black patients have higher incidence, worse prostate cancer severity at diagnosis, and higher risk of dying from prostate cancer,² Black patients should have higher or at least equivalent definitive treatment rates compared with White patients, indicating that our study may actually underestimate gaps in treatment rates. Fourth, we acknowledge that given lack of cancer severity data, there may be an unintended imbalance in disease severity distribution in this study sample, favoring decreased localized treatment as it pertains to a given practices' racial composition and with respect to ADT monotherapy use, indicating either delayed presentation or poor-quality localized treatment. Nonetheless, we used a 20% Medicare representative sample, which is randomly selected and used for Medicare population estimates.¹⁰ In addition, we used multivariable and multilevel statistical adjustments to adjust for differences in social determinants of health (eg, SES) that may be associated with cancer severity. This 20% Medicare sample is thus largely representative of population-level trends. Although practice patterns might have changed since our study period, outside of large-scale health policy changes, it is unlikely that there are subsequent major impacts to practice segregation and implications of the study. The data from this study were from 2010 to 2015. Although treatment disparities could be improving in more recent data, racial disparities in prostate cancer still exist as evidenced by robust ongoing debate on how to mitigate them.^37,38 A final limitation is inclusion of men age 65 years and older. Despite this, more than half (57%) of patients diagnosed with prostate cancer are 65 years and older, indicating that our findings are reflective of most prostate cancer care in the United States.³⁹

With these limitations notwithstanding, our findings have implications for patients, providers, and policymakers. For patients, we present ongoing disparities in treatment. Although some studies suggest that insurance coverage may reduce prostate cancer death among Black patients,^28,40 our finding of lower treatment rates among Black Medicare beneficiaries compared with White beneficiaries indicates that insurance coverage alone is insufficient to mitigate treatment disparities. Engagement, feedback, and patient-driven interventions are important for progress as Black patients are less likely to report high-quality care and good physician-patient communication.⁴¹ There is also evidence of benefit for noninsurance-related social services including community outreach, which can improve health literacy, utilization, and outcomes.⁴² For providers, interventions targeted at communication, culturally sensitive care, mitigating conscious and unconscious bias, and partnership with community health care workers are key.^43-47 Structural racism operates on a macrolevel with a greater influence compared with patient-level factors.²⁹ Although our contemporary findings indicate a narrowing gap in racial disparities in prostate cancer treatment among Medicare beneficiaries, the path forward requires a joint commitment to transforming and challenging policies perpetuating historical ideologies and persistent disparities. In other words, advocacy for policies and systems improves and promotes equity for affected communities.

In conclusion, we demonstrate ongoing disparities in prostate cancer treatment among Medicare beneficiaries, with a lens into the impact of racial composition of a practice. Although Black patients have higher rates of diagnosis and mortality from prostate cancer, Black Medicare beneficiaries had lower likelihood of definitive treatment, as did those with lower socioeconomic status, who were more often Black. Despite a disparity in every quartile, Black patients in higher quartiles received more treatment and perhaps better care. Practice segregation and structural racism are possible factors helping explain these findings. Evidence-based and multilevel interventions are needed to continue toward racial equity in prostate cancer care, including through evaluation and restructuring of the health and social care systems and policies, and cultural competency to enhance patient-provider interactions.

APPENDIX

TABLE A1.

Incident Prostate Cancer Treatment Rates by Urology Practice Racial Composition Quartile

TABLE A2.

Least Squares Means

AUTHOR CONTRIBUTIONS

Conception and design: Nnenaya Agochukwu-Mmonu, Brent K. Hollenbeck, Ted A. Skolarus

Financial support: Brent K. Hollenbeck

Administrative support: Danil Makarov, Brent K. Hollenbeck

Provision of study materials or patients: Brent K. Hollenbeck

Collection and assembly of data: Nnenaya Agochukwu-Mmonu, Brent K. Hollenbeck, Ted A. Skolarus

Data analysis and interpretation: All authors

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Understanding the Role of Urology Practice Organization and Racial Composition in Prostate Cancer Treatment Disparities

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = immediate family member, Inst = my institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/op/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).