Evaluation of the Success of ChatGPT-4 in Predicting Postoperative Intensive Care Needs and Mortality: Prospective Observational Study
Evaluation of the Success of ChatGPT-4 in Predicting Postoperative Intensive Care Needs and Mortality: Prospective Observational Study
This is a prospective, observational study to be conducted at Sağlık Bilimleri University Istanbul Kanuni Sultan Süleyman Training and Research Hospital and Başakşehir Çam and Sakura City Hospital. The study aims to record various preoperative and postoperative data of patients who have undergone surgeries, specifically those with ASA scores of III and IV or those indicated to potentially need postoperative intensive care. Data points include patient demographics, type of surgery, ASA score, comorbidities, lab and imaging findings, and both actual and ChatGPT version 4 predicted outcomes regarding postoperative intensive care needs, anesthesia methods, duration of stay in intensive care and the hospital, and 30-day mortality rates. ChatGPT version 4's predictions will be compared with actual outcomes and anesthesiologist decisions.
A Prospective, Observational Study to be Conducted at Sağlık Bilimleri University Istanbul Kanuni Sultan Süleyman Training and Research Hospital, Başakşehir Çam and Sakura City Hospital. In the study, age, surgery, additional diseases, abnormal laboratory findings, and imaging results of patients who have undergone preoperative anesthesia examination and received an ASA score of III and IV, or have been indicated to potentially need postoperative intensive care during the anesthesia examination will be recorded. Patients' surgeries are stated to be without complications, and predictions will be requested from the ChatGPT version 4 regarding the need for postoperative intensive care monitoring, recommended anesthesia method, strategies to reduce mortality, duration of stay in intensive care, and duration of hospital stay. These predictions will be compared with the decisions given by the anesthesiologist.
We will record these data:
Age: Patient's age. Gender: Patient's gender. Type of Surgery: The specific surgery the patient underwent. ASA Score: The American Society of Anesthesiologists (ASA) score indicating the patient's preoperative physical status.
Additional Diseases: Any comorbid conditions the patient has. Significant Laboratory Findings: Key lab results that could influence patient care.
Imaging Findings: Results from imaging studies relevant to the patient's condition or surgery.
ChatGPT-4's Intensive Care Prediction: Prediction made by ChatGPT version 4 regarding the need for postoperative intensive care.
Actual Need for Intensive Care: Whether the patient actually required postoperative intensive care.
Recommended Type of Anesthesia (ChatGPT-4): Anesthesia method suggested by ChatGPT version 4.
Type of Anesthesia Administered: The anesthesia method actually used during surgery.
Duration of Stay in Intensive Care: The actual length of time the patient spent in intensive care.
Intensive Care Stay Prediction (ChatGPT): ChatGPT version 4's prediction of how long the patient would need to stay in intensive care.
Total Hospital Stay Duration: The actual total length of the patient's hospital stay.
Total Hospital Stay Duration (ChatGPT): Prediction by ChatGPT version 4 of the total duration of the patient's hospital stay.
Mortality (Within 30 Days): Whether the patient died within 30 days of surgery. Mortality Prediction (ChatGPT): ChatGPT version 4's prediction regarding the patient's risk of mortality within 30 days post-surgery.
At Kanuni Sultan Süleyman Training and Research Hospital, the patient will be under the care of Specialist Doctor Engin İhsan Turan, and at Başakşehir Çam and Sakura City Hospital, under the care of Specialist Doctor Abdurrahman Engin Baydemir.
Primary Objective: To evaluate the success of ChatGPT-4 in predicting postoperative intensive care needs and mortality in adult patients with ASA scores of III and IV.
Secondary Objectives: To examine the effectiveness of ChatGPT-4's anesthesia method recommendations and additional suggestions in the clinical decision-making process.
Benefits: Understanding the contributions of artificial intelligence-based systems to clinical decision-making processes.
Risks: The potential for ChatGPT-4's recommendations to be misleading, however, the fact that doctors are the final decision-makers will mitigate this risk.
For the purpose of conducting statistical analysis, the data provided by ChatGPT will be compared with the actual data using IBM SPSS 21 software. For the categorical variable of the need for intensive care, the McNemar Exact test will be conducted.
Inclusion Criteria:
Exclusion Criteria:
Patients with ASA score I-II-V-VI. Patients under 18 years of age Emergency cases Operations with surgical complications Operations with anesthetic complications