Machine learning in spine: How an algorithm predicts risk of prolonged opioid use after surgery: 4 notes

Written by Laura Dyrda | June 12, 2019 | Print  |

A new study published in The Spine Journal examines machine learning algorithms for predicting the risk of prolonged opioid use after lumbar disc herniation surgery.

 

The researchers conducted a chart review of outcomes for patients who underwent lumbar disc herniation repair surgery from Jan .1, 2000, to March 1, 2018, at five medical centers. The data was then put through five different models designed to predict prolonged opioid prescription after surgery.

Four findings:

1. Of the 5,413 patients included in the study, 7.7 percent reported prolonged opioid prescriptions at 90 to 180 days post-surgical.

2. The most accurate prediction model was the elastic-net penalized logistic regression model. Researchers reported good calibration as well as overall performance, according to the study.

3. The important predictors associated with prolonged opioid prescription after lumbar disc herniation surgery were:

• Instruments used.
• Length of time the patient had opioid prescription before surgery.
• Depression.

4. Study authors concluded the predictive models can identify candidates at high risk for prolonged opioid use that may require "increased surveillance" postoperatively.

The models now exist online and can provide predictions and patient-specific explanation for results, which you can find here.

More articles on spine surgery:
Penn State, Geisinger partner to expand neurosurgical care
5 compensation models hospitals and health systems use when hiring new neurosurgeons
6 spine, neurosurgeons making the news

© Copyright ASC COMMUNICATIONS 2019. Interested in LINKING to or REPRINTING this content? View our policies here.

Top 40 Articles from the Past 6 Months