Racial disparities in urologist visits among elderly men with prostate cancer: A cohort analysis of patient-related and county of residence-related factors
Eberechukwu Onukwugha PhD,
Pharmaceutical Health Services Research Department, School of Pharmacy, University of Maryland, Baltimore
Corresponding author: Eberechukwu Onukwugha, PhD, Pharmaceutical Health Services Research Department, University of Maryland School of Pharmacy, 220 Arch St, Baltimore, MD 21201; Fax: (410) 706-5394; firstname.lastname@example.org
We thank Pharmaceutical Research Computing and Candice Yong for programming assistance on the primary data sets. The current study used the linked Surveillance, Epidemiology, and End Results (SEER)-Medicare database. The interpretation and reporting of these data are the sole responsibility of the authors. We acknowledge the efforts of the Applied Research Program of the National Cancer Institute; the Office of Research, Development and Information of the Centers for Medicare and Medicaid Services; Information Management Services, Inc; and the SEER Program tumor registries in the creation of the SEER-Medicare database. The ideas and opinions expressed herein are those of the author(s) and endorsement by the State of California Department of Public Health, the National Cancer Institute, and the Centers for Disease Control and Prevention or their contractors and subcontractors is not intended nor should be inferred.
Factors contributing to the lower likelihood of urologist follow-up among African American (AA) men diagnosed with prostate cancer may not be strictly related to patient factors. The authors investigated the relationship between crime, poverty, and poor housing, among others, and postdiagnosis urologist visits among AA and white men.
The authors used linked cancer registry and Medicare claims data from 1999 through 2007 for men diagnosed with American Joint Committee on Cancer stage I to III prostate cancer. The USA Counties and County Business Patterns data sets provided county-level data. Variance components models reported the percentage of variation attributed to county of residence. Postdiagnosis urologist visits for AA and white men were investigated using logistic and modified Poisson regression models.
A total of 65,635 patients were identified; 87% of whom were non-Hispanic white and 9.3% of whom were non-Hispanic AA. Approximately 16% of men diagnosed with stage I to III prostate cancer did not visit a urologist within 1 year after diagnosis (22% of AA men and 15% of white men). County of residence accounted for 10% of the variation in the visit outcome (13% for AA men and 10% for white men). AA men were more likely to live in counties ranked highest in terms of poverty, occupied housing units with no telephone, and crime. AA men were less likely to see a urologist (odds ratio, 0.65 [95% confidence interval, 0.6-0.71]; rate ratio, 0.94 [95% confidence interval, 0.92-0.95]). The sign and magnitude of the coefficients for the county-level measures differed across race-specific regression models of urologist visits.
Prostate cancer is the most common cancer diagnosed among men in the United States, and is a particularly significant problem in the African American (AA) population. Compared with white men, AA men exhibit higher rates of prostate cancer incidence, more advanced stage of disease at diagnosis, and higher disease-specific mortality.[1-3] Such disparities also extend to specialist follow-up visits. Specifically, older AA men with incident advanced-stage prostate cancer are less likely to visit a urologist after diagnosis compared with older white patients. To the best of our knowledge, to date limited attention has been paid to the role of community-level factors in explaining disparities in prostate cancer process measures such as urologist visits. Yet the patient's social environment may affect access to health care and impact health outcomes.[5-7] Prior studies focused on survival have indicated that an unequal burden of prostate cancer can arise if community-level poverty independently correlates with health status and AA individuals are generally more likely to live in poor communities. In addition, differences in the observed prostate cancer-related outcomes can arise if the experience of poverty differs between AA and white men living within the same community. Building on this reasoning, we investigated whether AA and white men differ with regard to the poverty-related metrics that characterize their community, whether community-level poverty is associated with the probability of a postdiagnosis urologist visit, and whether the relationship between community-level poverty and the urologist visit depends on race/ethnicity.
We used a broader conceptualization of poverty than has been used in most previous studies of prostate cancer. We used the following 5 domains from the English Indices of Multiple Deprivation (IMD)10: Income, Education, Access to Services, Living environment, and Crime. These 5 domains were used to characterize the individual's social environment and build the conceptual framework for the current study. The framework (Fig. 1) highlights the role of the social environment as an independent factor explaining the probability of a follow-up urologist visit after a diagnosis of prostate cancer. A visit to the urologist after a diagnosis of prostate cancer is affected by residence in an area that is ranked low in terms of the overall household income levels, educational attainment, access to services (health and social), housing quality, and/or safety. Figure 1 also indicates that there are potential differences in the contributing role of social environmental factors between AA and white men. These differences are due in part to differences between AA and white men in the likelihood of living in a deprived area and in their experience of living conditions.
There is theoretical motivation for considering the social experiences of AA men separately from that of white men. Feminist sociological theorists propose a perspective known as “intersectionality” to describe how systems of inequality (race/ethnicity, class, and sex) overlap to produce unequal outcomes in society. For example, the intersectionality perspective would explore sex differences within racial groups and racial differences within sex groups, as well as class. The intersectionality perspective underscores the value of investigating the experiences of AA men separately from the experiences of white men. AA men are unique due to their dual position in an advantaged group (men) and a disadvantaged group (racial minority). As a result of this dual position, the “compounding effect” of deprivation (due to residence in an area in which a majority of individuals are experiencing deprivation) may be further intensified for AA men compared with white men. If area-level deprivation is negatively associated with the probability of a postdiagnosis urologist visit (and AA men are more likely to live in deprived areas), the negative relationship between area-level deprivation and the probability of a visit may be intensified for AA men compared with white men.
The current study posited that the relationships between area-level deprivation and the likelihood of a urologist visit depend on race/ethnicity. Using American Joint Committee on Cancer (AJCC) stage I to III prostate cancer as a model, we investigated the role of patient-related and community-related factors with the long-term goal that the identification of differential effects may inform the development of patient-centered programs focused on addressing barriers to care.
MATERIALS AND METHODS
This retrospective analysis of linked cancer registry and Medicare claims data examined the treatment of men aged ≥66 years who were diagnosed with incident prostate cancer between 2000 and 2005 as listed in the Surveillance, Epidemiology, and End Results (SEER) cancer registry. Cases were limited to those men diagnosed with stage I to stage III prostate cancer as identified by the 3rd edition of the AJCC TNM classification and with continuous Medicare Parts A and B coverage for the 13 months before and including the SEER diagnosis month. Patients were required to have 12 months of follow-up data to be included in the sample. Treatment-related data from 1999 to 2007 were extracted from linked Medicare claims files. Exclusion criteria included: 1) health maintenance organization (HMO) enrollment during the 12 months before and including the month of diagnosis because HMO claims can be unreliable due to missing data; 2) a history of other cancers within 5 years before the diagnosis of prostate cancer; and 3) a prostate cancer diagnosis made at autopsy. Patients were censored if they enrolled in an HMO or lost coverage at any time after the diagnosis date, or if the end of the study period (December 2007) was reached. The SEER-Medicare data set was augmented with data from the USA Counties data set and the County Business Patterns data set.
The main outcome variable was a dichotomous variable modeling a postdiagnosis urologist visit occurring within 12 months of diagnosis. All claims, including Medicare Provider Analysis and Review, Carrier claims (National Claims History), and outpatient claims, were examined to identify evaluation and management records as well as the Health Care Financing Administration specialty code. Physician specialty was identified using the Health Care Financing Administration specialty codes provided in Medicare claims. The key independent variable used in the urologist visit models was the individual's race/ethnicity as documented in the SEER data set. The following potentially confounding measures (variables) at baseline were included in the urologist visit models: age group at diagnosis (reference category [RC], 66-69 years), marital status (RC, not married), poorly or not differentiated tumor (RC, well-differentiated or moderately differentiated tumor), positive number of months of state buy-in (RC, 0 months of state buy-in), comorbid conditions based on the Charlson comorbidity index (RC, score of 0), visit to a primary care physician (RC, no visit), a single proxy measure for poor performance status, and living area (RC, rural living area). The regression models also controlled for census tract-level measures of median income and percentage speaking English.
County-level measures were based on the English IMD, which include 7 domains. Two domains from the English IMD (Employment and Health) were assumed to be less relevant for a study involving a population of Medicare-insured individuals diagnosed with prostate cancer. The remaining 5 domains were used: Income, Education, Access to Services, Living environment, and Crime. We also included a county-level measure of the number of vehicles available at the household level. Continuous variables were converted into binary variables; cutpoints were created at the 75th or 25th percentiles for all individual variables entered directly in the regression models and for all variables that were considered in factor analyses. Some variables likely did not reflect a deprived state and were reverse-coded so that values > the 25th percentile represented the reference group (ie, medical and social services [including nonfederal physicians, total physicians, and health care and social assistance establishments]).
Statistical Analysis: County-Level Data Set
Exploratory factor analysis (EFA) was used to combine multiple domain indicators into a smaller number of conceptually and theoretically relevant factors. EFA was conducted for 3 domain areas: Crime, Facilities, and Access to Services. Several competing EFA models were developed and compared for each domain. Final models were chosen based on 3 criteria: 1) percentage variance explained; 2) achieving a “simple solution”; and 3) interpretability of the factors. Achieving a simple solution meant that each indicator only loaded on a single factor. Competing measurement structures for indicators were considered for quantifying deprivation; the 3 types of structures evaluated were continuous-level data, rank-ordered data using quintiles, and binary data using a 75th or 90th percentile. EFA models were evaluated using each type of data structure. The 90th percentile binary classification structure yielded the best EFA model based on the criteria outlined above. This measurement structure indicates whether counties fell in the top 10% of highest deprivation or the lower 90% for each indicator.
Statistical Analysis: Patient-Level Data Set
Descriptive statistics were calculated for the AA and white subgroups. Statistical analyses examined the association between AA race/ethnicity and the probability of a postdiagnosis urologist visit using logistic and “modified Poisson” regression to provide adjusted odds ratios (ORs) and rate ratios (RR). Race-specific regression models were estimated for the AA and white sample. Multilevel hierarchical logistic models were estimated to calculate the percentage of variation in urologist visits that is attributable to county-level variation for the entire group as well as for the AA and white subgroups. The median OR was reported, providing an OR interpretation for the effect size associated with the county-level residual.
The cutoff value for statistical significance was .05. All statistical analysis was conducted using SAS statistical software (version 9.1.3; SAS Institute Inc, Cary, NC). This study was approved by the University of Maryland Baltimore Institutional Review Board (HP-00042760).
Application of the inclusion criteria resulted in 65,635 patients diagnosed with incident AJCC stage I to III prostate cancer between 2000 and 2005. The average age at the time of diagnosis in the sample was 74 years, and 87% of patients were non-Hispanic white and 9.3% were AA. Overall, 16% of men diagnosed with stage I to III prostate cancer did not visit a urologist within 1 year after diagnosis (22% of AA men and 15% of white men). Descriptive statistics stratified by non-Hispanic white and non-Hispanic AA race (Table 1) highlight differences between AA and white men in their counties of residence. Additional detail regarding the specific county-level measures that contributed to each domain is available upon request.
Table 1. Descriptive Statistics by Race for Patient and County Characteristics
At least 1 urologist visit within 1 y after diagnosis
Missing CCI score
Hospitalization, walking aid, SNF, oxygen use, or wheelchair use in 12 mo before study period
Visit to primary care physician 12 mo before diagnosis
Percentage of population living below poverty level for 1999 (1 if >75th percentile)
Educational attainment among those aged ≥25 y completing <9th grade for 2000 (1 if >75th percentile)
Occupied housing units with no vehicles available for 2000 (1 if >75th percentile)
Occupied housing units with no telephone service available for 2000 (1 if >75th percentile)
Multivariable Regression Models
Using the full sample, AA race was associated with a statistically significantly lower likelihood of urologist visits compared with white patients (OR, 0.65 [95% confidence interval (95% CI), 0.6-0.71]; RR, 0.94 [95% CI, 0.92-0.95]). Results from race-specific modified Poisson regression models are reported in Table 2. In the AA sample, residence in a county with a high number of individuals living below the poverty level was found to be positively associated with a postdiagnosis urologist visit (OR, 1.75 [95% CI, 1.27-2.4]; RR, 1.1 [95% CI, 1.04-1.17]). Conversely, residence in a county with a high number of housing units lacking telephone service was found to be negatively associated with a postdiagnosis urologist visit (OR, 0.63 [95% CI, 0.47-0.85]; RR, 0.93 [95% CI, 0.89-0.99]). In the white subgroup, residence in a county with a high number of individuals living below the poverty level was not associated with a postdiagnosis urologist visit (OR, 0.95 [95% CI, 0.86-1.06]; RR, 0.99 [95% CI, 0.97-1.00]). Residence in a county with a high number of housing units lacking telephone service was negatively associated with a postdiagnosis urologist visit among white men (OR, 0.82 [95% CI, 0.74-0.9]; RR, 0.98 [95% CI, 0.96-0.99]).
Table 2. Modified Poisson Regression Model for Postdiagnosis Urologist Visit Within 1 Year: African American Sample (n=6115) and White Sample (n=56,995)
Generalized crime against persons factor (>90th percentile)
Access to services
Facilities factor (<10th percentile)
Services factor (<10th percentile)
Percent population below poverty level for 1999 (1 if >75th percentile)
Educational attainment among persons aged ≥25 years completing <9th grade for 2000 (1 if >75th percentile)
Occupied housing units with no telephone service available for 2000 (1 if >75th percentile)
Occupied housing units with no vehicles available for 2000 (1 if >75th percentile)
Poorly or not differentiated tumor
Well- or moderately differentiated tumor
No. of comorbidities (CCI)
Medicaid state buy-in
Age at diagnosis,y
Hospitalization, walking aid, SNF, oxygen use, or wheelchair use in previous 12 mo (morbidity)
Visit to primary care physician 12 mo before diagnosis
Large metro area
Random Intercept Logistic Regression Model
Variation across counties of patient residence accounted for 10% of the variation in the likelihood of a urologist visit. In subgroup analyses, variation at the county level accounted for 13% and 10%, respectively, of the variation in the likelihood of a urologist visit among AA and white men. The median OR (interval OR) associated with the county-level residual was 1.8 (95% CI, 0.3-2.9). The covariate-adjusted OR for the AA variable was 0.46 (95% CI, 0.42-0.5) in the multilevel model.
There is increasing interest in the impact of community-level factors on health outcomes for various cancers. In prostate cancer, studies have focused largely on mortality outcomes, in which the role of community-level socioeconomic status may account for some of the disparities in prostate cancer mortality between AA and white men.[8, 18, 19] Given that urologists are the gatekeepers in orchestrating the management of men diagnosed with earlier stages of prostate cancer, we focused on disparities in postdiagnosis urologist visits among men with stage I to stage III prostate cancer. Racial differences in specialist visits are of concern because they may lead to disparities in the receipt of treatment. For example, regardless of whether a patient diagnosed with localized prostate cancer opts for watchful waiting, active surveillance, or active treatment, contact with a urologist after the initial diagnosis of prostate cancer is important for ensuring that the patient makes informed decisions. We found that characteristics of the social environment and its relationship with urologist visits differed between AA and white men. Results regarding race differences in physician visits are consistent with prior findings concerning AA and white differences in physician visits among older adults.[20, 21]
A driving postulate in the current study was that community-level factors can influence a patient's health-related decisions and actions, including visits to a physician. We sought a more complete characterization of the patient's community than has been done in most prior studies. The county-level measure of the percentage of households without a telephone, for example, is one measure of an individual's environment. The hypothesis is not that an individual without a telephone is less likely to see a urologist after a diagnosis of prostate cancer. Rather, it is postulated that individuals living in a county ranked high in terms of the percentage of households without a telephone may have preexisting competing personal and social needs that reduce the likelihood that they will follow up with a urologist. The same reasoning applies to other area-level measures investigated, such as high crime or a high number of households without a vehicle. Thus, the focus is not whether a man diagnosed with prostate cancer has been a crime victim or lives in a house without a working telephone. Rather, it is the surrounding environment created by these factors that may influence a patient's personal priorities as they relate to seeking health care. Burdens borne at different times by those in an individual's broader support network may slowly realign the individual's priorities over time such that at the time he is diagnosed with a particular disease (eg, prostate cancer), a follow-up visit to a physician (eg, urologist) assumes a low priority. We found that more AA men than white men live in counties ranked high in terms of the measures of deprivation examined in the current study. As such, it is possible that a typical member of an AA man's support network is more likely to be a crime victim, without access to a vehicle, or without home telephone access, compared with the typical member of a white man's support network. Support networks are not strictly defined based on distance and may extend beyond the county of residence in ways that differ between AA and white men. The suggested differences in living experiences warrant further study.
Focusing on postdiagnosis urologist visits, the current study used a conceptual framework (Fig. 1) that emphasized the multilevel nature of factors that differ between AA and white men. Such multilevel focus is appropriate given that observed health behavior and outcomes are shaped by a complex interaction of individual and social environmental factors.[22-24] The multilevel view also highlights the interrelatedness between individuals and their community, as suggested by the finding that “the (mortality) consequences of neighborhood deprivation may be particularly exacerbated for Blacks, compared with whites.”
Those patients who did not see a primary care physician or who did not have reimbursed health services (despite insurance coverage) in the months leading up to their diagnosis of prostate cancer were less likely to follow up with the urologist after the cancer diagnosis. Furthermore, these effects were larger for AA men than for white men. These results highlight the importance of patient engagement in the health care system before the prostate cancer diagnosis. A patient who has not received health services or seen his primary care physician within the 12 months before the diagnosis of prostate cancer may be more difficult to engage in a management plan once a cancer diagnosis is made. Based on our conceptual framework, there could be potential differences in the factors characterizing the experience of AA men and white men during this 12-month baseline period. The finding with respect to urban versus rural residence is consistent with prior research that identified a lower likelihood of treatment receipt among rural residents compared with urban residents among older men diagnosed with early-stage prostate cancer. There were qualitative differences between white men and AA men in the role of poverty and the lack of services measured at the county level. These differences in effect between AA and white men are intriguing and warrant further investigation.
Although a relatively large population-based data set was used in the current study, one of its potential limitations is that it used Medicare claims data and therefore the results may not be generalizable to younger men diagnosed with prostate cancer. Furthermore, patient-level factors relating to culture and trust of the medical community may impact white and AA communities differently but were not available for analysis. Although the results of the current study indicated that there is a role for county-level factors, the area size is larger than the typical size for defining a community. Ideally, the measures of deprivation would be measured at a smaller area-level such as has been investigated in prior studies.[27, 28] The possibility of an ecologic fallacy exists when considering the measure of county-level poverty. Patient-level poverty may be poorly reflected by county-level poverty data and, furthermore, the difference may be related to patient race. Although the findings of the current study may also apply to certain other solid tumors such as those of the colon or lung, they may not be applicable to metastatic stages of cancer regardless of cancer type because prioritization among competing needs may differ once a patient is faced with advanced disease.
The current study examined patient-level and community-level factors that could in part account for race disparities in postdiagnosis urologist visits among older men with prostate cancer. The conceptual framework highlighted the race differences among men when considering their living environment as a factor potentially contributing to differences in follow-up visits with the urologist after a diagnosis of prostate cancer was made. Whether using a single-level model with contextual fixed effects or a multilevel model with separate estimates of patient-level and community-level variation, the consistent observation was that AA/white disparities in postdiagnosis urologist visits cannot be explained solely by patient-level factors. Rather, the patient's community is also important in understanding some of the disparities associated with individual health-related decisions and warrants further study, not only among patients with prostate cancer but perhaps those with other malignancies as well.
Supported by an American Cancer Society Institutional Research Grant to the University of Maryland Greenebaum Cancer Center. The collection of the California cancer incidence data used in the current study was supported by the California Department of Public Health as part of the statewide cancer reporting program mandated by California Health and Safety Code Section 103885; the National Cancer Institute's Surveillance, Epidemiology, and End Results Program under contract N01-PC-35136 awarded to the Northern California Cancer Center, contract N01-PC-35139 awarded to the University of Southern California, and contract N02-PC-15105 awarded to the Public Health Institute; and the Centers for Disease Control and Prevention's National Program of Cancer Registries under agreement #U55/CCR921930-02 awarded to the Public Health Institute.
CONFLICT OF INTEREST DISCLOSURES
Dr. Onukwugha received grants from Bayer Healthcare Pharmaceuticals and Amgen Inc as well as personal fees for a seminar on statistical methods from IMS Health, Janssen Analytics (a division of Johnson and Johson) for work performed outside of the current study. Dr. Mullins received grants from Amgen Inc and Bayer Healthcare Pharmaceuticals as well as personal fees from Bayer Healthcare Pharmaceuticals, Bristol-Myers Squibb, Genentech, Mundipharma, and Pfizer for work performed outside of the current study. Dr. Hussain has received research funding from Amgen Inc and Bayer Healthcare Pharmaceuticals for work performed outside of the current study.