RSNA 2014 

Abstract Archives of the RSNA, 2014


HPS169

Developing a Heuristic Score for Resident Selection

Scientific Posters

Presented on December 3, 2014
Presented as part of HPS-WEB: Health Services Wednesday Poster Discussions

Participants

Lawrence Cabusora MD, Presenter: Nothing to Disclose
Judah Burns MD, Abstract Co-Author: Nothing to Disclose
Mordecai Koenigsberg MD, Abstract Co-Author: Nothing to Disclose

PURPOSE

The NRMP Match program requires each residency program to rank candidates in the program's order of preference. Established methods of ranking residents are time- and effort-intensive, operate by an irreproducible process, and are subject to bias at multiple levels, often by design. A heuristic scoring model is introduced, offering an adjunct to decision-making, that streamlines and objectifies the process of resident ranking.  

METHOD AND MATERIALS

Candidates were rated on a 0-5 scale on each of seven components: USMLE score, medical school attended, medical school grades, research, professionalism, initiative, and sociability. Component rating was performed on the day of interview, but, to minimize cross-list bias, the ratings were not converted to a score at that time.  After completion of the Match rank list via the classical method, the candidate ratings were summed into a score. The candidates were sorted by score, and the resulting order was compared with the classical rank list via rank correlation coefficient analysis.

RESULTS

78 candidates were evaluated. In the comparison of the classical rank list vs. the score ordering, the Kendall tau-b equaled 0.74, an indication of strong positive rank-correlation. The distribution of individual absolute deviations between the classical rank list ordering vs. the score ordering was as follows: 0-5=42%, 5-10=28%, 10-15=19%, 15-20=6%, >20=4%.

CONCLUSION

Scores generated by a parsimonious model framework using easily-apprehended individual ratings (more reproducible than classical evaluations) are well-rank-correlated with results obtained via the laborious classical method. Further model refinement is desirable and possible, although caution is advised to avoid overfitting the extant data at the cost of predictive validity.

CLINICAL RELEVANCE/APPLICATION

A heuristic scoring model for the evaluation of residency candidates offers an efficient, objective adjunct to a radiology program's ranking process for the NRMP Match.

Cite This Abstract

Cabusora, L, Burns, J, Koenigsberg, M, Developing a Heuristic Score for Resident Selection.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14007574.html