What is the clinical evidence from this study?

Study design: Observational. Population: Atrial Fibrillation (n=754949). Intervention: Attention-based deep neural network vs. CHA2DS2-VASc score. Primary outcome: Prediction of ischemic stroke occurrence (AUROC).

What does this research mean for the field?

An attention-based deep neural network predicts the occurrence of ischemic stroke in patients with atrial fibrillation with greater accuracy (AUROC 0.727) than the traditional CHA2DS2-VASc score (AUROC 0.651) and other machine learning methods. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

March 4, 2022Open Access

Predicting Ischemic Stroke in Patients with Atrial Fibrillation Using Machine Learning

Key Result

An attention-based deep neural network predicted the occurrence of ischemic stroke in patients with atrial fibrillation with a higher AUROC (0.727) compared to the CHA2DS2-VASc score (0.651).

Study Design

Type

Observational (n=754,949)

Structured PICO

Does an attention-based deep neural network improve the prediction of ischemic stroke in patients with atrial fibrillation compared to the CHA2DS2-VASc score?

Population

754,949 patients with atrial fibrillation (AF) from the Korean National Health Insurance (KNHIS) database between 2005 and 2013.

Intervention

Attention-based deep neural network using 65-dimensional features (demographics, health examination, and medical history) to predict ischemic stroke.

Comparator

CHA2DS2-VASc score and other machine learning methods (logistic regression, XGBoost, random forest).

Outcome

Occurrence of ischemic stroke within 5 years after the diagnosis of AF.hard clinical

An attention-based deep neural network using national health insurance data predicted ischemic stroke in patients with atrial fibrillation with higher accuracy than the traditional CHA2DS2-VASc score.

Main Result

Absolute Event Rate: 0.727% vs 0.651%

Limitations

Potential increase in type I error (false positives) because only ICD-10 codes were used without prescription information to identify patients.
Prediction results can vary significantly depending on the definition of the case target (e.g., 5-year period vs acute stroke).
The model requires improvement for practical application due to low precision (13.2%), meaning a high rate of false positive predictions.

Abstract

BACKGROUND: Atrial fibrillation (AF) is a well-known risk factor for stroke. Predicting the risk is important to prevent the first and secondary attacks of cerebrovascular diseases by determining early treatment. This study aimed to predict the ischemic stroke in AF patients based on the massive and complex Korean National Health Insurance (KNHIS) data through a machine learning approach. METHODS: We extracted 65-dimensional features, including demographics, health examination, and medical history information, of 754,949 patients with AF from KNHIS. Logistic regression was used to determine whether the extracted features had a statistically significant association with ischemic stroke occurrence. Then, we constructed the ischemic stroke prediction model using an attention-based deep neural network. The extracted features were used as input, and the occurrence of ischemic stroke after the diagnosis of AF was the output used to train the model. RESULTS: -value < 0.001). When the proposed deep learning model was applied to 150,989 AF patients, it was confirmed that the occurrence ischemic stroke was predicted to be higher AUROC (AUROC = 0.727 ± 0.003) compared to CHA2DS2-VASc score (AUROC = 0.651 ± 0.007) and other machine learning methods. CONCLUSIONS: As part of preventive medicine, this study could help AF patients prepare for ischemic stroke prevention based on predicted stoke associated features and risk scores.

Bookmark

View Full Paper