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

Patent metadata

Field	Value
Publication number	US-11398299-B2
Application number	US-201715690703-A
Country	US
Kind code	B2
Filing date	Aug 30, 2017
Priority date	Jul 28, 2017
Publication date	Jul 26, 2022
Grant date	Jul 26, 2022

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 for predicting one or more future clinical events of a patient and identifying pertinent past medical events in an electronic health record of the patient that are relevant to the predicted one or more future clinical events and facilitating a healthcare provider in understanding the predicted one or more future clinical events, comprising:
a) a computer memory storing aggregated electronic health records from a multitude of patients of diverse age, health conditions, and demographics including at least medications, laboratory values, diagnoses, vital signs, and free text medical notes, wherein the aggregated electronic health records are converted into a single standardized data structure format and in an ordered arrangement per patient;
b) a computer executing one or more deep learning models trained on the aggregated health records converted into the single standardized data structure format and in the ordered arrangement to predict one or more future clinical events and identify pertinent past medical events related to the predicted one or more future clinical events based on an input electronic health record of a patient having the standardized data structure format, the electronic health record including a multitude of individual elements of at least free text medical notes, laboratory values, and medications, wherein the computer executing a particular one of the one or more deep learning models to predict one or more future clinical events and identify pertinent past medical events related to the predicted one or more future clinical events comprises generating a plurality of embedding vectors that represent the contents of the input electronic health record corresponding to respective different periods of time, and wherein generating a particular one of the embedding vectors that corresponds to a respective particular period of time comprises: (1) generating an embedding vector for at least one laboratory value of the input electronic health record that corresponds to the particular period of time and generating a first average embedding vector by determining a weighted average of the embedding vectors determined for the laboratory values that correspond to the particular period of time, (2) generating an embedding vector for at least one medication of the input electronic health record that corresponds to the particular period of time and generating a second average embedding vector by determining a weighted average of the embedding vectors determined for the medications that correspond to the particular period of time, (3) generating at least one embedding vector for at least one free text medical note of the input electronic health record that corresponds to the particular period of time and generating a third average embedding vector by determining a weighted average of the embedding vectors determined for the free text medical notes that correspond to the particular period of time, and (4) generating the particular one of the embedding vectors by concatenating at least the first, second, and third average embedding vectors; and
c) an electronic device for use by the healthcare provider treating the patient configured with a healthcare provider-facing interface displaying concurrently both (1) the predicted one or more future clinical events and (2) the identified pertinent past medical events of the patient in the input electronic health record, whereby the concurrent 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 assists the healthcare provider understanding both the predicted one or more future clinical events and focusing their attention 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 aggregated health records comprise health records arranged in at least two different data formats.

3. The system of claim 1 , wherein the standardized data structure format comprises Fast Health Interoperability Resources (FHIR).

4. The system of claim 1 , wherein the one or more deep learning models each contain attention mechanisms indicating how attention the one or more models gave to elements in the electronic health record to predict the one or more future clinical events and summarize pertinent past medical events related to the predicted one or more future clinical events.

5. The system of claim 4 , wherein the one or more deep learning models comprise at least one Long-Short-Term Memory (LSTM) model, at least one time aware Feed-Forward Model (FFM), and at least one embedded boosted time-series model.

6. The system of claim 5 , wherein each of the deep learning models are trained on data sets making up the aggregated electronic health records separately.

7. The system of claim 5 , wherein predictions of one or more future clinical events and summarized pertinent past medical events related to the predicted one or more future clinical events are obtained from an ensemble average of the deep learning models.

8. The system of claim 1 , wherein the predicted one or more future clinical events include at least one of unplanned transfer to intensive care unit, length of stay in a hospital greater than 7 days, unplanned readmission within 30 days after discharge of the patient, inpatient mortality, primary diagnosis, a complete set of primary and secondary billing diagnoses, or atypical laboratory values.

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

10. The system of claim 9 , wherein the one or more deep learning models each contain an attention mechanism indicating how much attention the one or more models gave to elements in the electronic health record to predict the one or more future clinical events and identify pertinent past medical events related to the predicted one or more future clinical events, and wherein the display of the notes or excerpts thereof are displayed in a manner indicating results from the application of the attention mechanism.

11. A method for predicting and summarizing medical events from electronic health records, comprising:
a) aggregating electronic health records from a multitude of patients of diverse age, health conditions, and demographics, the electronic health records including at least medications, laboratory values, diagnoses, vital signs, and free text medical notes;
b) converting the aggregated electronic health records into a single standardized data structure format and ordered per patient into an ordered arrangement;
c) training one or more deep learning models on the aggregated health records converted into the single standardized data structure format and in the ordered arrangement;
d) using the trained one or more deep learning models to predict one or more future clinical events and identify pertinent past medical events related to the predicted one or more future clinical events from an input electronic health record of a patient having the standardized data structure format and ordered into a chronological order, the electronic health record including a multitude of individual elements of at least free text medical notes, laboratory values, and medications wherein using a particular one of the one or more deep learning models to predict one or more future clinical events and identify pertinent past medical events related to the predicted one or more future clinical ev

Assignees

• Google Llc

Inventors

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

Classifications

• G06N3/045

  Combinations of networks · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G06N3/0442

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

• G06N3/08

  Learning methods · CPC title

• G06N3/09

  Supervised learning · CPC title