Back

Predicting Effectiveness of Antihypertensive Medications for Heart Failure based on Longitudinal Patient Records and Deep Learning

Chowdhury, S.; Chen, Y.; Ma, X.; Dai, Q.; Yu, Y.; Zong, N.

2022-11-14 health informatics
10.1101/2022.11.01.22281804 medRxiv
Show abstract

Drug treatment for heart failure (HF) condition includes different medications. As patients could respond variably to a particular medication, being able to predict drug effectiveness is crucial for personalized treatment. Laboratory tests in EHR summarize different aspects of the patients physiological process related to a diagnosis, where blood pressure (BP) is deemed a critical hemodynamic parameter for HF prognosis. This work first proposes a novel method based on combinations of different clinical end points to generate the positive and negative samples corresponding to HF patients on whom the drug is effective and not effective respectively. We then formulate drug effectiveness prediction as a time series classification problem and experiment with several deep learning models, leveraging the temporal BP laboratory measurements from EHR as the features. Over thorough comparative evaluations among 3 categories of HF medications and two types of lab features, we achieved the best F1 performance of [~]0.97.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.