An Intelligent Patient Admission Model of Day Surgery Using Heterogeneous Data with Semi-Supervised Learning: A Case Study of Laparoscopic Cholecystectomy (Preprint)
Preprint
- 12 November 2021
- preprint
- Published by JMIR Publications Inc. in JMIR Preprints
Abstract
BACKGROUND Day surgery has many advantages including shortening hospital stay, decreasing the risk of hospital-associated infections, and increasing cost efficiency over traditional surgery, it has gained a great reputation and popularity in recent years. However, the patients’ admission criteria of day surgery at present were mainly based on expert experience, which was a lack of scientific evidence. OBJECTIVE Our study is to investigate the day surgery patient’s admission criteria and build an intelligent machine learning model of day surgery patients who underwent laparoscopic cholecystectomy, to ensure patients’ safety and medical quality, providing reference and inspiration for other day surgery admission decisions. METHODS We analyzed the clinical data of day surgery patients who underwent laparoscopic cholecystectomy at West China Hospital from Jan 1st 2009 to Dec 31st 2021 and developed a semi-supervised artificial intelligence algorithm, SDSPA algorithm, which is built by self-training and uses both structured data like patient characteristics and unstructured clinical diagnosis to assist surgeons to make quick admission decisions. RESULTS After comparing several classifiers with self-training in our experiment, the performance of LightGBM with unstructured text processed by BERT were the best, obtaining an accuracy of 0.85 and an f1-score of 0.83, as well as reaching 0.97 on the precision score, which is an important indicator related to patients’ safety. CONCLUSIONS The application of our SDSPA algorithm can make the patient admission of day surgery more intelligent, and maximize the utilization of medical resources while ensuring patients’ safety.Keywords
This publication has 21 references indexed in Scilit:
- Extreme Gradient Boosting Model Has a Better Performance in Predicting the Risk of 90-Day Readmissions in Patients with Ischaemic StrokeJournal of Stroke and Cerebrovascular Diseases, 2019
- Emergency department triage prediction of clinical outcomes using machine learning modelsCritical Care, 2019
- Natural language processing of symptoms documented in free-text narratives of electronic health records: a systematic reviewJournal of the American Medical Informatics Association, 2019
- Effective Identification of Similar Patients Through Sequential Matching over ICD Code EmbeddingJournal of Medical Systems, 2018
- Admission Control Policies for Surgery Patients.Journal of Medical Systems, 2017
- A Self-Training Subspace Clustering Algorithm under Low-Rank Representation for Cancer Classification on Gene Expression DataIEEE/ACM Transactions on Computational Biology and Bioinformatics, 2017
- XGBoost2016
- Pattern classification and clustering: A review of partially supervised learning approachesPattern Recognition Letters, 2014
- MissForest—non-parametric missing value imputation for mixed-type dataBioinformatics, 2011
- A STATISTICAL INTERPRETATION OF TERM SPECIFICITY AND ITS APPLICATION IN RETRIEVALJournal of Documentation, 1972