Artificial Intelligence Outperforms Doctors in Predicting Surgical Complications

Recent advancements in artificial intelligence (AI) have demonstrated a breakthrough in preoperative risk assessment, particularly by analyzing standard electrocardiogram (ECG) tests. Researchers at Johns Hopkins University developed AI models that can detect subtle signals in routine heart tests, which significantly enhance the prediction of postoperative complications such as heart attacks, strokes, or death within 30 days after surgery. This innovative approach outperforms traditional risk scores that are currently used by clinicians, which tend to be accurate only around 60% of the time.

The study involved analyzing preoperative ECG data from 37,000 patients who underwent surgery at Beth Israel Deaconess Medical Center. Two AI models were trained: one based solely on ECG data and a second, fusion model, combining ECG results with patient information like age, gender, and existing health conditions. While the ECG-only model already exceeded existing risk prediction methods in accuracy, the fusion model achieved an impressive 85% accuracy in predicting postoperative complications.

Senior author Robert D. Stevens emphasized that the ECG contains vast, often overlooked information about a patient’s cardiovascular health and general physiology. AI techniques, especially deep learning, enable extraction of this hidden data, revealing insights into inflammation, endocrine function, metabolism, and electrolyte balance—all factors influencing surgical outcomes.

This technological leap could redefine surgical risk evaluation, turning a quick, inexpensive, routine test into a powerful predictive tool. Patients and doctors can engage in better-informed discussions about surgical risks, potentially leading to improved decision-making and personalized care plans.

Future steps include validating the models on more extensive datasets and conducting prospective trials to assess their effectiveness in ongoing clinical settings. Researchers also aim to uncover additional predictive features from ECGs, further enriching the understanding of patient health through AI.

The findings, published in the British Journal of Anaesthesia, highlight the transformative potential of AI in healthcare, promising safer surgical procedures and better patient outcomes.

source: https://medicalxpress.com/news/2025-09-ai-complications-surgery-doctors.html