Artificial intelligence could help improve the efficiencies and outcomes of spine surgery while reversing the field's ''unsustainable" trajectory of rising costs in the U.S., according to a paper published Jan. 6 in the Global Spine Journal.
Amid the shift to value-based care, there has been much emerging research on the use of AI to improve spine surgery outcomes while lowering costs. Michelle Lee, Matthew Grabowski, MD, Ghaith Habboub, MD, and Thomas Mroz, MD, of the Cleveland Clinic Foundation reviewed work published in the past year to evaluate the use of AI in spine care.
Seven takeaways from their paper:
1. Machine learning systems designed to accurately predict length of stay, discharge to non-home facility and early unplanned readmissions after spine surgery could help identify high-risk patients and factors that contribute to risk. These projections could enable hospitals to more efficiently allocate resources, reduce costly lengths of stay and readmissions, and maintain or improve care quality.
2. Because spine surgeries aren't easily studied through traditional methods, clinicians could use predictive algorithms derived from vast data repositories to standardize care.
3. Some researchers have developed machine learning models that accurately predict positive outcomes after surgery for degenerative cervical myelopathy, probability of failure of nonoperative management in spinal epidural abscesses, and 90-day post-discharge mortality in patients with spinal metastatic disease.
4. Clinical researchers have also used AI technology to predict negative outcomes, such as surgical complications in patients undergoing elective anterior cervical discectomy and fusion, posterior lumbar fusion, and adult spinal deformity surgeries. Some models even predicted prolonged opioid prescription after surgery for lumbar disc herniation.
5. Surgeons' subjectivity remains important for reviewing spine imaging and determining whether a patient should undergo surgery, but AI and machine learning algorithms are designed to help in that decision-making process. For instance, one model was able to predict the diagnosis and severity of cervical spondylotic myelopathy with high sensitivity and specificity.
6. A fully integrated AI platform can automate burdensome administrative tasks such as conducting postsurgical checks and ordering medications, thereby enhancing efficiencies, outcomes, patient satisfaction and employee engagement, the authors said.
7. The paper acknowledged that a lack of large, quality data sets and consensus about how to categorize issues could be barriers to using AI-driven platforms in spine care. These models can make false associations if based upon inadequate data, and work must be done to make the algorithms transfer from one facility to another.