American Statistical Association New York City Metropolitan Area Chapter Memorial Sloan-Kettering Cancer Center Biostatistics Seminar

Diagnostic and prognostic models are typically evaluated with measures of accuracy such as sensitivity, specificity, and area under the curve. Such measures often tell us little or nothing about a models clinical value, such as its impact on number of cancers found or unnecessary biopsies avoided. Decision-analytic techniques may allow assessment of clinical outcomes but they generally require collection of additional information, such as patient preferences or costs of treatment. We developed a novel method decision curve analysis to overcome the drawbacks of traditional statistical and decision analysis.

In this talk, I will: