Crystal Ball or Magic8 Ball? Reply Hazy, Try Again

Excerpt

In this issue of Annals of Surgical Oncology, Zabor et al.1 present a comparison of three online melanoma prediction tools. By applying data inputs from a single validation set, the authors demonstrate variability in prediction of individual survival across the tools. The findings highlight important distinctions not only in how the tools were developed, but also in the potential for the tools to contain inherent bias, the breadth of the tool’s use or applicability, and the utility of a tool for its intended population (e.g., its ability to correctly discriminate patients with different levels of risk). The article also emphasizes the importance of understandable information representation as it applies to data interpretation and risk communication for health care providers and patients. We applaud the authors for highlighting several practical concerns about prediction tools and their unregulated availability, including the lack of documentation on their assumptions and performance. …