# Development and Technical Validation of an Integrated Risk Calculator for Acute Coronary Syndrome Using ChatGPT-Assisted Coding

## TL;DR

This paper introduces a digital tool that combines three risk models for heart attacks into one interface, using AI to speed up development and improve accuracy.

## Contribution

The novel contribution is an integrated calculator for TIMI, GRACE, and Mehran scores using AI-assisted coding, validated against clinical registries.

## Key findings

- The calculator showed perfect agreement with TIMI and GRACE scores from clinical registries.
- Manually reviewed Mehran scores were consistent with the tool's output.
- The tool reduced data entry duplication and maintained accuracy.

## Abstract

Introduction

Acute coronary syndrome (ACS) remains a major contributor to cardiovascular morbidity and mortality, necessitating efficient and accurate risk-stratification tools. Widely used models such as the Thrombolysis in Myocardial Infarction (TIMI), Global Registry of Acute Coronary Events (GRACE), and Mehran scores require separate manual inputs, leading to workflow inefficiencies and transcription errors. This study introduces an integrated digital tool that consolidates these scoring systems into a single interface and explores the feasibility of AI-assisted development in clinical software design.

Objective

The objective of this study is to develop and technically validate an integrated ACS risk calculator that streamlines the computation of TIMI, GRACE, and Mehran scores while demonstrating the utility of AI-assisted coding in clinical applications.

Methods

A web-based calculator was developed using Hypertext Markup Language (HTML)/JavaScript, with AI-assisted code prototyping via ChatGPT (o3-mini-high) (OpenAI, San Francisco, CA). The interface standardizes shared clinical variables for all three risk scores. Validation was conducted using 226 ACS cases from the Sheikh Khalifa Specialty Hospital registry. TIMI and GRACE scores generated by the tool were compared against Get With The Guidelines-Coronary Artery Disease (GWTG-CAD) registry values using Pearson correlation. The Mehran score was internally validated through manual review. Congestive heart failure was inferred using Killip class > I to align inputs across models.

Results

The tool showed complete agreement with registry-based TIMI and GRACE scores (TIMI: r = 1.000, p < 0.001; GRACE: r = 1.000, p < 0.001). Manually reviewed Mehran scores demonstrated consistent output. The calculator reduced data entry duplication and preserved computational accuracy.

Conclusion

This study validates an integrated ACS risk calculator that unifies three established models within a single digital tool. It enhances workflow efficiency and demonstrates the practical value of AI-assisted, clinician-led development in cardiovascular decision support. Further clinical and usability validation is warranted.