A team of Harvard researchers used machine learning to predict which patients undergoing a procedure for lower back pain should be observed for opioid dependence, reports AI in Healthcare.
The research, published in The Spine Journal, develops an algorithm to forecast the likelihood of opioid dependency. It incorporates information such as prior diagnosis of depression and whether a long-term painkiller prescription is required by the patient.
Four insights:
1. They studied the details of 5,400 patients who underwent surgery to relieve pain from herniated discs of the lower back.
2. Researchers found that 416 patients had opioid prescriptions of at least 90 to 180 days post-surgery.
3. The data was used to create five algorithms and explanatory data models. Limitations included a lack of data on opioid dose in oral morphine equivalents in old EHRs and a presumption that the opioids were taken as prescribed.
4. The researchers concluded that the prediction of increased risk of prolonged opioid prescription can result in management changes, offering further support to patients before surgery.