Back

Finding Rare Disease Patients in EHR Databases via Lightly-Supervised Learning

Colbaugh, R.; Glass, K.

2020-07-07 health informatics
10.1101/2020.07.06.20147322 medRxiv
Show abstract

There is considerable interest in developing computational models capable of detecting rare disease patients in population-scale databases such as electronic health records (EHRs). Deriving these models is challenging for several reasons, perhaps the most daunting being the limited number of already-diagnosed, labeled patients from which to learn. We overcome this obstacle with a novel lightly-supervised algorithm that leverages unlabeled and/or unreliably-labeled patient data - which is typically plentiful - to facilitate model induction. Importantly, we prove the algorithm is safe: adding unlabeled/unreliably-labeled data to the learning procedure produces models which are usually more accurate, and guaranteed never to be less accurate, than models learned from reliably-labeled data alone. The proposed method is shown to substantially outperform state-of-the-art models in patient-finding experiments involving two different rare diseases and a country-scale EHR database. Additionally, we demonstrate feasibility of transforming high-performance models generated through light supervision into simpler models which, while still accurate, are readily-interpretable by non-experts.

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.