Urban/Rural Differences in Survival Among Medicare Beneficiaries with Breast Cancer Lisa R. Shugarman, Ph.D. Haijun Tian, Ph.D. Arvind Jain, M.S. J. Scott Ashwood, M.A. Melony E.S. Sorbero, Ph.D. RAND Corporation Funded by Health Resources and Services Administration Office of Rural Health Policy Background – Breast Cancer Background – Rural Health • Rural areas are characterized by: • High incidence of breast cancer – Most common cancer type in women – Probability of diagnosis increases with age • Second leading cause of cancer death in women • American Cancer Society estimates for 2007 – 178,480 women will be diagnosed with invasive breast cancer – 40,460 will die • • • • – Lower population density – Large distances between individuals and communities – Large distances from urban centers Experience challenges recruiting and retaining providers Hospitals and other facilities not capable of providing all services Populations in rural areas travel further and wait longer for outpatient care Rural elders more likely to be poor and near poor than urban elders Sorbero 3 6-3-07 Objectives Sorbero 4 6-3-07 Methods - Data • Three data sources • To examine urban/rural differences in survival among women age 65 and older who have been diagnosed with breast cancer • Survival differences may exist due to – Urban/rural socioeconomic differences – Lower local supply of cancer services and providers in rural areas – Surveillance, Epidemiology, and End Result (SEER) Data (1995-1999) • 14 cancer registries representing 26% US population – Linked Medicare administrative data (claims and enrollment database) (1994-2003) – Area Resource File (selected years for supply variables) Sorbero 5 6-3-07 Sorbero 6 6-3-07 1 Methods - Sample Methods – Defining Urban/Rural • Inclusion criteria – Breast cancer was the first diagnosed cancer – Female – Continuously enrolled in both Medicare Part A & B for 1-year before diagnosis through 8 months after diagnosis • Exclusion criteria – Enrolled in managed care (N=12,843) – Eligible for Medicare for ESRD diagnosis or disability (N=16,326) – Breast cancer diagnosed via autopsy or death certificate (N=47) • N=32,626 • County-based definitions create a single label for counties with hetergeneous population densities • 1990 Rural-Urban Commuting Area (RUCA) Codes – Based on Census Bureau’s definitions of urbanized areas and urban places (population density and commuting patterns) – Acknowledges great variation across rural areas – Developed based on census tract and crosswalked to zip code • Four categories created: Urban, Large Rural, Small Rural, and Isolated Rural communities Sorbero 7 6-3-07 Sorbero 8 6-3-07 Methods – Survival Analysis Methods – Survival Analysis • Variables entered into model in stages – RUCA codes and demographic variables • Cox proportional hazard models – Hi(t) = λ 0(t ) exp{βj1xi1+...+ βjkxik} • Age, gender, race, marital status, number of comorbidities – Breast cancer variables – Parametric tests of proportional hazards assumption • Overall survival time in months – Date of diagnosis (mid-point of month) to date of death – Survivors censored at end of study period • Year of diagnosis, stage, ER and PR status – Sociodemographic and supply variables • 15% + of population not speaking English well, median household income, and Medicaid status • HPSA Residence, number of radiation oncologists and number of hospital oncology services per 10,000 population 65+ Sorbero 9 6-3-07 Sorbero 10 6-3-07 Results – Sample Characteristics Results – Sample Characteristics Whole Sample Urban Large Rural Small Rural Isolated Rural In situ 13.6 14.2 10.8 10.2 11.0 1 46.8 46.6 48.4 47.0 47.7 31.1 Variable Whole Sample Urban Large Rural Small Rural Isolated Rural Mean Age** 76.0 (6.9) 75.9 (6.8) 75.8 (7.0) 76.7 (7.1) 76.7 (7.2) Variable Stage (%) ** % Married** 43.9 43.2 47.1 47.0 48.6 2 29.0 28.7 29.7 31.1 % Black** 6.3 7.5 0.6 0.1 0.1 3 5.0 5.0 5.2 5.8 4.5 % Medicaid 11.5 11.7 10.8 11.1 10.4 4 3.3 3.3 3.6 3.3 2.6 Mean Comorbidity* 1.8 (1.7) 1.8 (1.7) 1.6 (1.5) 1.5 (1.5) 1.6 (1.5) Mean Survival** 65.4 (26.4) 65.6 (26.5) 64.4 (26.2) 64.1 (26.0) 64.5 (26.6) ** p<0.01; * p<0.05 Unstaged 2.3 2.2 2.3 2.7 3.2 ER positive (%)** 59.3 58.4 62.6 65.6 63.4 PR positive (%)** 49.1 48.0 53.8 56.4 54.9 ** p<0.01; * p<0.05 Sorbero 11 6-3-07 Sorbero 12 6-3-07 2 Results – Sample Characteristics Multivariate Results Whole Sample Urban Large Rural Small Rural Isolated Rural Not speaking English well (%)** 14.8 16.5 10.8 3.0 3.6 Median income <30,000 (%)** 33.0 23.8 65.3 92.3 HPSA (%)** 78.9 82.4 58.9 Mean N radiation oncologists.** 1.3 (1.1) 1.5 (1.1) Mean N hospitalbased oncology services** 0.8 (0.8) 0.7 (0.3) Variable RUCA + Demographic + Breast Cancer + SES & Supply Variable Hazard Ratio Hazard Ratio Hazard Ratio Urban Referent Referent Referent 84.1 Large Rural 1.19**** 1.13** 1.06 60.6 63.9 Small Rural 1.13*** 1.06 0.95 0.4 (0.6) 0.2 (0.5) 0.1 (0.4) Isolated 1.07 1.04 0.92 1.0 (0.9) 1.8 (1.8) 1.8 (2.2) **** p<.0001; *** p<.001; ** p<.01 ** p<0.01; * p<0.05 Sorbero 13 6-3-07 Sorbero 14 6-3-07 Multivariate Results - Supply Summary Full Model Variable Hazard Ratio HPSA County 1.06* Radiation Oncologists - Middle Tertile 0.99 Radiation Oncologists - Highest Tertile 1.05 Hospitals-based Oncology Services – Middle Tertile 0.90*** Hospitals-based Oncology Services – Highest Tertile 0.96 • Rural residence defined by RUCA categories not consistently associated with mortality following a breast cancer diagnosis • Controlling for demographics, higher mortality in large rural and small rural categories • Residing in county with partial or whole HPSA designation associated with increased mortality, while increased supply of hospital-based oncology services associated with decreased mortality **** p<.0001; *** p<.001; ** p<.01; * p<.05 Sorbero 15 6-3-07 Limitations Sorbero 16 6-3-07 Conclusions • Measures of supply based on county not RUCA • Women with breast cancer in rural areas experience codes • Hospital-based oncology services a proxy for all such providers • Did not examine disease-free survival • Findings may not be generalizable to Medicare beneficiaries enrolled in managed care or nonelderly Sorbero 17 6-3-07 greater mortality • Individual and regional socioeconomic factors associated with risk of mortality • Some evidence provider supply associated with mortality in elderly breast cancer patients • Policies should be developed to address provider shortages in both rural and urban areas Sorbero 18 6-3-07 3