Saja Al-Alawneh abstract: Providing radiologists with feedback has been shown to improve their performance in mammography diagnosis. In 1992, the Mammography Quality Standards Act (MQSA) was enacted to improve the quality of mammography using audit and feedback procedures. However, no standard audit and feedback system for radiologists has been installed in the United States. Instead, auditing typically requires human effort to manually correlate radiology and pathology results. Furthermore, the procedures that are currently in place focus primarily on compliance rather than providing individualized feedback. We conducted a descriptive study to inform development of a system to automate the correlation of breast imaging and pathology reports. We will discuss the results of this study, development of methods that build further on this initial study, and the design of a more complete evaluation of this methodology. This work lays a foundation for building an automated audit and feedback system for radiology diagnostic performance improvement.
Sam Rosko abstract: While in vitro drug mechanism studies are an important part of pre-market drug development, there remains a debate among drug interaction experts about the trustworthiness of using data from these kinds of studies to infer potential drug-drug interactions (PDDIs) for use in clinical decision support systems. In this study, our objective was to empirically determine the influence of in vitro data on PDDI inference accuracy. Toward that goal, we inferred PDDIs with and without mechanistic assertions based on in vitro metabolic enzyme, substrate, or inhibitor studies. We then compared the performance of various evidence strategies against a validation set of clinical pharmacokinetic studies. We also explored the generalizability of our results by comparing the drugs used in our study with both the universe of currently-marketed drugs and the set of drugs listed in the FDA’s organ impairment drug interaction database. The results show that there was a slight decrease in F-measure (0.031) and a slight increase in coverage of the validation set (0.057) with evidence strategies that use in vitro evidence. Our generalizability analysis showed that our results are applicable to most Cytochrome P-450 mediated drugs, but not to transport protein-mediated drugs. We found the tradeoff of F-measure for coverage to be favorable, but conclude that there is not a significant enough difference to warrant inclusion or exclusion of in vitro data. As future work, we plan to use large-scale electronic health records and claims data to assess the validity of any unevaluated PDDI pairs we generated by looking for changes in clinical outcomes in the presence of these pairs.