The following is the summary of “Predicting Bone Health Using Machine Learning in Patients undergoing Spinal Reconstruction Surgery” published in the January 2023 issue of Spine by Shen, et al.
Analysis of past results using present-day data. The purpose of this research is to use machine learning to develop a prediction model of patients’ bone health before they undergo adult spinal reconstruction (ASR) using machine learning (ML). However, spine surgeons lack the resources to risk stratify patients preoperatively to determine who should receive bone health screening because of the knowledge that bone health affects the results of spine surgery. To assess the likelihood of poor bone health after ASR, ML is used to mine data patterns. In this analysis, researchers looked at scans from 211 patients over the age of 30 who had undergone spinal reconstruction surgery and had been subjected to dual-energy X-ray absorptiometry.
Both human and machine labor was used to extract information from the electronic health records. Models for multiclass categorization into “healthy,” “osteopenia,” and “osteoporosis” (OPO) categories were developed with the help of the Weka program. Using dual energy X-ray absorptiometry T scores, bone status was classified in accordance with World Health Organization (WHO) guidelines. Area under the receiver operating curve (AUC), accuracy, sensitivity, and specificity were determined. The model’s capacity to generalize was tested using new data. There was a 23.22% rate of OPO and a 52.61% rate of osteopenia. On the training set, the random forest model performed best, with an average sensitivity of 0.81, specificity of 0.95, and AUC of 0.96.
On average, the model’s test set results were an AUC of 0.69, a sensitivity of 0.64, and a specificity of 0.78. In terms of forecasting OPO in patients, the model performed exceptionally well. Predictive value was found for a wide variety of patient characteristics, including BMI, insurance, serum salt, serum creatinine, prior bariatric surgery, and drug use (particularly selective serotonin reuptake inhibitors). In ASR patients, bone health status prediction can be achieved by an ML-based strategy. Moreover, data mining with ML can unearth previously undiscovered risk variables for bone health in patients undergoing ASR surgery.