PERSONALIZED LAB TEST RESPONSE PREDICTION WITH KNOWLEDGE AUGMENTATION
Personalized medical systems are rapidly gaining traction as opposed to “one size fits all” systems. The ability to predict patients’ lab test responses and provide justification for the predictions would serve as an important decision support tool and help clinicians tailor treatment regimes for patients. This requires one to model the complex interactions among different medications, diseases, and lab tests. We also need to learn a strong patient representation, capturing both the sequential information accumulated over the visits and information from other similar patients. Further, we model the drug-lab interactions and diagnosis-lab interactions in the form of graphs and design a knowledge-augmented approach to predict patients’ response to a target lab result. We also take into consideration patients' past lab responses to personalize the prediction. Experiments on the benchmark MIMIC-III and a real-world outpatient dataset demonstrate the effectiveness of the proposed solution in reducing prediction errors by a significant margin. Case studies show that the identified top factors for influencing the predicted lab results are consistent with the clinicians' understanding.
PDF Abstract
MIMIC-III
