Predicting neurosurgical referral outcomes in patients with chronic subdural hematomas using machine learning algorithms - A multi-center feasibility study.

Background: Chronic subdural hematoma (CSDH) incidence and referral rates to neurosurgery are increasing. Accurate and automated evidence-based referral decision-support tools that can triage referrals are required. Our objective was to explore the feasibility of machine learning (ML) algorithms in predicting the outcome of a CSDH referral made to neurosurgery and to examine their reliability on external validation. Methods: Multicenter retrospective case series conducted from 2015 to 2020, analyzing all CSDH patient referrals at two neurosurgical centers in the United Kingdom. 10 independent predictor variables were analyzed to predict the binary outcome of either accepting (for surgical treatment) or rejecting the CSDH referral with the aim of conservative management. 5 ML algorithms were developed and externally tested to determine the most reliable model for deployment. Results: 1500 referrals in the internal cohort were analyzed, with 70% being rejected referrals. On a holdout set of 450 patients, the artificial neural network demonstrated an accuracy of 96.222% (94.444-97.778), an area under the receiver operating curve (AUC) of 0.951 (0.927-0.973) and a brier score loss of 0.037 (0.022-0.056). On a 1713 external validation patient cohort, the model demonstrated an AUC of 0.896 (0.878-0.912) and an accuracy of 92.294% (90.952-93.520). This model is publicly deployed: https://medmlanalytics.com/neural-analysis-model/. Conclusion: ML models can accurately predict referral outcomes and can potentially be used in clinical practice as CSDH referral decision making support tools. The growing demand in healthcare, combined with increasing digitization of health records raises the opportunity for ML algorithms to be used for decision making in complex clinical scenarios.

Natural language processing for the automated detection of intra-operative elements in lumbar spine surgery.

Background: The aim of this study was to develop natural language processing (NLP) algorithms to conduct automated identification of incidental durotomy, wound drains, and the use of sutures or skin clips for wound closure, in free text operative notes of patients following lumbar surgery. Methods: A single-centre retrospective case series analysis was conducted between January 2015 and June 2022, analysing operative notes of patients aged >18 years who underwent a primary lumbar discectomy and/or decompression at any lumbar level. Extreme gradient-boosting NLP algorithms were developed and assessed on five performance metrics: accuracy, area under receiver-operating curve (AUC), positive predictive value (PPV), specificity, and Brier score. Results: A total of 942 patients were used in the training set and 235 patients, in the testing set. The average age of the cohort was 53.900 ± 16.153 years, with a female predominance of 616 patients (52.3%). The models achieved an aggregate accuracy of >91%, a specificity of >91%, a PPV of >84%, an AUC of >0.933, and a Brier score loss of ≤0.082. The decision curve analysis also revealed that these NLP algorithms possessed great clinical net benefit at all possible threshold probabilities. Global and local model interpretation analyses further highlighted relevant clinically useful features (words) important in classifying the presence of each entity appropriately. Conclusions: These NLP algorithms can help monitor surgical performance and complications in an automated fashion by identifying and classifying the presence of various intra-operative elements in lumbar spine surgery.