The Predictability of the Necessity for Cardiology Consultation in Patients Scheduled for Non-Cardiac Surgery Using Artificial Intelligence Models in Preoperative Anesthesia Assessment

The Effectiveness of Using Artificial Intelligence (Chat GPT) in Cardiac Assessment During Anesthesia Examination of Preoperative Cases

Structured Summary Title

Predictability of Cardiology Consultation Requirement in Patients Undergoing Non-Cardiac Surgery Using Artificial Intelligence Models

Background

Preoperative cardiac risk assessment is essential for minimizing perioperative morbidity and mortality in patients undergoing non-cardiac surgery. Cardiology consultations are often requested to assess surgical eligibility and reduce complication risks. However, unnecessary consultations may contribute to inefficient healthcare resource utilization and procedural delays.

Recent advances in artificial intelligence, particularly large language models, have demonstrated potential in clinical decision support systems. The European Society of Cardiology (ESC) 2024 guidelines provide a structured framework for evaluating perioperative cardiac risk. This study aims to investigate whether AI-based models can assist in predicting the need for cardiology consultation and to examine the effect of prompted versus non-prompted input formats on AI recommendations.

Study Design

Prospective, observational, comparative study.

Ethical Approval

The study has been approved by the Bursa City Hospital Ethics Committee and will be conducted in accordance with the Declaration of Helsinki.

Sample Size

Sample size was calculated using G*Power software based on anticipated effect size and statistical power requirements.

Participants

Inclusion Criteria:

Adults aged 18 years or older

ASA physical status I-IV

Scheduled for non-cardiac surgery

Evaluated by anesthesia residents with less than two years of clinical experience

Exclusion Criteria:

Pediatric patients

Patients declining participation

Incomplete clinical data

Data Collection

The following patient data will be recorded:

Demographics (age, sex, BMI)

Medical history (comorbidities, medication use, allergies, substance use)

Functional capacity (METs score)

ECG findings

Chest radiography findings

Planned surgical procedure characteristics

AI Model Evaluation

Multiple AI language models will be tested using standardized patient scenarios. Each scenario will be presented in two formats:

Prompted format:

"You are a 10-year experienced anesthesiologist. According to ESC 2024 guidelines, evaluate whether this patient requires cardiology consultation."

Non-prompted format:

"Evaluate whether this patient requires cardiology consultation."

AI recommendations will not influence clinical decision-making.

Outcome Measures

Primary and secondary analyses will include:

Agreement between AI recommendations and expert anesthesiologist evaluations

Readability of AI-generated responses

Quality assessment of responses

Classification performance comparisons across models

Statistical Analysis

Statistical analyses will be performed using appropriate comparative and agreement tests. Readability and quality scores will be analyzed using non-parametric methods where applicable. ROC analysis will be used to assess classification ability. A significance level of p < 0.05 will be applied.

Study Objective

The objective of this study is to explore the feasibility of AI-assisted decision support systems in predicting cardiology consultation requirements and to evaluate whether prompt engineering influences AI performance.

Study Overview

• Status: Active, not recruiting
• Conditions: USE OF ARTIFICIAL INTELLIGENCE IN ANESTHESIA; PREOPERATIVE CARDIOLOGY CONSULTATION REQUIREMENT
• Intervention / Treatment: Other: Patient scenarios were presented to different AI models (ChatGPT 4.5, ChatGPT 5, Copilot, Deepseek, Grok, Claude, Gemini Flash, Gemini Pro) with and without prompts.

Detailed Description

Structured Summary Title

Predictability of Cardiology Consultation Requirement in Patients Undergoing Non-Cardiac Surgery Using Artificial Intelligence Models

Background

Preoperative cardiac risk assessment is essential for minimizing perioperative morbidity and mortality in patients undergoing non-cardiac surgery. Cardiology consultations are often requested to assess surgical eligibility and reduce complication risks. However, unnecessary consultations may contribute to inefficient healthcare resource utilization and procedural delays.

Recent advances in artificial intelligence, particularly large language models, have demonstrated potential in clinical decision support systems. The European Society of Cardiology (ESC) 2024 guidelines provide a structured framework for evaluating perioperative cardiac risk. This study aims to investigate whether AI-based models can assist in predicting the need for cardiology consultation and to examine the effect of prompted versus non-prompted input formats on AI recommendations.

Study Design

Prospective, observational, comparative study.

Ethical Approval

The study has been approved by the Bursa City Hospital Ethics Committee and will be conducted in accordance with the Declaration of Helsinki.

Sample Size

Sample size was calculated using G*Power software based on anticipated effect size and statistical power requirements.

Participants

Inclusion Criteria:

Adults aged 18 years or older

ASA physical status I-IV

Scheduled for non-cardiac surgery

Evaluated by anesthesia residents with less than two years of clinical experience

Exclusion Criteria:

Pediatric patients

Patients declining participation

Incomplete clinical data

Data Collection

The following patient data will be recorded:

Demographics (age, sex, BMI)

Medical history (comorbidities, medication use, allergies, substance use)

Functional capacity (METs score)

ECG findings

Chest radiography findings

Planned surgical procedure characteristics

AI Model Evaluation

Multiple AI language models will be tested using standardized patient scenarios. Each scenario will be presented in two formats:

Prompted format:

"You are a 10-year experienced anesthesiologist. According to ESC 2024 guidelines, evaluate whether this patient requires cardiology consultation."

Non-prompted format:

"Evaluate whether this patient requires cardiology consultation."

AI recommendations will not influence clinical decision-making.

Outcome Measures

Primary and secondary analyses will include:

Agreement between AI recommendations and expert anesthesiologist evaluations

Readability of AI-generated responses

Quality assessment of responses

Classification performance comparisons across models

Statistical Analysis

Statistical analyses will be performed using appropriate comparative and agreement tests. Readability and quality scores will be analyzed using non-parametric methods where applicable. ROC analysis will be used to assess classification ability. A significance level of p < 0.05 will be applied.

Study Objective

The objective of this study is to explore the feasibility of AI-assisted decision support systems in predicting cardiology consultation requirements and to evaluate whether prompt engineering influences AI performance.

• Study Type: Observational
• Enrollment (Estimated): 183

Contacts and Locations

Study Locations

• Turkey (Türkiye)
  • Bursa
    • Bursa, Bursa, Turkey (Türkiye), 16001
      • Bursa Şehir Hastanesi

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Probability Sample

Study Population

This study will be conducted with the participation of patients who will be referred to cardiology following an examination at the anesthesia clinic prior to undergoing non-cardiac surgery.

Description

Inclusion Criteria:

Adults aged 18 years or older

ASA physical status classification I-IV

Scheduled for non-cardiac surgery

Patients evaluated preoperatively by anesthesia residents with less than two years of clinical experience

Availability of complete clinical data including medical history, ECG findings, and chest radiography

Ability to provide informed consent

Exclusion Criteria:

Patients younger than 18 years of age

Patients undergoing cardiac surgery

Patients with incomplete clinical data

Patients who declined participation

Emergency surgery cases

Patients unable to undergo standard preoperative evaluation

Study Plan

How is the study designed?

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Agreement Between AI Model Recommendations and Expert Anesthesiologist Decision Regarding Cardiology Consultation Requirement	The level of agreement between artificial intelligence model recommendations and expert anesthesiologist evaluations for cardiology consultation necessity will be assessed using Cohen's Kappa coefficient based on ESC 2024 guidelines.	At baseline preoperative evaluation (Day 1)

Other Outcome Measures

Outcome Measure	Measure Description	Time Frame
Readability of AI-Generated Responses	The readability of AI-generated responses will be assessed using the Ateşman Readability Index to determine clarity and comprehensibility of consultation recommendations.	Immediately after AI-generated response evaluation (Day 1)

Collaborators and Investigators

• Sponsor: Bursa City Hospital
• Investigators: Principal Investigator: eralp çevikkalp, Bursa Şehir Hastanesi

Publications and helpful links

General Publications

• 1.M. Graeßner et al., "Enabling personalized perioperative risk prediction by using a machinelearning model based on preoperative data," Scientific Reports, vol. 13, no. 1, May 2023, doi: 10.1038/s41598-023-33981-8. 2.B. Choi et al., "Prediction Model for 30-Day Mortality after Non-Cardiac Surgery Using MachineLearning Techniques Based on Preoperative Evaluation of Electronic Medical Records," Journal of Clinical Medicine, vol. 11, no. 21, p. 6487, Nov. 2022, doi: 10.3390/jcm11216487. 3.M. Vine et al., "Innovative approaches to preoperative care including feasibility, efficacy, and ethical implications: a narrative review," AME Surgical Journal, vol. 4. AME Publishing Company, p. 1, Feb. 01, 2024. doi: 10.21037/asj-23-41. 4.P. Chung, C. T. Fong, A. M. Walters, N. Aghaeepour, M. Yetişgen, and V. N. O'Reilly-Shah, "Large Language Model Capabilities in Perioperative Risk Prediction and Prognostication," JAMA Surgery, vol. 159, no. 8, American Medical Association, p. 928, Jun. 05, 2024. doi: 10.1001/jamasurg.2024.1621 5.T. Yurttas, R. Hidvegi, and M. Filipovic, "Biomarker-Based Preoperative Risk Stratification for Patients Undergoing Non-Cardiac Surgery," Journal of Clinical Medicine, vol. 9, no. 2, p. 351, Jan. 2020, doi: 10.3390/jcm9020351 6.J. Stones and D. Yates, "Clinical risk assessment tools in anaesthesia," BJA Education, vol. 19, no. 2. Elsevier BV, p. 47, Dec. 15, 2018. doi: 10.1016/j.bjae.2018.09.009. 7. Engineering as an Important Emerging Skill for Medical Professionals: Tutorial. J Med Internet Res. 2023 Oct 4;25:e50638. doi: 10.2196/50638. PMID: 37792434; PMCID: PMC10585440. 8. ATEŞMAN, Ender. (1997). Türkçe'de okunabilirliğin Ölçülmesi. A.Ü. Tömer Dil Dergisi, sayı:58,s.171174. 9. Coskun B, Ocakoglu G, Yetemen M, Kaygisiz O. Can ChatGPT, an Artificial Intelligence Language Model, Provide Accurate and High-quality Patient Information on Prostate Cancer? Urology. 2023 Oct;180:35-58. doi: 10.1016/j.urology.2023.05.040. Epub 2023 Jul 4. PMID: 37406

Study record dates

Study Major Dates

• Study Start (Actual): January 1, 2025
• Primary Completion (Actual): June 30, 2025
• Study Completion (Estimated): March 15, 2026

Study Registration Dates

• First Submitted: December 20, 2025
• First Submitted That Met QC Criteria: January 31, 2026
• First Posted (Actual): February 9, 2026

Study Record Updates

• Last Update Posted (Actual): February 9, 2026
• Last Update Submitted That Met QC Criteria: January 31, 2026
• Last Verified: January 1, 2026

More Information

Terms related to this study

• Keywords: THE ROLE OF ARTIFICIAL INTELLIGENCE IN ANESTHESIA; ARTIFICIAL INTELLIGENCE IN CONSULTATION NEEDS
• Other Study ID Numbers: Bursa Şehir Hastanesi

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO
• IPD Plan Description: Due to our commitment to patient rights, data privacy, and the consent form

Drug and device information, study documents

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