Abstract
PURPOSE: We aimed to develop a clinical decision support tool for clinicians counseling patients with localized prostate cancer. The tool would provide estimates of patient life expectancy based on age, comorbidities and tumor characteristics. We reviewed the literature to find suitable prediction models. MATERIALS AND METHODS: We searched the literature for prediction models for life expectancy. Models were evaluated in terms of whether they provided an estimate of risk, incorporated comorbidities, were clinically feasible and gave plausible estimates. Clinical feasibility was defined in terms of whether the model provided coefficients and could be used in the initial consultation for men across a wide age range without an undue burden of data gathering. RESULTS: Models in the literature were characterized by the use of life years rather than a risk of death, questionable approaches to comorbidities, implausible estimates, questionable recommendations and poor clinical feasibility. We found tools that involved applying an unvalidated approach to assessing comorbidities to a clearly erroneous life expectancy table, or requiring that a treatment decision be made before life expectancy could be calculated, or giving highly implausible estimates such as a substantial risk of prostate cancer specific mortality even for a highly comorbid 80-year-old with Gleason 6 disease. CONCLUSIONS: We found gross deficiencies in current tools that predict risk of death from other causes. No existing model was suitable for implementation in our clinical decision support system.