The purpose of the proposed study is to: 1) Assess the validity of claims data for estimating incidence, treatment, expenditures and outcomes for lung, colon and prostate cancer in the elderly; 2) Analyze trends in screening, initial therapy, followup monitoring and treatment of recurrences for these cancers; 3) Measure the effects of policy factors, including types of cancer facilities, state laws and competitive pressures; 4) Estimate the effects of these alternative treatment patterns on costs and outcomes of care; and 5) Develop comprehensive simulation models from the empirical findings to estimate the consequences of policy and treatment changes affecting the management of cancer in the U.S. The first objective addresses a National Cancer Institute Program Announcement, """"""""Cancer Surveillance Using Health Claims-Based Data Systems"""""""". These research questions will be examined using a large, national data base and instrumental variable (IV) techniques to control for omitted variables bias. The analysis is in three parts. The first part focuses on the reliability of Medicare claims data for epidemiologic and economic research on cancer. The second analysis focuses on trends in medical treatment patterns, expenditures and outcomes, and the policy and demographic factors influencing those trends. The third part of the analysis uses instrumental variables (i.e., variables that are correlated with choice of treatment, but uncorrelated with unobserved factors affecting health outcomes, such as patient health status) to identify the marginal effect of treatments on outcomes. The investigators have reduced the number of cancers to be studied from six to three: lung, prostate and colon cancer. Stage of cancer at diagnosis is divided into two basic categories: localized and metastatic. Cancer treatment is divided into four phases: detection, initial therapy, followup surveillance for patients with localized cancers and later treatment likely associated with recurrence. Treatments are limited to: invasive screening procedures (for prostate and colon cancer), degree of aggressive surgical management and the use of a broader range of aggressive measures (surgery, chemotherapy and/or radiation therapy). The principal outcome measures will be survival time and treatment-free survival months (months without a hospitalization or a cancer treatment). Expenditure measures will include total insured reimbursement (and patient deductibles and copayments) and estimated costs. (Table 8 on page 42 summarizes the operational definition of variables in the analysis.) The proposed analytic methods build on the investigators' approach to analysis of the effectiveness of cardiac catheterization. Because patients are not assigned randomly to treatments, the investigators are concerned that omitted variables, such as unobserved dimensions of patient health, might influence both treatment choice and patient outcomes. Their approach is to identify a variable, such as access to facilities that offer particular treatment technologies, to stand in as an """"""""instrument"""""""" for actual treatment in the analysis. If access to treatment is correlated with choice of treatment, but not with unobserved measures of patient health, then comparing treatment outcomes for patients with different levels of access can provide an unbiased estimate of the effect of different treatments on outcomes.
