Perioperative Predictors of Early Spinal Cord Stimulator Removal: A Machine Learning-Assisted Retrospective Cohort Study

Background: Spinal cord stimulators can offer an effective treatment in chronic pain refractory to conventional medical management. However, with a failure rate of up to 44% and an annual explant rate of 6-9%, there is a need to better identify patients at high risk of therapeutic failure. The objective of this retrospective cohort study was to determine predictors of early SCS explant following device placement. Methods: The Medical Informatics Operating room Vitals and Events Repository database was queried for patients with a spinal cord stimulator and at least two years of follow up (n = 56). A multivariate logistic regression was fitted. Recursive factor elimination and bootstrap validation were used to minimize risk of overfitting. The model was used to predict risk factors for explant, odds ratio (OR), and 95% confidence interval (CI). Results: The final model displayed good performance with a bootstrap mean Area Under the Receiver Operating Curve of 0.91 (bootstrap CI: 0.74 – 1.0) and bootstrap mean accuracy of 82.1% (bootstrap CI: 63.0% – 97.7%). Fibromyalgia (OR: 2.38; CI: 2.34 – 2.41), hyperlipidemia (OR: 2.24; CI: 2.20 – 2.27), sleep disorder (OR: 2.10; CI: 2.07 – 2.14), obesity (OR: 1.93; CI: 1.90 – 1.95), and irritable bowel syndrome (OR: 1.63; CI: 1.61 – 1.65) displayed statistically significant increased risk of explantation. Conclusions: A medical history of fibromyalgia, hyperlipidemia, sleep disorders, obesity, and irritable bowel syndrome are novel risk factors for spinal cord stimulator explantation. While further prospective studies are needed, our study would suggest these factors may be worth considering in pre-operative evaluation.