What is the clinical evidence from this study?

Study design: Observational. Population: Postoperative sore throat (POST) (n=834). Intervention: Neural network (NN) prediction model vs. Extreme gradient boosting (XGBoost) and Random forest (RF) models. Primary outcome: Area under the receiver operating characteristic curve (AUROC) for predicting POST in the external validation cohort (AUROC 0.81, 95% CI 0.74-0.89).

October 22, 2023Open Access

Leveraging artificial intelligence to identify high-risk patients for postoperative sore throat: An observational study

Key Result

The neural network model demonstrated superior generalization capabilities in predicting postoperative sore throat compared to XGBoost and random forest models in external validation, achieving an AUROC of 0.81.

Study Design

Type

Observational (n=834)

Multicenter

Yes

Structured PICO

Does a neural network model accurately predict postoperative sore throat in adult patients undergoing general anesthesia with endotracheal intubation?

Population

834 adult patients (≥18 years) with ASA physical status 1-3 who underwent general anesthesia with endotracheal intubation. Exclusions: mental disorder, recent upper respiratory tract infection, >1 intubation attempt, transferred to ICU with endotracheal intubation after operation, or lost to follow-up.

Intervention

Neural network (NN) machine learning model using 5 predictors (age, sex, endotracheal tube cuff pressure, endotracheal tube insertion depth, and time interval between extubation and first drinking of water)

Comparator

Extreme gradient boosting (XGBoost) and random forest (RF) machine learning models

Outcome

Presence of postoperative sore throat (POST) assessed on the first postoperative morning (roughly 12-24 h post-extubation) using the POST questionnairepatient reported

A neural network model using five clinical variables accurately predicts postoperative sore throat, outperforming other machine learning models and offering a tool for targeted prevention.

Main Result

Effect estimate: AUROC 0.81 (95% CI 0.74-0.89)

Absolute Event Rate: 0.81% vs 0.78%

Limitations

Sample size may be insufficient for thorough validation of the models
Retrospective data collection could introduce potential biases and confounders
Did not consider certain factors that may influence POST, such as the use of different airway devices, lubrication, and the anesthetist's experience
Did not assess the impact of implementing the models in clinical practice on patient outcomes and resource utilization

Abstract

Postoperative sore throat (POST) is a prevalent complication after general anesthesia and targeting high-risk patients helps in its prevention. This study developed and validated a machine learning model to predict POST. A total number of 834 patients who underwent general anesthesia with endotracheal intubation were included in this study. Data from a cohort of 685 patients was used for model development and validation, while a cohort of 149 patients served for external validation. The prediction performance of random forest (RF), neural network (NN), and extreme gradient boosting (XGBoost) models was compared using comprehensive performance metrics. The Local Interpretable Model-Agnostic Explanations (LIME) methods elucidated the best-performing model. POST incidences across training, validation, and testing cohorts were 41.7%, 38.4%, and 36.2%, respectively. Five predictors were age, sex, endotracheal tube cuff pressure, endotracheal tube insertion depth, and the time interval between extubation and the first drinking of water after extubation. After incorporating these variables, the NN model demonstrated superior generalization capabilities in predicting POST when compared to the XGBoost and RF models in external validation, achieving an area under the receiver operating characteristic curve (AUROC) of 0.81 (95% CI 0.74–0.89) and a precision–recall curve (AUPRC) of 0.77 (95% CI 0.66–0.86). The model also showed good calibration and clinical usage values. The NN model outperforms the XGBoost and RF models in predicting POST, with potential applications in the healthcare industry for reducing the incidence of this common postoperative complication.

Bookmark

View Full Paper

Bookmark

View Full Paper

Cite This Study

Zhou et al. (Sun,) conducted a observational in Postoperative sore throat (POST) (n=834). Neural network (NN) prediction model vs. Extreme gradient boosting (XGBoost) and Random forest (RF) models was evaluated on Area under the receiver operating characteristic curve (AUROC) for predicting POST in the external validation cohort (AUROC 0.81, 95% CI 0.74-0.89). The neural network model demonstrated superior generalization capabilities in predicting postoperative sore throat compared to XGBoost and random forest models in external validation, achieving an AUROC of 0.81.

synapsesocial.com/papers/6a07b3e29090d046d755d03a https://doi.org/https://doi.org/10.17305/bb.2023.9519

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: