An Artificial Intelligence Model for Intensive Care Length of Stay, Neurological Outcome and Costs Estimation After Cardiopulmonary Resuscitation: a Cohort Study.

N/ANot Yet RecruitingNCT07210866

Bezmialem Vakif University5,000 enrolled

Overview

The study aims to overview patients registered to Bezmialem Vakıf University Hospital Intensive Care Unit after successive cardiac arrest resuscitation from October 2010 to September 2025. The goal is to determine length of stay in reanimation, neurological clinical outcome and costs of these patients at discharge from the department. All these data is intended to be evaluated by artificial intelligence to evaluate a predictive model.

Study Type

OBSERVATIONAL

Enrollment

5,000

Conditions

Cost Analysis Length of ICU Stay Neurological Outcome

Interventions

no physical or medical interventionsOTHER

Data from patients after successive rescucitaion will be evaluated by machine learning programs.

Eligibility

Sex: ALLMin age: 18 Years

Medical Language ↔ Plain English

Inclusion Criteria: * age\>18 years * successive cardiopulmonary resuscitation * at least 1 hour long admission to ICU after Return Of Spontaneous Circulation (ROSC) Exclusion Criteria: * age \< 18 years * \>80% missing data in patient records * patients with no ROSC

Outcomes

Primary Outcomes

Machine Learning Python programme

The created database will be analyzed using a machine learning artificial intelligence algorithm with the Python programming language. After processing missing and incomplete data by artificial intelligence, the database will be divided into two parts: model training and model validation. Meaningful data will be selected through model training, and a prediction model will be built based on these data. To increase the interpretability of the prediction model and help users understand how and why certain predictions are made, the SHapley Additive exPlanations (SHAP) algorithm will be used. In machine learning, the SHAP technique is used to interpret the decision-making processes of complex machine learning models.

Time frame: 3 months