Contrast Between Traditional Regression Model and AI in Predicting Prolonged Stay Stay After Head and Neck Tumors

Sun Yat-sen University

Status: Enrolling

Conditions: Head and Neck Cancer

Study type: Observational

Funder types: Other

Identifiers

NCT06570486

SYSKY-2024-403-01

Details and patient eligibility

About

This experiment is an observational study of cohort. By establishing a cohort of patients with head and neck tumors transferred to ICU after surgery, investigators compared the prediction effect of AI and the traditional prediction model on whether patients can be transferred to ICU within 24 hours of head and neck tumors. First retrospective analysis of patients after head and neck tumor surgery, medical records were collected, the test results are divided into training group and validation group according to 7:3, divided into 2 groups according to the patient ICU stay time is greater than 24 hours, the prediction model after the ICU duration of head and neck tumor surgery after more than 24 hours. At the same time, clean the data, train the AI with the data, and compare the effectiveness of both sides with the ROC. After the establishment of prediction model and AI training, the patients included in the cohort were evaluated by prediction model and AI immediately after being transferred to the ICU, predicting the possibility of transferring out of the ICU within 24 hours.

Enrollment: 700 estimated patients

Sex: All

Ages: 18+ years old

Volunteers: No Healthy Volunteers

Inclusion criteria

1. Patients after head and neck tumors;
2. older than 18 years.

Exclusion criteria

1. Patients transferred to ICU twice after head and neck tumors;
2. Patients with unplanned transfer to the ICU.

Trial design

700 participants in 3 patient groups

training group

Description:

For training models

test group

Description:

Used to validate the model

Prospective cohort

Description:

For the prospective validation of the model