Unbiased Outcome Estimates for Conservative vs. Aggressive Treatment of Early Stage Prostate Cancer from Retrospective Data: An Instrumental Variables Approach
Abstract
The purpose of this study is to use instrumental variables techniques to estimate the outcome differences between aggressive treatment and conservative management among marginal patients with early stage prostate cancer; estimate true cost-effectiveness ratios; and determine whether and what type of patients may be safely shifted from aggressive to conservative treatment. In this annual report we describe the factors that are related to the sorting of patients to receive conservative or aggressive treatments. We have demonstrated that selected access to care variables (the candidate instrumental variables) group patients such that the groups have different rates of aggressive treatment but do not differ meaningfully in demographic, tumor, or comorbidity characteristics. Thus, these variables appear to satisfy the two criteria that an instrument must fulfill: (1) the variable must be related to the possibility of patients receiving a particular treatment; and (2) the variable must have no effect on outcomes either directly or indirectly (e.g., through relationships with unmeasured confounding factors such as patient severity and unrecorded treatments). If the candidate instruments had failed this test, it would not have been possible to proceed to Task 2, estimating unbiased treatment effects for conservative vs. aggressive treatment.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1999
- Accession Number
- ADB262564
Entities
People
- Elizabeth A. Chrischilles
Organizations
- University of Iowa