What question did this study set out to answer?

This research aims to identify patients who may benefit from a paroxysmal nocturnal hemoglobinuria workup using machine learning techniques.

December 8, 2025BloodOpen Access

Early identification of patients likely to benefit from paroxysmal nocturnal hemoglobinuria workup using machine learning on large-scale real-world data.

View Full Paper

Ask AI

Bookmark

This research aims to identify patients who may benefit from a paroxysmal nocturnal hemoglobinuria workup using machine learning techniques.
Utilized Atropos Health Aracadia data asset across the U.S. for analysis.
Analyzed 416K patients with a hemoglobin (Hb) index date, monitoring them for 3 to 12 months post-workup.
Applied XGBoost model to predict likelihood of diagnosis, focusing on precision recall metrics.
Included clinical features from up to 24 months prior and employed a train/test split approach.
Identified 306 patients with paroxysmal nocturnal hemoglobinuria from the larger sample.
Achieved AUPRC and AUROC scores of 0.09 and 0.77 for the full model, and 0.08 and 0.77 for the model with 13 features.
Key predictive features were history of anemia, overall disease burden, age, and haptoglobin levels.
Physician reviews indicated a majority of 'false positives' warranted further diagnostic workup.

No takes yet. Share an insight, caveat, or question.

Machine learning predicts diagnosis likelihood for paroxysmal nocturnal hemoglobinuria, indicating potential treatment acceleration.

W. T. Pike (2025) studied this question.

View Full Paper

Ask AI

Bookmark