Doctor AI: Predicting Clinical Events via Recurrent Neural Networks

Edward Choi, Mohammad Taha Bahadori, Andy Schuetz, Walter F. Stewart, Jimeng Sun
; Proceedings of the 1st Machine Learning for Healthcare Conference, PMLR 56:301-318, 2016.

Abstract

Leveraging large historical data in electronic health record (EHR), we developed Doctor AI, a generic predictive model that covers observed medical conditions and medication uses. Doctor AI is a temporal model using recurrent neural networks (RNN) and was developed and applied to longitudinal time stamped EHR data from 260K patients and 2,128 physicians over 8 years. Encounter records (e.g. diagnosis codes, medication codes or procedure codes) were input to RNN to predict (all) the diagnosis and medication categories for a subsequent visit. Doctor AI assesses the history of patients to make multilabel predictions (one label for each diagnosis or medication category). Based on separate blind test set evaluation, Doctor AI can perform differential diagnosis with up to 79% recall@30, significantly higher than several baselines. Moreover, we demonstrate great generalizability of Doctor AI by adapting the resulting models from one institution to another without losing substantial accuracy.

Cite this Paper


BibTeX
@InProceedings{pmlr-v56-Choi16, title = {Doctor AI: Predicting Clinical Events via Recurrent Neural Networks}, author = {Edward Choi and Mohammad Taha Bahadori and Andy Schuetz and Walter F. Stewart and Jimeng Sun}, pages = {301--318}, year = {2016}, editor = {Finale Doshi-Velez and Jim Fackler and David Kale and Byron Wallace and Jenna Wiens}, 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/Choi16.pdf}, url = {http://proceedings.mlr.press/v56/Choi16.html}, abstract = {Leveraging large historical data in electronic health record (EHR), we developed Doctor AI, a generic predictive model that covers observed medical conditions and medication uses. Doctor AI is a temporal model using recurrent neural networks (RNN) and was developed and applied to longitudinal time stamped EHR data from 260K patients and 2,128 physicians over 8 years. Encounter records (e.g. diagnosis codes, medication codes or procedure codes) were input to RNN to predict (all) the diagnosis and medication categories for a subsequent visit. Doctor AI assesses the history of patients to make multilabel predictions (one label for each diagnosis or medication category). Based on separate blind test set evaluation, Doctor AI can perform differential diagnosis with up to 79% recall@30, significantly higher than several baselines. Moreover, we demonstrate great generalizability of Doctor AI by adapting the resulting models from one institution to another without losing substantial accuracy.} }
Endnote
%0 Conference Paper %T Doctor AI: Predicting Clinical Events via Recurrent Neural Networks %A Edward Choi %A Mohammad Taha Bahadori %A Andy Schuetz %A Walter F. Stewart %A Jimeng Sun %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-Choi16 %I PMLR %J Proceedings of Machine Learning Research %P 301--318 %U http://proceedings.mlr.press %V 56 %W PMLR %X Leveraging large historical data in electronic health record (EHR), we developed Doctor AI, a generic predictive model that covers observed medical conditions and medication uses. Doctor AI is a temporal model using recurrent neural networks (RNN) and was developed and applied to longitudinal time stamped EHR data from 260K patients and 2,128 physicians over 8 years. Encounter records (e.g. diagnosis codes, medication codes or procedure codes) were input to RNN to predict (all) the diagnosis and medication categories for a subsequent visit. Doctor AI assesses the history of patients to make multilabel predictions (one label for each diagnosis or medication category). Based on separate blind test set evaluation, Doctor AI can perform differential diagnosis with up to 79% recall@30, significantly higher than several baselines. Moreover, we demonstrate great generalizability of Doctor AI by adapting the resulting models from one institution to another without losing substantial accuracy.
RIS
TY - CPAPER TI - Doctor AI: Predicting Clinical Events via Recurrent Neural Networks AU - Edward Choi AU - Mohammad Taha Bahadori AU - Andy Schuetz AU - Walter F. Stewart AU - Jimeng Sun BT - Proceedings of the 1st Machine Learning for Healthcare Conference PY - 2016/12/10 DA - 2016/12/10 ED - Finale Doshi-Velez ED - Jim Fackler ED - David Kale ED - Byron Wallace ED - Jenna Wiens ID - pmlr-v56-Choi16 PB - PMLR SP - 301 DP - PMLR EP - 318 L1 - http://proceedings.mlr.press/v56/Choi16.pdf UR - http://proceedings.mlr.press/v56/Choi16.html AB - Leveraging large historical data in electronic health record (EHR), we developed Doctor AI, a generic predictive model that covers observed medical conditions and medication uses. Doctor AI is a temporal model using recurrent neural networks (RNN) and was developed and applied to longitudinal time stamped EHR data from 260K patients and 2,128 physicians over 8 years. Encounter records (e.g. diagnosis codes, medication codes or procedure codes) were input to RNN to predict (all) the diagnosis and medication categories for a subsequent visit. Doctor AI assesses the history of patients to make multilabel predictions (one label for each diagnosis or medication category). Based on separate blind test set evaluation, Doctor AI can perform differential diagnosis with up to 79% recall@30, significantly higher than several baselines. Moreover, we demonstrate great generalizability of Doctor AI by adapting the resulting models from one institution to another without losing substantial accuracy. ER -
APA
Choi, E., Bahadori, M.T., Schuetz, A., Stewart, W.F. & Sun, J.. (2016). Doctor AI: Predicting Clinical Events via Recurrent Neural Networks. Proceedings of the 1st Machine Learning for Healthcare Conference, in PMLR 56:301-318

Related Material