# Development and validation of identification models for aortic dissection and non-ST-segment elevation acute coronary syndrome in the emergency department

## TL;DR

This study develops and validates two models to quickly distinguish aortic dissection from non-ST-segment elevation acute coronary syndrome in emergency departments.

## Contribution

The novel contribution is the creation of two clinical prediction models with high accuracy for rapid differential diagnosis of two critical conditions.

## Key findings

- The whole-model achieved an AUC of 0.973 in training and 0.980 in validation sets.
- The whole-model demonstrated high sensitivity (0.930) and specificity (0.946) for accurate diagnosis.
- Both models are practical tools for emergency department use.

## Abstract

Aortic dissection (AD) and non-ST-segment elevation acute coronary syndrome (NSTE-ACS) are critical illnesses whose prompt identification within the emergency department is challenging. This study aimed to establish rapid discrimination models to differentiate between these conditions. Patients of the training set and validation set were collected from January 2020 to June 2023. All patients used their final diagnosis. Discriminant models were constructed via univariate and multivariate logistic regression analyses. Based on the results of the two models, two web calculators were developed. A total of 1314 patients were included in the study, with 997 patients (399 AD patients and 598 NSTE-ACS patients) and 317 patients (132 AD patients and 185 NSTE-ACS patients) in the training and validation sets, respectively. The semi-model consisted of six clinical characteristics (age, heart rate, pulse pressure, temperature, hypertension, diabetes), with an area under the ROC curve (AUC) of 0.792 and 0.823 in the training and validation sets. The whole-model included five clinical characteristics (age, pulse pressure, hypertension, diabetes) and two point-of-care test data (high sensitivity troponin I, D-dimer). It had a higher predictive value compared to the semi-model, with AUCs of 0.973 and 0.980 in the training and validation sets, respectively. Given the optimal cutoff point, the semi-model demonstrated a sensitivity of 0.716 and a specificity of 0.734, whereas the whole-model displayed a sensitivity of 0.930 and a specificity of 0.946. Both identification models can be used as reliable tools for rapidly identifying AD and NSTE-ACS.

The online version contains supplementary material available at 10.1038/s41598-025-31275-9.