Machine Learning Model to Predict Postoperative Respiratory Failure

Development and Prospective Evaluation of a Machine Learning Model to Predict Postoperative Respiratory Failure

The main objective of this study is to develop a machine learning model that predicts postoperative respiratory failure within 7 postoperative day using a real-world, local preoperative and intraoperative electronic health records, not administrative codes.

Study Overview

• Status: Completed
• Conditions: Noncardiac Surgery
• Intervention / Treatment: Diagnostic test: Prediction of postoperative respiratory failure using a machine learning

Detailed Description

Postoperative pulmonary complications are known to increase the length of hospital stay and healthcare cost. One of the most serious form of these complications is postoperative respiratory failure, which is also associated with morbidity and mortality. A lot of risk stratification models have been developed for identifying patients at increased risk of postoperative respiratory failure. However, these models were built by using a traditional logistic regression analysis. A logistic regression analysis had disadvantages of assuming the relationship between dependent and independent variables as linear. Recent advances in artificial intelligence make it possible to manage and analyze big data. Prediction model using a machine learning technique and large-scale data can improve the accuracy of prediction performance than those of previous models using traditional statistics. Furthermore, a machine learning technique may be a useful adjuvant tool in making clinical decisions or real-time prediction if it is integrated into the healthcare system. However, to our knowledge, there was no study investigating the predictive factors of postoperative respiratory failure using a machine-learning approach. Therefore, the main objective of this study is to develop a machine learning model that predicts postoperative respiratory failure within 7 postoperative day using a real-world, local preoperative and intraoperative electronic health records, not administrative codes and evaluate its performance prospectively.

• Study Type: Observational
• Enrollment (Actual): 22250

Contacts and Locations

Study Locations

• Korea, Republic of
  • Seoul, Korea, Republic of
    • Hyun-Kyu Yoon

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: 18 years and older (ADULT, OLDER_ADULT)
• Accepts Healthy Volunteers: No
• Genders Eligible for Study: All

• Sampling Method: Non-Probability Sample

Study Population

Adult patients undergoing general anesthesia for noncardiac surgery

Description

Inclusion Criteria:

• Adults patients undergoing general anesthesia for noncardiac surgery

Exclusion Criteria:

• Age under 18 years
• Surgery duration < 1 hr
• Cardiac surgery
• Surgery performed only regional or local anesthesia, peripheral nerve block, or monitored anesthesia care
• Organ transplantation
• Patient with preoperative tracheal intubation
• Patients who had tracheostoma prior to surgery
• Patients scheduled for tracheostomy
• Surgery performed outside the operating room
• Length of hospital stay < 24 h

If the patients had multiple surgeries during the same hospital stays, we included the first surgical cases in the dataset.

Study Plan

How is the study designed?

Number of groups / cohorts

1

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
AI_PRF; Adults patients undergoing general anesthesia	Diagnostic test: Prediction of postoperative respiratory failure using a machine learning; The performance of a machine learning model to predict postoperative respiratory failure after general anesthesia within postoperative day 7 was tested prospectively.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
the incidence of postoperative respiratory failure after general anesthesia	Postoperative respiratory failure which was defined as mechanical ventilation >48 h or any reintubation after surgery	within postoperative day 7

Collaborators and Investigators

• Sponsor: Seoul National University Hospital

Study record dates

Study Major Dates

• Study Start (ACTUAL): May 26, 2021
• Primary Completion (ACTUAL): May 25, 2022
• Study Completion (ACTUAL): June 25, 2022

Study Registration Dates

• First Submitted: August 21, 2020
• First Submitted That Met QC Criteria: August 21, 2020
• First Posted (ACTUAL): August 26, 2020

Study Record Updates

• Last Update Posted (ACTUAL): September 1, 2022
• Last Update Submitted That Met QC Criteria: August 29, 2022
• Last Verified: August 1, 2022

More Information

Terms related to this study

• Keywords: Validation; Risk Factors; Artificial Intelligence; Postoperative respiratory failure
• Additional Relevant MeSH Terms: Respiratory Tract Diseases; Respiration Disorders; Respiratory Insufficiency
• Other Study ID Numbers: AI_PRF

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No