SEARCH

SEARCH BY CITATION

Keywords:

  • prostate cancer;
  • health resource use;
  • race;
  • ethnicity;
  • cost

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. CONFLICT OF INTEREST
  9. REFERENCES

Study Type – Health economics (resource use) Level of Evidence 2b

OBJECTIVE

To analyse the racial and ethnic variation in health resource use (HRU) and direct medical care (DMC) cost in elderly men with prostate cancer.

PATIENTS AND METHODS

This was a retrospective case-control study using the linked Surveillance, Epidemiology, and End Results Medicare database. Patients with prostate cancer diagnosed between 1995 and 1998 (50 147 men) were identified and followed retrospectively for 1 year before and 5 years after the diagnosis. Phase-specific HRU and DMC costs were compared between racial and ethnic groups using parametric and nonparametric analysis. To compute the incremental cost of prostate cancer, a matched non-cancer control group was extracted from Medicare database. Poisson and general linear models (log-link) were used to identify the association of race and ethnicity with HRU and DMC cost, after controlling for potentially influential clinical and demographic covariates.

RESULTS

The African-American group was more likely to have emergency-room visits (odds ratio 1.19, 95% confidence interval 1.12–1.28) and less likely to have outpatient visits (0.96, 0.96–0.97) than whites. However, the Hispanic group was more likely to have inpatient and outpatient visits (odds ratio 0.88, 0.83–0.91; and 0.93, 0.91–0.95) than whites. Adjusted DMC cost showed racial and ethnic variation in all phases except the treatment and terminal phases. Factors associated with DMC cost varied among racial and ethnic groups.

CONCLUSION

The incremental burden of prostate cancer remains significant in the long term. Overall, the cost of prostate cancer care was higher among African-American men than white and Hispanic men. This indicates the need for further research on care-level factors to comprehend the racial and ethnic disparity in HRU and cost.


Abbreviations
AA

African-American

HRU

health resource use

DMC

direct medical care

SEER

Surveillance, Epidemiology, and End Results

PEDSF

Patient Entitlement and Diagnosis Summary File

ER

emergency room

RP

radical prostatectomy

EBRT

external beam radiotherapy

GLM

generalized linear model.

INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. CONFLICT OF INTEREST
  9. REFERENCES

Prostate cancer is the sixth most common cancer in the world, with 782 600 new cases diagnosed in 2007 alone, and the third most common cancer in men and the most common cancer in men in Europe and North America [1,2]. Disparities in the incidence and outcomes of prostate cancer are a characteristic of the global pattern of prostate cancer, with men of African descent suffering disproportionately from this disease. There is a significant disconnection between the available treatment options and delivery of care to the population in general and the population from certain racial and ethnic groups, socio-economic strata and geographical location. The cause of this disparity and its effect on the pattern of health care use and cost of care has been inadequately explored.

Prostate cancer is the most prevalent cancer among elderly men in the USA, with an estimated 192 280 new cases diagnosed during 2009 [3]. Prostate cancer is also associated with higher annual Medicare expenditure than any other cancer among elderly men, and is expected to increase with the ageing of the USA population. For Medicare patients diagnosed with prostate cancer in 2004, the 5-year aggregated cost of care was estimated to be US$2.3 billion [4]. The introduction of the PSA test has resulted in the earlier detection of prostate cancer, a doubling of its incidence rate and a parallel increase in the number of men treated for the disease [3–5]. This trend has many implications for health resource use (HRU) and direct medical care (DMC) cost patterns. There are disparities in care for prostate cancer across institutions, regions, age, racial and ethnic groups [4–12]. Patterns of HRU and DMC cost are an indicator of the total resource devoted to the care of patients. Thus, determinants of racial and ethnic variations in HRU and associated costs are crucial for developing effective healthcare policy to improve the quality of care for older adults [12].

Prostate cancer poses many challenges to healthcare, due to its high incidence, lack of consensus on screening benefits, uncertainty about the best course of treatment, and costs (direct and indirect) associated with detection and treatment [13–15]. While studies indicate that race is an important predictor of treatment for prostate cancer [13,14], the effects of prostate cancer care on short- and long-term DMC costs, HRU and factors associated with such ethnic disparity remain to be investigated [12,15]. There is racial and ethnic disparity in prostate cancer care across all phases of care, beginning with prevention, to screening, diagnosis, treatment and follow-up care. Such disparities in prostate cancer care might translate into higher costs to both the healthcare system and society. Analysing and understanding the disparities in costs of prostate cancer care is essential in developing strategies to eliminate or reduce such disparities. However, systematic and comprehensive assessment of such costs remains unexplored [13–15]. Wilson et al.[16] showed that prostate-related costs per person are substantial and sustained over time. The short-term treatment cost comparisons do not truly reflect the cost of treatment choices over the long-term. Also, a population-based study showed that cancer survivors visited their specialist more often than the comparable normative population [17]. Thus, analysing the long-term cost associated with prostate cancer care is crucial for understanding the effects of disparity in treatment. The incremental cost of illness improves the understanding of cost attributable to prostate cancer. Most men with prostate cancer are elderly and often have multiple diagnoses that need a complex combination of care. Allocating costs correctly across diagnoses for a person is a problem of joint product, making it difficult to attribute specific services and costs to each diagnosis. This is further compounded by concern about efficacy of treatments for prostate cancer. Thus, the incremental-cost method using matched controls is essential to estimate the costs incurred due to prostate cancer. The objective of this study was to analyse the racial and ethnic variation in short- and long-term HRU pattern and DMC cost of elderly patients after controlling for demographic and clinical characteristics. The linked Surveillance, Epidemiology, and End Results (SEER) Medicare database was used to examine the phase-specific (before treatment, treatment, after treatment and terminal phases) implications of racial and ethnic disparity in HRU and DMC cost for elderly African-American (AA), Hispanic and white men diagnosed with prostate cancer. We also analysed the factors associated with HRU and DMC cost for each racial and ethnic group separately. We hypothesise that there is significant racial and ethnic variation in the pattern of HRU and cost of care among elderly patients with prostate cancer.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. CONFLICT OF INTEREST
  9. REFERENCES

We used the SEER-Medicare linked data of the National Cancer Institute; these data bring together Medicare administrative claims data and clinical tumour registry data for Medicare recipients, and offer an excellent opportunity for meaningful outcomes research in prostate cancer care [18]. The SEER programme collects data on cancer incidence, treatment and mortality rate from 13 SEER sites and encompasses 14% of the USA population. With the exception of individuals who are enrolled in health maintenance organizations or do not have Part B coverage (13%), Medicare data provides information about all inpatient and outpatient use of medical care for residents of the USA who are aged ≥65 years. The SEER-Medicare file contains one record for each Medicare beneficiary enrolled in the SEER programme. Of persons diagnosed with cancer at age ≥65 years enrolled in SEER registries, 93% have been matched with their Medicare enrolment records in a linked customized file, the Patient Entitlement and Diagnosis Summary File (PEDSF). The SEER data provide characteristics of the tumour that are crucial to adequately adjust for prostate cancer severity, including histology, stage and grade [18].

For this retrospective case-control design, we created a cohort of all AA (non-Hispanic black), Hispanic and white (non-Hispanic white) men, aged ≥66 years, and diagnosed with prostate cancer between 1995 and 1998 (57 128 men) from SEER-Medicare data. Men of other ethnicity or those aged <66 years at the time of diagnosis were excluded to ensure that the data file included sufficient claims for medical care before a diagnosis to allow for comorbidity adjustments. The prostate cancer cohort was followed retrospectively for 1 year before diagnosis (pre-diagnosis phase) and up to 5 years after the diagnosis. The first year after diagnosis was considered the ‘treatment phase’, and the other 4 years were considered as the ‘follow-up phase’. For those who died during the treatment or follow-up phase, we considered the 1-year period before the date of death as the ‘terminal phase’[4,19]. A cancer-free control group (1:1) matched by age, ethnicity, comorbidity and postal codes (a proxy for income) was extracted from the 5% Medicare non-cancer files. As the cancer-free controls had no ‘date of diagnosis’ we assigned a random ‘pseudo-diagnosis’ (anchor) date and determined phases similar to the prostate cancer cohort [4], i.e. 1-year before the anchor date (pre-treatment), 5 years after (1-year treatment phase and 4-year follow-up-phase) and a terminal phase.

MEASUREMENT STRATEGY

Outcome variables

HRU; the linked record included service codes from three Medicare files: (i) inpatient file; (ii) hospital outpatient standard analytical file (claims for outpatient facility services); and (iii) physician part B file (claims for physician and other medical provider services). Health resources are categorized as inpatient (length of stay, number of admission, surgical and diagnostic procedures), outpatient (laboratory and emergency room, ER), skilled nursing facility, durable medical equipment, home health services and hospice care.

DMC costs were defined as the reimbursements received from Medicare by respective healthcare organization [19,20]. The total DMC costs include costs of care provided by physicians and other health professionals, care provided in hospitals, outpatient and ER costs, inpatient medications, and laboratory services. Sources of costs data were: (i) inpatient file (Medicare Provider Analysis and Review); (ii) hospital outpatient standard analytical file (claims for outpatient facility services); (iii) physician part B file (claims for physician and other medical provider services); (iv) home health agency; (v) durable medical equipment; and (vi) hospice care.

Of all the payments recorded for a patient with prostate cancer in the Medicare database, only a portion is attributable to treatment for the disease. To identify the costs related to prostate cancer we used a case-control approach to compute the incremental cost of prostate cancer [19,20]. This involves comparing phase-specific costs of cases to those from a matched sample of cancer-free Medicare patients. The incremental cost analysis will deal with the joint-product issue that often affects cost- of-illness studies [4,19,20]. Also, given the complex nature of SEER-Medicare database, the case-control approach provides a viable approach to compute phase-specific care patterns and costs [4,19,20].

EXPLANATORY VARIABLES

For each eligible patient in the SEER-Medicare linked file, primary and secondary procedure codes were searched to identify International Center for Disease-9-CM codes for prostate cancer treatments. Primary procedures were; surgery (radical prostatectomy, RP), radiotherapy (external beam, EBRT, or brachytherapy), hormone therapy and watchful waiting.

Disease severity was adjusted for using information on prostate cancer stage, grade and histology, provided in the SEER database. The Charlson comorbidity index was used to assess medical comorbidity, using inpatient and outpatient Medicare claims data [21,22]. We used diagnostic information from all encounters 1 year before the month of prostate cancer diagnosis to adjust for comorbidity [21,22].

We obtained demographic and income data from the PEDSF file, which is in a one-record-per-person format and includes individuals with a cancer diagnosis as captured by one of the SEER registries, and whose records have been linked to the Medicare Enrolment Database.

We first tested for underlying differences in the demographic and clinical characteristics of AA, Hispanic and white men from the study cohort using t-tests and chi-square tests for continuous and categorical data, respectively. For categorical data on HRU, we used chi-square tests to compare across racial and ethnic groups for each phase. The annual HRU and costs for the pre-treatment, treatment, follow-up and terminal phases were compared across racial and ethnic groups using anova[23]. Poisson regression models (with zero inflation correction) were used to assess HRU patterns [24]. The dependent variables were data on the number of inpatient, outpatient and ER visits. The independent variables were age, ethnicity, income, TNM stage, geographical area, SEER registry area, Charlson comorbidity and marital status. Mean costs were compared between racial and ethnic groups using parametric and nonparametric tests. As cost data often violate normality assumptions, nonparametric tests such as Wilcoxon rank-sum test (median costs) and t-test on log-transformed data were used [23,24]. To analyse the association of race and ethnicity with DMC cost, we used a generalized linear model (GLM) with a log-link and γ-distribution variance function [24,25]. To estimate the incremental cost between prostate cancer and non-cancer groups, we fitted a GLM. The incremental cost of prostate cancer is the difference between the total costs of care for the prostate cancer group and non-cancer controls, after adjusting for potentially influential covariates. All costs were standardized to 2009 values using a 5% discount rate. Finally, the factors associated with total DMC cost were analysed separately for each racial and ethnic group using the GLM.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. CONFLICT OF INTEREST
  9. REFERENCES

The study sample comprised 50 147 fee-for-service elderly Medicare beneficiaries who were diagnosed with prostate cancer between 1995 and 1998 (40 876 white, 5867 AA and 3404 Hispanic patients). As shown in Table 1, there were significant differences in the demographic and clinical characteristics across racial and ethnic groups. The mean age at the diagnosis of prostate cancer was lower for AA and Hispanic men than for white men. Of AA patients, >90% lived in large metropolitan areas, were less likely to be married and had a lower annual income. For clinical characteristics, more AA men presented with greater medical comorbidity at the time of diagnosis and were diagnosed at a later stage of cancer. Also, treatment patterns differed across racial and ethnic groups; AA men were less likely to have RP, whereas Hispanic men were more likely to have RP than were white men.

Table 1.  Baseline demographic and clinical characteristics of the 50147 men
Mean (sd) or % variableWhiteAAHispanicP
No. of patients40 876 5 867 3 404 
Age at diagnosis, years    74.5 (6.2)    73.7 (6.1)    73.7 (6.2)<0.001
Geographical area    
 Large metropolitan    61.31    91.63    73.14 
 Metropolitan    24.96     6.41    18.37<0.001
 Urban     5.32     0.46     2.09 
 Less Urban     6.84     1.33     6.05 
 Rural     1.58     0.17     0.35 
Marital status    
 Single     6.78    15.05     7.93 
 Married    70.75    55.51    69.89<0.001
 Separated     0.23     1.41     0.50 
 Divorced     3.95     8.28     5.49 
 Widowed    10.90    12.43    10.08 
 Unknown     7.39     7.31     6.11 
Charlson comorbidity score    
 0    83.24    73.29    96.06<0.001
 1–2    12.11    19.46     2.94 
 3–5     4.57     7.06     1.00 
 >5     0.08     0.19     0.00 
Annual median income of   census tract, $41 507 (150 008)25 617 (110 380)32 959 (15 000)<0.001
Histology    
 In situ     0.04     0.10     0.00<0.001
 Distant     6.50     9.70     8.49 
 Localized/Regional    84.43    77.77    84.05 
 Unstaged     9.03    12.43     7.46 
Grade    
 Well differentiated    10.07     7.41     11.46 
 Moderately differentiated    58.85    57.29    55.52<0.001
 Poorly differentiated    20.06    21.78    22.68 
 Undifferentiated     0.52     0.48     0.59 
Stage    
 ≤T2a    41.06    36.71    41.36<0.001
 T2b and T2c    36.25    33.00    33.73 
 ≥T3a    22.69    30.29    24.91 
Treatment    
 RP    28.41    24.05    34.62<0.001
 RP + RT     3.33     4.40     3.33 
 RP + RT + hormone     1.30     1.37     1.47 
 RP + hormone     4.03     3.51     4.89 
 EBRT    24.91    28.04    21.07 
 Brachytherapy alone     3.32     2.10     1.13 
 EBRT + hormone    15.57    14.01    13.50 
 Brachytherapy + hormone     2.05     0.99     0.95 
 Hormone    17.09    21.54    19.04 

Comparisons of unadjusted HRU (ER, inpatient and outpatient visits and length of stay) patterns indicated significant variation in usage between racial and ethnic groups and phases (results not presented). AA men had more ER visits than whites during all phases of care. By contrast, outpatient visits were higher among white men across all phases. Inpatient visits were more numerous among whites during the treatment phase. The mean length of stay was longer among AA men during all phases of care.

For the costs of inpatient HRU, AA and Hispanic men reported higher inpatient pharmacy costs during all phases of care than white men (results not presented). However, medical and surgical supply costs were highest for whites during all phases, except the terminal phase. Finally, inpatient physical therapy and laboratory costs were higher among AA men during all phases of care. The results of Poisson regression models indicated that AA men had more ER visits (odds ratio, 1.19, 95% CI 1.12–1.28) and fewer outpatient visits (0.96, 0.96–0.97) than whites. However, Hispanic men had fewer inpatient and outpatients visits (odds ratio 0.88 95% CI 0.83–0.91; and 0.93, 0.91–0.95) than whites (Table 2).

Table 2.  Poisson regression models (treatment phase)
VariableOdds ratio (sem), P, for number of
ER visitsInpatient visitsOutpatient visits
Intercept0.003 (0.2021), <0.0010.12 (0.1126), <0.0014.2 (0.0449), <0.001
Age1.06 (0.0015), <0.0011.04 (0.0009), <0.0011.00 (0.0004), <0.001
Income0.99 (0.0010), <0.0010.99 (0.001), <0.0010.99 (0.0001), <0.001
Marital status: married0.87 (0.0224), <0.0010.95 (0.0121), <0.0011.03 (0.0046), <0.001
TNM stage   
 T2b and T2c0.79 (0.0452), <0.0011.09 (0.0172), <0.0011.08 (0.0072), <0.001
 ≥T3a1.88 (0.0283), <0.0011.66 (0.0162), <0.0011.13 (0.0074), <0.001
 ≤T2a (reference)   
Geographical area   
 Metropolitan1.20 (0.0907), 0.0420.96 (0.0447), 0.2700.56 (0.0144), <0.001
 Urban1.27 (0.0894), 0.0071.06 (0.0442), 0.1680.77 (0.0141), <0.001
 Rural (reference)   
Comorbidity: Charlson score   
 1–21.03 (0.0388), 0.3891.00 (0.0230), 0.8760.99 (0.0074), 0.999
 ≥31.09 (0.0501), 0.0971.05 (0.0300), 0.1311.03 (0.0097), 0.010
 0 (reference)   
Treatment   
 RP1.21 (0.0257), <0.0012.23 (0.0123), <0.0011.03 (0.0055), <0.001
 RT0.68 (0.0275), <0.0010.65 (0.0148), <0.0011.63 (0.0046), <0.001
 Hormone1.08 (0.0244), 0.0021.00 (0.0134), 0.8151.19 (0.0046), <0.001
 None (reference)   
Race   
 AA1.19 (0.0337), <0.0011.03 (0.0190), 0.1730.96 (0.0073), <0.001
 Hispanic0.99 (0.0499), 0.9050.88 (0.0237), <0.0010.93 (0.0107), <0.001
 white (reference)   
SEER Registry Area   
 San Francisco0.53 (0.1422), <0.0010.82 (0.0838), 0.0160.75 (0.0341), <0.001
 Connecticut0.98 (0.1405), 0.8790.97 (0.0841), 0.6741.21 (0.0338), <0.001
 Detroit1.14 (0.1389), 0.3391.04 (0.0835), 0.6751.91 (0.0335), <0.001
 Hawaii0.67 (0.2765), 0.1400.73 (0.1494), 0.0320.76 (0.0600), <0.001
 Iowa0.94 (0.1350), 0.6131.01 (0.0811), 0.8781.44 (0.0329), <0.001
 New Mexico0.71 (0.1417), 0.0160.86 (0.0838), 0.0741.08 (0.0340), 0.028
 Seattle0.52 (0.1405), <0.0010.72 (0.0831), <0.0011.07 (0.0335), 0.0336
 Utah0.58 (0.1446), <0.0010.87 (0.0842), 0.1081.29 (0.0340), <0.001
 Atlanta1.01 (0.1418), 0.9170.97 (0.0848), 0.7581.07 (0.0343), 0.035
 San Jose0.62 (0.1494), 0.0010.80 (0.0866), 0.0120.76 (0.0354), <0.001
 Los Angeles0.59 (0.1351), <0.0010.87 (0.0825), 0.0810.72 (0.0336), <0.001
 Rural Georgia (reference)   

Comparisons of unadjusted phase-specific DMC costs and incremental costs showed that, overall, the costs were higher during the treatment phase than in the pre-diagnosis phase, and declined over the follow-up phase (Table 3). This trend was apparent for all racial and ethnic groups, but the costs were highest during the terminal phase for all racial and ethnic groups. AA men had higher DMC costs during all phases of care. The patterns were similar for incremental cost after adjusting for covariates. Also, the multivariate regression model for incremental cost analysis showed that overall, the prostate cancer group had higher costs (odds ratio 1.23, 95% CI 1.12–1.36) than the control group. Unadjusted comparison of the costs associated with inpatient, outpatient, ER visits, home health agency services, hospice care and physician cost during the follow-up phases of care indicated that overall, AA patients had higher DMC costs during all phases of care (Table 3). As seen in Table 4, the GLM indicated that AA men had higher total DMC costs for all post-treatment phases, after controlling for covariates. The β values reported in Tables 4 and 5 are parameter estimates that can be converted as 100 (eβ− 1) to obtain the percentage change in the expected value of the dependent variable for a one-unit increase in the independent variable [26]. For example, the β value of 0.12 associated with the AA group for post-treatment (year 2) can be interpreted as a 12.7% increase in total DMC costs, compared to the white group. The factors associated with total DMC costs differed for each racial and ethnic group (Table 5). Age, marital status, TNM stage and treatment were associated with higher total DMC costs for white patients. For AA men, being married was associated with lower costs, where TNM stage >T3, radiotherapy and comorbidity score of >2 were associated with higher costs. Finally, TNM stage and radiation and hormone treatment were associated with higher total costs for Hispanic patients.

Table 3.  Phase-specific total DMC costs, as the mean (sd) (in 1000s US$) and median (range) and incremental cost (IC), as the mean (sem)
PhaseWhiteAAHispanic
CasesControlsIC*CasesControlsIC*CasesControlsIC*
  • *

    Values are in US$;

  • P≤ 0.001.

Pre-diagnosis4.90 (20.3)4.39 (12.01)   517 (82)7.09 (26.7)4.17 (12.8) 2 922 (401)2.85 (12.0)2.83 (8.59)  18 (273)
Median*153754 126188 03 
range (1000s)0–1194.30–379.95 0–808.80–327.5 0–303.50–98.6 
Treatment         
Year 121.4 (79.9)6.34 (15.55)15 070 (402)26.5 (92.40)7.04 (20.49)19 473 (1287)10.13 (28.97)4.88 (16.06)5249 (621)
Median*50411078 4527353 4484 
range (1000s)0–223.010–408.99 0–222.100–38.27 0–681.280–254.73 
After treatment         
Year 27.81 (29.04)4.91 (13.26) 2 901 (157)10.27 (39.81)5.55 (16.51) 4 710 (583)5.52 (20.90)3.65 (13.04)1888 (459)
Median*298635 11184 00 
range (1000s)0–1051.670–345.99 0–1490.840–255.08 0–440.180–206.17 
Year 35.93 (15.68)4.96 (14.21)   961 (104)6.84 (21.55)5.46 (16.79) 1 378 (366)5.58 (26.48)3.54 (12.69)2339 (553)
Median*343498 018 00 
range (1000s)0–479.990–607.89 0–459.780–281.56 0–1049.610–235.15 
Year 46.88 (18.54)5.04 (16.49) 1 837 (122)7.78 (24.40)5.69 (17.39) 2 091 (392)6.50 (24.21)3.83 (12.53)2668 (512)
Median*545439 00 00 
range (1000s)0–1110.590–1208.61 0–317.020–376.65 0–658.900–230.88 
Year 57.53 (20.63)5.27 (15.92) 2 258 (128)7.75 (24.31)5.50 (17.28) 2 251 (406)6.18 (21.88)4.57 (15.23)1603 (495)
Median*574372 00 00 
range (1000s)0–896.300–943.31 0–580.570–272.13 0–468.600–297.25 
Terminal (1 year)23.88 (62.20)10.12 (19.50)10 190 (437)30.07 (68.75)14.20 (26.26) 11 602 (1312)18.75 (42.87)10.19 (22.85)5566 (1616)
Median*72911743 94482165 39490 
range (1000s)0–2097.940–277.29 0–1887.690–272.01 0–625.070–121.18 
Table 4.  Association of race and ethnicity and phase-specific DMC cost
Variableβ (sem), P*, for phase after diagnosis
1 year2 years3 years4 years5 yearsTerminal (1 year)
  • *

    Entries with no P value are P < 0.001, and others are shown to two decimal places.

Intercept8.8 (0.14)6.6 (0.20)7.2 (0.10)7.2 (0.19)8.1 (0.19)10.7 (0.22)
Age0.004 (0.001)0.03 (0.01)0.03 (0.01)0.03 (0.002)0.03 (0.002)−0.01 (0.01)
Income−0.001 (0.001)−0.001 (0.001)−0.001 (0.001)−0.001 (0.001)−0.001 (0.001)−0.001 (0.001), 0.01
Marital status (married)−0.09 (0.02)−0.13 (0.02)−0.09 (0.02)−0.07 (0.02)−0.03 (0.02), 0.100.03 (0.02), 0.26
Stage: ≤T2a (ref)      
T2b and T2c0.11 (0.02)−0.03 (0.03), 0.300.02 (0.03), 0.58−0.04 (0.03)0.01 (0.03), 0.840.15 (0.05)
≥T3a0.27 (0.02)0.37 (0.04)0.35 (0.04)0.37 (0.04)0.32 (0.04)0.14 (0.03)
Geographical area      
Metro0.07 (0.06), 0.200.07 (0.07), 0.340.08 (0.07), 0.26−0.10 (0.07), 0.260.03 (0.07), 0.630.17 (0.09), 0.05
Urban0.06 (0.05), 0.29−0.04 (0.07), 0.620.02 (0.07), 0.74−0.10 (0.07), 0.480.001 (0.07), 0.990.05 (0.09), 0.58
Rural (ref)      
Treatment      
Surgery (RP)0.17 (0.02)−0.15 (0.02)−0.20 (0.02)−0.17 (0.02)−0.13 (0.02)0.07(0.03), 0.007
Radiation0.37 (0.02)−0.13 (0.02)−0.17 (0.02)−0.08 (0.02), 0.70−0.05 (0.02), 0.0050.07 (0.03), 0.009
Hormone0.09 (0.02)0.24 (0.02)0.23 (0.02)0.21 (0.02), 0.0010.18 (0.02)0.12 (0.03)
None (ref)      
Charlson score      
1–20.04 (0.03), 0.09−0.02 (0.03), 0.65−0.03 (0.03), 0.420.04 (0.03), 0.78−0.05 (0.03), 0.16−0.02 (0.04), 0.72
≥30.04 (0.03), 0.250.05 (0.05), 0.28−0.02 (0.04), 0.720.04 (0.04), 0.80−0.03 (0.04), 0.490.08 (0.06), 0.16
0 (ref)      
Race      
AA0.01 (0.03), 0.640.17 (0.03)0.24 (0.03)0.19 (0.03)0.10 (0.03), 0.0010.05 (0.04), 0.19
Hispanic−0.05 (0.03), 0.15−0.07 (0.04), 0.130.07 (0.04), 0.080.03 (0.04), 0.41−0.09 (0.04), 0.03−0.05 (0.05), 0.39
white (ref)      
SEER Registry Area      
San Francisco0.38 (0.11)0.55 (0.15)0.11 (0.14), 0.440.21 (0.13), 0.04−0.43 (0.13), 0.0010.17 (0.15), 0.25
Connecticut1.8 (0.10)1.5 (0.14)0.07 (0.14), 0.610.13 (0.13), 0.19−0.27 (0.13), 0.040.94 (0.15)
Detroit1.7 (0.10)1.1 (0.14), 0.350.05 (0.14), 0.730.13 (0.13), 0.41−0.29 (0.13), 0.030.71 (0.15)
Hawaii0.30 (0.18), 0.10−0.28 (0.30), 0.68−0.42 (0.28), 0.14−1.1 (0.25), 0.23−1.3 (0.23)−0.23 (0.26), 0.38
Iowa−0.09 (0.10), 0.37−0.06 (0.14), 0.60−0.22 (0.13), 0.09−0.23 (0.13), 0.23−0.68 (0.17)−0.23 (0.14), 0.11
New Mexico−0.11 (0.11), 0.370.08 (0.14), 0.60−0.27 (0.15), 0.05−0.37 (0.13), 0.1−0.78 (0.13)−0.27 (0.15), 0.07
Seattle−0.05 (0.10), 0.620.08 (0.14), 0.56−0.24 (0.14), 0.08−0.24 (0.13), 0.67−0.69 (0.13)0.11 (0.15), 0.47
Utah−0.21 (0.11), 0.04−0.02 (0.14), 0.9−0.26 (0.14), 0.05−0.19 (0.13), 0.19−0.60 (0.13)−0.20 (0.15), 0.18
Atlanta−0.18 (0.11), 0.090.28 (0.15), 0.06−0.07 (0.14), 0.570.04 (0.13), 0.96−0.34 (0.13), 0.01−0.001 (0.16), 1.0
San Jose0.37 (0.11)0.63 (0.15)0.12 (0.14), 0.410.21 (0.14), 0.03−0.27 (0.14), 0.050.22 (0.16), 0.17
Los Angeles0.58 (0.10)0.89 (0.14)0.49 (0.13)0.52 (0.003), 0.0070.08 (0.15), 0.560.57 (0.15)
Rural Georgia (ref)      
Table 5.  Factors associated with DMC costs
Variableβ (sem), P*
WhiteAAHispanic
  • *

    Entries with no P value are P < 0.001, and others are shown to two decimal places.

Intercept8.5 (0.17)8.8 (0.53)9.4 (0.53)
Age0.01 (0.001)−0.0003 (0.004), 0.940.007 (0.005), 0.13
Income−0.0001 (0.001)−0.001 (0.001)−0.001 (0.001), 0.28
Marital status (married)−0.07 (0.02)−0.17 (0.04)−0.01 (0.06), 0.83
Stage   
T2b and T2c0.12 (0.03)−0.06 (0.08), 0.490.21 (0.09), 0.015
≥T3a0.26 (0.03)0.19 (0.07), 0.010.38 (0.09)
≤T2a (ref)   
Geographical area   
Metropolitan0.06 (0.06), 0.280.95 (0.54), 0.08−0.13 (0.38), 0.74
Urban0.06 (0.06), 0.300.81 (0.47), 0.09−0.28 (0.38), 0.47
Rural (ref)   
Treatment   
Surgery (RP)0.21 (0.02)0.09 (0.05), 0.10−0.12 (0.07), 0.06
Radiation0.37 (0.02)0.39 (0.05)0.31 (0.07)
Hormone0.11 (0.02)−0.03 (0.05), 0.560.11 (0.07), 0.02
No treatment (ref)   
Comorbidity: Charlson score   
1–20.07 (0.03), 0.01−0.11 (0.06), 0.080.10 (0.23), 0.66
≥3−0.002 (0.04), 0.950.27 (0.08), 0.02−0.47 (0.29), 0.10
0 (ref)   
SEER Registry Area   
San Francisco0.48 (0.14)0.19 (0.31), 0.53−0.44 (0.10)
Connecticut1.9 (0.14)1.6 (0.31)1.1 (0.19)
Detroit1.8 (0.14)1.2 (0.31)0.97 (0.27)
Hawaii0.43 (0.21), 0.04−1.4 1(0.98), 0.17−1.1 (1.2), 0.38
Iowa0.004 (0.14), 0.98−0.15 (0.34), 0.66−1.1 (0.40), 0.01
New Mexico−0.01 (0.14), 0.93−0.51 (0.34), 0.14−0.69 (0.09)
Seattle0.04 (0.14), 0.75−0.17 (0.31), 0.59−0.82 (0.29), 0.005
Utah−0.11 (0.14), 0.43−1.1 (0.48), 0.03−0.79 (0.19)
Atlanta−0.12 (0.14), 0.41−0.35 (0.30), 0.25−0.76 (0.31), 0.02
San Jose0.47 (0.14), 0.0010.38 (0.37), 0.31−0.38 (0.31)
Los Angeles0.62 (0.14)0.48 (0.31), 0.110.0001 (0.001), 0.99
Rural Georgia (ref)   

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. CONFLICT OF INTEREST
  9. REFERENCES

The economic impact of prostate cancer is often substantial for patients and families, employers, payers, and society at large. Measuring the burden of the disease and analysing the factors associated with observed racial and ethnic variation in treatment, HRU and cost is a crucial step for health researchers and policy makers. Important findings of this study are: (i) There is significant variation in the HRU pattern and costs of prostate cancer care between racial and ethnic groups; (ii) incremental cost analysis indicated that the diagnosis of prostate cancer has a significant burden during all phases of care; (iii) AA patients have higher care costs for prostate cancer; (iv) HRU and costs were comparable between Hispanic and white patients; (v) factors associated with total cost varied between racial and ethnic groups; (vi) the incremental cost of prostate cancer care was higher for the AA group; and (vii) during treatment and the terminal phase of care, race and ethnicity were not associated with the total cost of care.

Racial and ethnic disparities in HRU and cost are recognized as a major quality problem [12]. In the past, SEER-Medicare linked data have been used in outcomes research on various cancer presentations, including prostate cancer [10,11,13–15,19,27–30]. A study using a 5% random sample of Medicare beneficiaries found that AA men undergoing RP or EBRT had higher charges than white men for early-stage cancer [31]. Although the direction of these previous results is in accordance with our results, the magnitude differed, mainly due to the use of charges instead of reimbursement. In the present study we used reimbursement, as Medicare costs are derived mainly from the reimbursement [19,20]. Other investigators using an administrative database from one healthcare system showed that comorbidity was a predictor of the type of treatment cost. It was also noted that provider characteristics are important in the variation in cost and treatment [32,33]. Many factors are associated with the observed variation in HRU and cost. AA and white patients are to a large extent treated by different physicians and type of hospitals. The characteristics and treatment of patients with prostate cancer differ significantly by age and ethnicity. Some studies indicate that treatments for a given stage of prostate cancer vary by age and ethnicity [4,13,14], and others have addressed the variations in cost of prostate cancer and disease-specific mortality [11,31,34]. Using the SEER-Medicare database, Chu et al.[35] reported similar patterns of incidence, survival and mortality in black and white men in the USA. In another study using SEER-Medicare data [29] racial disparities were evident both in overall survival and prostate cancer-specific survival. Black patients had poorer overall survival among patients undergoing surgery. In summary, studies have analysed racial and ethnic disparity in prostate cancer treatment and mortality, but only a few have explored disparities in phase-specific HRU and DMC cost [4]. Our study aimed to minimize some of these observed gaps in research by analysing the HRU and DMC cost of care related to prostate cancer using national-level data.

Despite the strengths of using a population-based national sample from SEER-Medicare, our study has limitations. The data created in this study included only AA, Hispanic and white men aged >66 years who live in a SEER area and are not enrolled in a health maintenance organization. Although SEER is designed to provide a representative sample of the USA, only a relatively small Hispanic population is included. Our findings might not be generally applicable to men aged <66 years or to men enrolled in a health maintenance organization. The age and sex distribution for individuals aged ≥66 years in the SEER areas is comparable with that of the elderly population of the USA. However, the SEER area distribution differs from that in the USA elderly population, in that SEER areas have a lower proportion of whites and a higher proportion of persons of other races. Routinely collected statistics on cancer occurrence commonly reported for the major racial and ethnic population might mask wide variations in the cancer burden by county of origin. The use of fee-for-service Medicare claims data provides an excellent opportunity to analyse prostate cancer care in a broad population. However, claims data have limitations, such as misclassification of procedure codes. Finally, although we controlled for treatment and other covariates in the analysis, it is possible that some biases remain unaccounted for and we plan to address these in our future studies using an instrumental variable approach.

A major concern of policy makers in the USA is the escalating cost of healthcare and disparities in HRU [12,15]. In this study we identified racial and ethnic variations in phase-specific HRU and cost of care among elderly men with prostate cancer. AA patients had higher HRU and DMC costs than white and Hispanic groups. However, during the treatment and terminal phase, DMC cost did not differ between the racial and ethnic groups after controlling for demographic and clinical covariates. Factors associated with cost of care and HRU also varied between racial and ethnic groups. This shows a substantial opportunity to minimize the racial and ethnic variations in cost of care and HRU. Any health-policy measures aimed at effectively reducing racial and ethnic disparity in care should address the issues related to variation in HRU and cost. Our results also suggest the need for further research on various care-level factors using a systems approach in understanding the under-use, overuse and misuse of health resources across racial and ethnic groups.

ACKNOWLEDGEMENTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. CONFLICT OF INTEREST
  9. REFERENCES

Supported in part by the National Cancer Institute, National Institutes of Health-Grant ♯ 5RO3CA 121338-2 and Linda and Laddie Montague research fund.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ACKNOWLEDGEMENTS
  8. CONFLICT OF INTEREST
  9. REFERENCES