Development and validation of an artificial intelligence system for surgical case length prediction.
Adhitya Ramamurthi1, Bhabishya Neupane2, Priya Deshpande3
1Division of Surgical Oncology, Department of Surgery, Medical College of Wisconsin, Milwaukee, WI. Electronic address: https://twitter.com/anaikothari.
Surgery
|November 29, 2024
Summary
An artificial intelligence system significantly improves surgical case length prediction, outperforming current electronic health record estimates by 62%. This AI model enhances operating room efficiency through more accurate surgical time forecasting.
Area of Science:
- Artificial Intelligence in Healthcare
- Machine Learning for Predictive Analytics
- Natural Language Processing in Medicine
Background:
- Accurate surgical case length estimation is crucial for operating room (OR) optimization.
- Current methods for predicting surgical duration suffer from significant inaccuracies.
- Improving prediction accuracy can lead to better resource allocation and efficiency in surgical settings.
Purpose of the Study:
- To develop and validate an artificial intelligence (AI) system for enhanced surgical case length prediction.
- To leverage natural language processing (NLP) and machine learning (ML) techniques for improved predictive accuracy.
- To address the limitations of existing solutions in surgical duration forecasting.
Main Methods:
- Utilized a dataset of 125,493 inpatient elective surgical cases (2017-2023).
- Trained and evaluated multiple ML models, including Linear Regression, CategoricalBoost, and Feed-Forward Neural Networks.
- Employed bidirectional encoder representations from transformers (BERT) embeddings pretrained on clinical text for feature extraction.
Main Results:
- The top-performing model, CategoricalBoost Regressor with BERT embeddings, achieved a mean absolute error (MAE) of 46.4 minutes.
- This AI model's MAE was substantially lower than existing electronic health record (EHR) estimates (120.0 minutes, P < 0.001).
- The AI model achieved 48% accuracy in predicting case length within ±20% of actual duration, compared to 17% for EHR estimates.
Conclusions:
- An AI-driven system for surgical case length prediction demonstrates superior performance over current EHR-based predictions.
- The developed AI model improved estimate accuracy by 62% and correctly estimated approximately 2.8 times more cases.
- This study highlights the successful application of advanced NLP and ML techniques for more precise surgical duration forecasting.
