Clinical Tagging with Joint Probabilistic Models

Yoni Halpern, Steven Horng, David Sontag
Proceedings of the 1st Machine Learning for Healthcare Conference, PMLR 56:209-225, 2016.

Abstract

We describe a method for parameter estimation in bipartite probabilistic graphical models for joint prediction of clinical conditions from the electronic medical record. The method does not rely on the availability of gold-standard labels, but rather uses noisy labels, called anchors, for learning. We provide a likelihood-based objective and a moments-based initialization that are effective at learning the model parameters. The learned model is evaluated in a task of assigning a heldout clinical condition to patients based on retrospective analysis of the records, and outperforms baselines which do not account for the noisiness in the labels or do not model the conditions jointly.

Cite this Paper


BibTeX
@InProceedings{pmlr-v56-Halpern16, title = {Clinical Tagging with Joint Probabilistic Models}, author = {Halpern, Yoni and Horng, Steven and Sontag, David}, booktitle = {Proceedings of the 1st Machine Learning for Healthcare Conference}, pages = {209--225}, year = {2016}, editor = {Doshi-Velez, Finale and Fackler, Jim and Kale, David and Wallace, Byron and Wiens, Jenna}, volume = {56}, series = {Proceedings of Machine Learning Research}, address = {Northeastern University, Boston, MA, USA}, month = {18--19 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v56/Halpern16.pdf}, url = {https://proceedings.mlr.press/v56/Halpern16.html}, abstract = {We describe a method for parameter estimation in bipartite probabilistic graphical models for joint prediction of clinical conditions from the electronic medical record. The method does not rely on the availability of gold-standard labels, but rather uses noisy labels, called anchors, for learning. We provide a likelihood-based objective and a moments-based initialization that are effective at learning the model parameters. The learned model is evaluated in a task of assigning a heldout clinical condition to patients based on retrospective analysis of the records, and outperforms baselines which do not account for the noisiness in the labels or do not model the conditions jointly.} }
Endnote
%0 Conference Paper %T Clinical Tagging with Joint Probabilistic Models %A Yoni Halpern %A Steven Horng %A David Sontag %B Proceedings of the 1st Machine Learning for Healthcare Conference %C Proceedings of Machine Learning Research %D 2016 %E Finale Doshi-Velez %E Jim Fackler %E David Kale %E Byron Wallace %E Jenna Wiens %F pmlr-v56-Halpern16 %I PMLR %P 209--225 %U https://proceedings.mlr.press/v56/Halpern16.html %V 56 %X We describe a method for parameter estimation in bipartite probabilistic graphical models for joint prediction of clinical conditions from the electronic medical record. The method does not rely on the availability of gold-standard labels, but rather uses noisy labels, called anchors, for learning. We provide a likelihood-based objective and a moments-based initialization that are effective at learning the model parameters. The learned model is evaluated in a task of assigning a heldout clinical condition to patients based on retrospective analysis of the records, and outperforms baselines which do not account for the noisiness in the labels or do not model the conditions jointly.
RIS
TY - CPAPER TI - Clinical Tagging with Joint Probabilistic Models AU - Yoni Halpern AU - Steven Horng AU - David Sontag BT - Proceedings of the 1st Machine Learning for Healthcare Conference DA - 2016/12/10 ED - Finale Doshi-Velez ED - Jim Fackler ED - David Kale ED - Byron Wallace ED - Jenna Wiens ID - pmlr-v56-Halpern16 PB - PMLR DP - Proceedings of Machine Learning Research VL - 56 SP - 209 EP - 225 L1 - http://proceedings.mlr.press/v56/Halpern16.pdf UR - https://proceedings.mlr.press/v56/Halpern16.html AB - We describe a method for parameter estimation in bipartite probabilistic graphical models for joint prediction of clinical conditions from the electronic medical record. The method does not rely on the availability of gold-standard labels, but rather uses noisy labels, called anchors, for learning. We provide a likelihood-based objective and a moments-based initialization that are effective at learning the model parameters. The learned model is evaluated in a task of assigning a heldout clinical condition to patients based on retrospective analysis of the records, and outperforms baselines which do not account for the noisiness in the labels or do not model the conditions jointly. ER -
APA
Halpern, Y., Horng, S. & Sontag, D.. (2016). Clinical Tagging with Joint Probabilistic Models. Proceedings of the 1st Machine Learning for Healthcare Conference, in Proceedings of Machine Learning Research 56:209-225 Available from https://proceedings.mlr.press/v56/Halpern16.html.

Related Material