Abstract:
|
Statistical analyses of health interventions typically target standard parameters of efficacy and harm: Sensitivity, specificity and AUC's for diagnostic tests; mortality risk ratios for cancer screening tests, and hazard ratios for treatment. But for policy panels deciding how to intervene at the population level, alternative metrics of harm and benefit are often of greater interest. In the case of cancer screening, trials yield relative estimates of reduction in disease-specific deaths, but policy panels are often more interested in absolute benefits (number of lives saved per population screened). Similarly, although tests of novel diagnostic markers yield estimates of performance characteristics, markers will ultimately only be useful if they impact downstream outcomes such as (quality-adjusted) life expectancy in a cost-effective manner. This Roundtable will discuss opportunities, methods and challenges that accompany the process of translating standard statistical inferences into policy-relevant consequences. We will consider examples from the fields of disease prevention, screening and treatment and real dilemmas from panels such as the USPSTF in the US and NICE in the UK.
|