NEW research has shown that a deep learning model may be able to predict disability and pain in patients up to 1 year after lumbar disc herniation surgery. These findings suggest that such models may be able to aid decision-making for patients and surgeons alike, as well as providing accurate prognosis information to clinicians. Thus far, though machine learning algorithms are used increasingly often for analysing data, their use for outcome prediction in spinal surgery remains largely unknown.
Bjørnar Berg, Oslo Metropolitan University, Norway, and team, used retrospective data from the Norwegian Registry for Spine Surgery (NORspine), to assess the results of 21,161 patients (mean age of 47 years; 12,952 [57%] males), who underwent 22,707 lumbar disc herniation surgeries between January 2007–May 2021. Researchers used a validated machine learning model in order to predict treatment success in terms of disability and pain 1 year after surgery. Treatment success was defined in this study as improvements in the Oswestry Disability Index (ODI) of 22 points or more, Numeric Rating Scale (NRS) for back pain of at least two points, and NRS for leg pain of four points or more. Performance of the deep learning model was assessed through discrimination and calibration.
The team found that 33% of cases were unsuccessful according to ODI, 27% according to NRS for back pain, and 31% according to NRS for leg pain. The machine learning model showed consistent discrimination and calibration across all regions, according to researchers. Only minor heterogeneity was found in calibration slopes and intercepts.
This research provides valuable insights into the applications of deep learning algorithms, suggesting that such models may be able to inform patients and clinicians about individual prognosis in the future. Additionally, deep learning models may one day “aid in surgical decision-making to ultimately reduce ineffective and costly spine care,” stated the team.
