Identifying patients at risk of increased health utilization following lumbar spine surgery

Adequate preoperative identification of patients at risk of significant healthcare utilization after surgery could help guide preoperative decision-making as well as postoperative patient management. While several studies have proposed mechanisms and risk factors for healthcare utilization, no studies have developed a prognostic machine learning model to quantify and functionalize predictions.

A cohort of lumbar fusion and lumbar decompression surgeries was queried from a tertiary academic medical center from 2002 to 2022. Patient and operative characteristics were systematically extracted for each surgery. Several machine learning algorithms were employed and optimized to predict high healthcare utilizers using an aggregate 90-day index. SHAP feature importance values were computed for the top performing model.

A total of 10,128 unique lumbar decompression surgeries and 2,890 unique lumbar fusion surgeries were included. The Random Forest model had the highest performance of tested models (AUROC of 0.766 for lumbar decompression, 0.727 for lumbar fusion). Both models outperformed the ASA benchmark model. The top three predictors of high health utilization in the decompression cohort included preoperative lumbar stenosis, preoperative benzodiazepine use, and preoperative neuromodulator use. The top three predictors for the fusion cohort included preoperative opioid use, preoperative neuromodulator use, and preoperative benzodiazepine use. Other Key variables spanned several domains including preoperative medication usage, patient demographics, and operative indications and characteristics.

This study demonstrates the successful creation of a prognostic machine learning model for prediction of high healthcare utilization within 90 days of lumbar spine surgery. These models, after external validation, have the potential to be instrumental aspects of a spine surgeon's workflow.