System and method for predicting and summarizing medical events from electronic health records

Patent metadata

Field	Value
Publication number	US-11935634-B2
Application number	US-201715690721-A
Country	US
Kind code	B2
Filing date	Aug 30, 2017
Priority date	Jul 28, 2017
Publication date	Mar 19, 2024
Grant date	Mar 19, 2024

Abstract

A system for predicting and summarizing medical events from electronic health records includes a computer memory storing aggregated electronic health records from a multitude of patients of diverse age, health conditions, and demographics including medications, laboratory values, diagnoses, vital signs, and medical notes. The aggregated electronic health records are converted into a single standardized data structure format and ordered arrangement per patient, e.g., into a chronological order. A computer (or computer system) executes one or more deep learning models trained on the aggregated health records to predict one or more future clinical events and summarize pertinent past medical events related to the predicted events on an input electronic health record of a patient having the standardized data structure format and ordered into a chronological order. An electronic device configured with a healthcare provider-facing interface displays the predicted one or more future clinical events and the pertinent past medical events of the patient.

First claim

We claim:

1. A system comprising, in combination,
a) a computer executing one or more deep learning models trained on aggregated health records converted into a single standardized data structure format and presented to the one or more deep learning models in a time sequence ordered arrangement per patient, the aggregated health records including a combination of at least free text medical notes, laboratory values, and medications; the one or more deep learning models predicting one or more future clinical events and identifying one or more pertinent past medical events related to the predicted one or more future clinical events for a patient based on an input electronic health record of the patient having the standardized data structure format, wherein the electronic health record of the patient comprises a plurality of individual health record elements, wherein the computer executing a particular model of the one or more deep learning models based on the input electronic health record comprises (i) determining an embedding vector for each bag of a plurality of bags of the individual elements of the input electronic health record, wherein determining an embedding vector for a particular bag of the plurality of bags comprises determining a first average embedding vector that averages embedding vectors representing free text medical notes of individual elements of the input electronic health record that are in the particular bag, determining a second average embedding vector that averages embedding vectors representing laboratory values of individual elements of the input electronic health record that are in the particular bag, determining a third average embedding vector that averages embedding vectors representing medications of individual elements of the input electronic health record that are in the particular bag, and concatenating the first, second, and third average vectors to generate the embedding vector for the particular bag, and (ii) applying each of the embedding vectors individually and sequentially to the particular model in chronological order ; wherein at least one of the one or more deep learning models includes an attention mechanism indicating how much attention the at least one of the one or more models gave to the individual elements in the electronic health record to predict the one or more future clinical events; and
b) a healthcare provider-facing interface of an electronic device for use by a healthcare provider treating the patient configured to display both (1) the predicted one or more future clinical events and (2) the identified one or more pertinent past medical events of the patient in the input electronic health record, including at least a free text medical note or excerpt thereof, whereby the display of the identified one or more pertinent past medical events of the patient in the input electronic health record together with the predicted one or more future clinical events facilitates the provider focusing on elements in the electronic health record which are relevant to the predicted one or more future clinical events.

2. The system of claim 1 , wherein the interface of the electronic device includes a display of:
(1) an alert to the one or more future clinical events,
(2) key medical problems or conditions related to the alert, and
(3) free text notes or excerpts thereof related to the alert.

3. The system of claim 2 , wherein the display of the free text notes or excerpts thereof are displayed in a manner indicating results from the application of the attention mechanism.

4. The system of claim 2 , wherein the display further comprises a display of at least one of inferred information from the patient electronic health record and a timeline of a probability or risk of certain events occurring in the future.

5. The system of claim 1 , wherein the display permits a user of the electronic device to select one of the predicted future clinical events and the selection triggers further display of information pertinent to the selected future clinical event.

6. The system of claim 5 , wherein the further display comprises display of medications prescribed to the patient and notes or excerpts thereof related to the selected future clinical event.

7. The system of claim 3 , wherein the display of the notes or excerpts thereof indicating results from the application of the attention mechanism comprises display of the notes or excerpts thereof using at least one of the following to provide highlighting or gradations of emphasis on particular words, phrases or other text in the notes: font size, font color, shading, bold, italics, underline, strikethough, blinking, highlighting with color, and font selection.

8. A method of assisting a health care provider in providing care for a patient, comprising the steps of:
a) using a predictive model trained on aggregated electronic health records converted into a single standardized data structure format and presented to the predictive model in a time sequence ordered arrangement per patient to generate, based on an input electronic health record of the patient having the standardized data structure format, (1) a prediction of a future clinical event for the patient and (2) identify pertinent past medical events from an input electronic health record for the patient which are related to the prediction, wherein the aggregated health records include a combination of at least free text medical notes, laboratory values, and medications, wherein the input electronic health record of the patient comprises a plurality of individual health record elements, and wherein using a particular model of the predictive models trained on the aggregated electronic heath records comprises (i) determining an embedding vector for each bag of a plurality of bags of the individual elements of the input electronic health record, wherein determining an embedding vector for a particular bag of the plurality of bags comprises determining a first average embedding vector that averages embedding vectors representing free text medical notes of individual elements of the input electronic health record that are in the particular bag, determining a second average embedding vector that averages embedding vectors representing laboratory values of individual elements of the input electronic health record that are in the particular bag, determining a third average embedding vector that averages embedding vectors representing medications of individual elements of the input electronic health record that are in the particular bag, and concatenating the first, second, and third average vectors to generate the embedding vector for the particular bag, and (ii) applying each of the embedding vectors individually and sequentially to the particular model in chronological order;
b) generate data related to both the prediction and the identified pertinent past medical events;
c) transmit the generated data to an electronic device used by the health care provider for display on the electronic device;
wherein:
the predictive model uses an attention mechanism to indicate how much attention the predictive model gave to elements in the input electronic health record to predict the future clinical event and identify pertinent past medical events and wherein the generated data includes the results of the attention mechanism; and
d) displaying both (1) the one or more future clinical events and (2) the identified pertinent past medical events on the electronic device including free text medical notes or excerpts thereof, whereby the display of the past medical events of facilitates the health care provider focusing on elements in the electronic health record which are relevant to the predicted one or more future clinical events.

9. The m

Assignees

• Google Llc

Inventors

• Mossin Alexander
• Rajkomar Alvin
• Oren Eyal
• Wilson James
• Wexler James
• Sundberg Patrik
• Dai Andrew
• Cui Yingwei
• Corrado Gregory
• Yee Hector
• Marcus Jacob
• Dean Jeffrey
• Irvine Benjamin
• Chen Kai
• Zhang Kun
• Hardt Michaela
• Sun Xiaomi
• Hajaj Nissan
• Liu Peter Junteng
• Le Quoc
• Liu Xiaobing
• Zhang Yi

Classifications

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title