AI-driven Models with Effective Feature Selection Accurately Predict ICU Admission after Laparoscopic Cholecystectomy Open Access

Abstract

Laparoscopic cholecystectomy (LC) is widely recognized for its advantages over open cholecystectomy, including shorter hospital stays, reduced postoperative pain, and lower mortality. However, despite its favorable safety profile, some patients still experience complications that require intensive care unit (ICU) admission. This study aims to develop interpretable, AI-driven models to predict ICU admission following LC and to identify key patient-specific risk factors.We retrospectively analyzed data from 1,411 patients who underwent LC at Irvine Medical Center between 2017 and 2022. The primary outcome was ICU admission after surgery. A total of 51 variables were collected, including demographics, comorbidities, medications, laboratory results, and postoperative complications. Recursive feature elimination with cross-validation (RFECV) was applied to select the most predictive features.Five machine learning models were developed: Random Forest, Decision Tree, Support Vector Machine, Neural Network, and Logistic Regression. Model performance was assessed using area under the receiver operating characteristic curve (AUROC), recall, accuracy, F1 score, and Matthews Correlation Coefficient (MCC). Among all models, the Random Forest showed the highest performance, with an AUROC of 0.83.To enhance interpretability, SHapley Additive exPlanations (SHAP) were used to evaluate the contribution of each feature to model predictions. SHAP force plots provided individualized explanations, highlighting how specific features influenced ICU risk on a per-patient basis. Top predictors included prolonged length of stay, extended anesthesia duration, and non-routine discharge disposition.This study is the first to integrate interpretable AI models with SHAP visualizations for predicting ICU admission following LC. Our findings suggest that machine learning models can effectively identify high-risk patients and provide transparent, clinically relevant insights to guide decision-making.Future work should focus on external validation and real-time integration of these models into clinical decision support systems to improve risk stratification and optimize resource utilization in postoperative care.

Table of Contents

CHAPTER  1: INTRODUCTION 1

CHAPTER  2: METHODS 3

2.1  Data Source and Study Population 3

2.2  Study Variables and Primary Outcome 3

2.3  Data Preprocessing and Missing Data Handling 3

2.4  Model Development and Performance Evaluation 4

CHAPTER  3: RESULTS 6

3.1  Baseline Characteristics 6

3.2  Feature Selection 6

3.3  Individual Risk Analysis 7

3.4  Evaluation Metrics and Model Performance with Train and Test Datasets 7

CHAPTER  4: DISCUSSION 9

Appendix 12

Reference 20

About this Master's Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Rollins School of Public Health
Department	Biostatistics
Subfield / Discipline	Biostatistics - MPH & MSPH
Degree	M.P.H.
Submission	Master's Thesis
Language	English
Research Field	Artificial Intelligence Biology, Biostatistics Mathematics
Keyword	Artificial intelligence, laparoscopic cholecystectomy, ICU admission, random forest, risk prediction, Shapley additive exPlanation (SHAP).
Committee Chair / Thesis Advisor	Rameshbabu Manyam, Emory University
Committee Members	Zhaohui Qin, Emory University

Last modified

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	AI-driven Models with Effective Feature Selection Accurately Predict ICU Admission after Laparoscopic Cholecystectomy ()	2025-04-10 23:45:45 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions