In two recent influential reports, the National Cancer Policy Board of the Institute of Medicine emphasized both the inadequacy of our present understanding of cancer care quality and the important role of the National Program of Cancer Registries (NPCR) in monitoring, examining variations in, and supporting improvement in cancer care. In response, the goals of the proposed study are to: describe patterns of care for female breast cancer and for prostate cancer among residents of the State of Georgia; compare how breast cancer and prostate cancer care vary across racial/ethnic (e.g., Caucasian and African-American) and socio-demographic (e.g., rural and urban) population strata; examine disease-, patient-, provider-, and health system-related factors influencing treatment choice; investigate the degree to which breast cancer and prostate cancer treatment in Georgia is guideline-concordant; and determine the extent to which the findings from each of first four pursuits above vary depending on whether the data for analysis (a) are derived solely from the Georgia Comprehensive Cancer Registry (GCCR), or (b) consist of GCCR data that have been augmented by information from inpatient and outpatient sources in the community. That is, in the spirit of sensitivity analysis, we propose to examine in what ways conclusions about patterns of care and quality of care are altered, and enhanced, as a result of expanding the empirical base beyond what is currently captured in the GCCR. This will yield insights into the value-added from specific approaches to improving the accuracy, completeness, and comprehensiveness of registry data. ? ? The empirical base for the study will be a stratified random sample from the GCCR of 2,500 breast cancer cases (stage l-lll) and 2,500 prostate cancer cases (all stages) that occurred between Jan 1, 2002 and Dec. 31, 2003 and that meet specified inclusion criteria. The sampling strata will be selected to ensure adequate representation of Caucasians and African-Americans from both rural and urban areas in the state. Trained data abstractors based in each of the five GCCR regions will abstract required information from medical records and submit completed data forms to the project data manager in Atlanta. Stringent data quality control checks will be in place throughout the study. Patterns-of-care variations in breast and prostate cancer and the extent to which care is guideline concordant will be examined through descriptive, univariate, stratified, and multivariable regression analyses. Binary and multinomial logic models will be used as appropriate to assess the main effects and interactions among patient-, provider-, and health system factors that are hypothesized to influence treatment selection and implementation, and also whether the care that is received is guideline concordant. ? ? ?
