INDIANAPOLIS, Ind. – Algorithms and analytics are now common used by professional sports, in sales forecasts, lending decisions and by car insurance providers. Managers and other decision makers no longer simply “go with their gut.” But doctors often remain reluctant to introduce such information when making medical decisions for patients.
In an article published in Science, Helen Colby, an assistant professor of marketing at the Indiana University Kelley School of Business in Indianapolis, and a co-author note that it is time for many doctors to stop relying on their use of mental shortcuts, usually called decision rules or heuristics, when making decisions about patient care with limited cognitive resources.
Colby, who researches health decision-making, and Meng Li, an associate professor and director of undergraduate studies in the Department of Health and Behavioral Sciences at the University of Colorado Denver, were invited to provide commentary on a paper also published in Science by Manasvini Singh, an assistant professor of health economics at the University of Massachusetts.
Colby and Li’s article, “Physicians’ flawed heuristics in the delivery room; Complicated childbirths can influence later decision-making,” also appears in the Oct. 15 edition of the journal.
Using electronic health records of more than 86,000 deliveries, Singh found that delivering physicians were influenced not just by the indications of the current patient but also by the outcome of their most recent previous delivery. For example, when a physician experienced a negative outcome with a vaginal delivery they were more likely to choose to deliver the next baby by cesarean section and vice versa.
“Most of the time, the heuristics do save time and resources and they produce pretty good outcomes. But in some situations, pretty good is not good enough,” Colby said. “When lives are on the line, any improvement in decision making can have life-saving consequences.”
Colby and Li highlight that a “win-stay-lose-shift” heuristic has been identified in other contexts as a learning strategy, but said it only works well in certain settings.
“In the medical context, this heuristic would be rational only if the specifics of the prior patient matched the specifics of the current patient and thus provided a useful learning experience. In that case, if one patient’s delivery went wrong, it can tell the physician that the same delivery plan may not work well with another patient with very similar characteristics and indications,” they wrote. “However, two patients who happen to have consecutive deliveries by the same physician are not expected to be highly similar.”