Opioid Resource Center
Opioid Resource Center
Perspective from David Dickerson, MD
Source/Disclosures
Source: Soens MS, et al. Development and comparison of machine learning models for prediction of postoperative opioid use. Presented at: Anesthesiology 2020; Oct. 2-5, 2020 (virtual meeting).
Disclosures: Healio Primary Care could not confirm relevant financial disclosures at the time of publication.
October 16, 2020
2 min read
Save

AI predicts postoperative opioid use

Perspective from David Dickerson, MD
Source/Disclosures
Source: Soens MS, et al. Development and comparison of machine learning models for prediction of postoperative opioid use. Presented at: Anesthesiology 2020; Oct. 2-5, 2020 (virtual meeting).
Disclosures: Healio Primary Care could not confirm relevant financial disclosures at the time of publication.
You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact customerservice@slackinc.com.

An artificial intelligence tool based on data from patients’ electronic medical records correctly identified which patients would need high doses of opioids after surgery, according to findings presented at Anesthesiology 2020.

“Being able to target the right treatment to the right patient is important to not only to reduce opioid use, but also to ensure that patients receive the treatment that is right for them,” Mieke S. Soens, MD, an instructor of anesthesia at Brigham and Women's Hospital, told Healio Primary Care.

 The quote is: "The tool would also be available for primary care physicians, as long as they have access to the patient’s EMR." The source of the quote is: Mieke S. Soens, MD.

In a two-part study, Soens and colleagues gathered data from 5,994 patients scheduled to receive general anesthesia and undergo surgical procedures without a peripheral nerve block. Of these, 1,287 were administered more than 90 morphine milligram equivalents in the first 24 hours after surgery.

In the first part of the study, the researchers used 163 potential factors to predict high pain after surgery based on a literature search and input from experts. Based on those factors, they created three machine learning algorithm models — logistic regression, random forest and artificial neural networks. The algorithms compiled data from patient medical records and shortened the list of predictive factors down to 21 that most accurately predicted patients’ pain severity and their potential need for opioids after surgery.

In the second part of the study, Soens and colleagues compared the models’ predictions for opioid use and the patients’ actual opioid use. They found that all three models had comparable accuracy overall: 81% for logistic regression and random forest, and 81% for artificial neural networks.

“Our model will allow the surgical and anesthesia teams to create a tailored personalized approach for each patient that maximize nonopioid analgesic strategies for patients, including nerve blocks and epidurals,” Soens said. “Patients can experience less pain and get optimal doses of opioid pain medication after surgery, and we also hope to reduce the risk for chronic opioid use.”

The researchers hope to partner with EMR vendors to integrate their model into more health systems, Soens said.

“The amount of work that has to be done and associated costs would depend on the EMR system,” she said. “While this tool is mostly designed to help perioperative care teams create an individualized pain management approach for the surgical patient, the tool would also be available for primary care physicians, as long as they have access to the patient’s EMR.”