Explaining differences in chemotherapy utilization in ovarian cancer between health service areas

TitleExplaining differences in chemotherapy utilization in ovarian cancer between health service areas
Publication TypeJournal Article
Year of Publication2004
AuthorsPolsky, Daniel E., Katrina A. Armstrong, Thomas C. Randall, Richard N. Ross, X. Zhang, Orit Even-Shoshan, Sanford J. Schwartz, Paul R. Rosenbaum, and Jeffrey H. Silber
JournalJournal of Clinical Oncology
AbstractBackground: Identifying excess variation has proved to be a useful tool to better understand potential strengths and weaknesses of the health care system. This study aims to report on the extent of geographic variation in chemotherapy utilization in ovarian cancer between 38 U.S. Health Service Areas (HSA) utilizing the 11 tumor registries from SEER-Medicare, and to determine HSA characteristics associated with this variation. Methods: Medicare patients with a diagnosis of ovarian cancer between 1990 and 1999 were selected from the SEER registries (N=10,986). After exclusions for HMO status (N=1778), death < 30 days (N=810), and HSAs < 30 subjects (N=466), the sample size was 7932. Chemotherapy within a year of diagnosis was identified by Medicare billing codes. Sociodemographic and disease characteristics were obtained from SEER and Census data. HSA data were obtained from the Area Resource File of the Federal Bureau of Health Professions. Results: In order to study differences across HSAs, we fit a logit model describing each of the 38 HSAs, and explored the variation in the odds of receiving chemotherapy inside HSAs, controlling for patient age, stage and year of diagnosis. There were large and significant differences between HSAs, with ORs varying 4.6-fold, (chi-square for HSA was 151, df=37, p=0.000). In a second model we added race, income, and education. ORs for HSAs now varied by 5.9 fold (chi-square for HSA was 148, df=37, p=0.000). Finally, we removed the 38 fixed effects and replaced them with 7 HSA descriptors (%HMO, pop. density, MD/pop., oncologist/MD, ln pop., N hospitals, %hosp with oncology services). The chi-square for these HSA characteristics was 21 (df=7, p=0.004). The only HSA descriptor with a significant odds ratio was % hospitals with oncology services (OR=1.82, p=.012). Conclusions: Medicare patients with the same age, stage, and year of diagnosis had very different chemotherapy utilization within one year of diagnosis in different geographic areas as defined by the HSAs. An important HSA factor contributing to this excess variation in utilization of chemotherapy between HSAs was the density of oncology facilities inside each HSA. No significant financial relationships to disclose.