Instinct Outweighs Data When ICU Physicians Order Diagnostic Tests

But data become more important as time goes on.

Even in an age of clinical decision support, new research finds that physicians rely heavily on “gut instinct” when making choices about which tests to order.

Researchers at the Massachusetts Institute of Technology (MIT) wanted to better understand the degree to which physicians depend on patient data when making clinical decisions. To answer the question, scientists analyzed more than 60,000 records from patients admitted to the intensive care unit (ICU) of a Boston hospital over a 10-year period. Those records, including physician notations, were subjected to sentiment analysis, a language-processing algorithm that can determine whether the text suggests positive or negative sentiments. The idea was to compare what effect, if any, a physician’s feelings had on the number of diagnostic imaging tests ordered.

>> READ: Leo Celi and the ‘Holy Grail of Personalized Medicine’

If the analysis found no correlation between sentiment and tests ordered, it would suggest that physicians were relying solely on patient data to drive their test ordering decisions.

But researchers found otherwise. Instead, the data suggested that physician sentiment was a major indicator of how many tests would be ordered, particularly at the beginning of a hospital stay.

Co-lead author Tuka Alhanai, a Ph.D. candidate in MIT’s Department of Electrical Engineering and Computer Science, told Healthcare Analytics News™ that she and her colleagues were unsure what to expect when they began the research. They were “very surprised” by the distinct effect of sentiment.

“We think this shows that even when physicians have access to standard data, like age and severity of illness of a patient, they're still observing something more: Could it be the flushness of the patient's skin? Or reaction to treatment?” she said. “It could be that sentiment is a proxy to something more that is being observed by the physician.”

The fact that sentiment and gut instinct played a lesser role as time went on (particularly after the first two days of a patient’s hospital stay) suggests that physicians rely more heavily on data as more information becomes available through tests and observation.

The type of sentiment — positive or negative — also played a significant role in the behavior of physicians. When a physician had a positive sentiment about the patient’s condition, they tended to order fewer tests. However, when the doctor’s notes indicated a negative sentiment, the physician generally ordered more tests. That effect trailed off, however, when the physician had a very negative sentiment, perhaps indicating that physicians order fewer tests when they expect the patient to have a low chance of survival.

Although the study affirmed the importance of physicians’ instincts and observations when treating patients, Alhanai said she believes computer scientists will one day be able to encode sentiment and “gut instinct” data into clinical decision tools, potentially fostering integration of the currently inimitable expertise that comes with physician’s personal experience.

And that’s not all.

“We also think it may be possible to try to capture more of what the physician is observing — for example, by analyzing their speech or gaze,” she added.

For now, Alhanai and colleagues believe their research should be a starting point for additional study of the role of instinct and sentiment in the decision-making of physicians.

The study is titled, “How is the Doctor Feeling? ICU Provider Sentiment is Associated with Diagnostic Imaging Utilization.” The paper can be read in full at this link.

Get the best insights in healthcare analytics directly to your inbox.


How 1 Small Change Reduced Post-Op Opioids at 3 Hospitals

EHRs and Behavioral Interventions Can Fight Antibiotic-Resistant Superbugs

Rage Against the Machines