Patient predictive admission, discharge, and monitoring tool

Patent metadata

Field	Value
Publication number	US-11646105-B2
Application number	US-201816490815-A
Country	US
Kind code	B2
Filing date	Mar 2, 2018
Priority date	Mar 3, 2017
Publication date	May 9, 2023
Grant date	May 9, 2023

Abstract

Predictions can be provided for admission, discharge, pathway, and units for a patient during a stay at a healthcare facility. In some aspects, a computing device receives initial data regarding the patient and generates an admission or discharge prediction indicating a probability that the patient will be admitted to or discharged from the healthcare facility. The computing device segments portions of the initial data into segmented data and stores the segmented data in one or more of a plurality of segmentation categories. The computing device compares the segmented data to one or more patterns of a predictive model and, based on the comparing, generates a total length-of-stay (LOS) prediction indicating a duration of time for which the patient will stay at the healthcare facility.

First claim

What is claimed is:

1. A system for predicting patient admission and stay at a healthcare facility, the system comprising one or more processors;
a raw patient database;
a normalized and categorized patient database storing normalized and categorized data associated with each of a plurality of patients of the healthcare facility;
a prediction database; and
a memory in communication with the one or more processors, the memory storing instructions, wherein, when executed by the one or more processors, the instructions cause the one or more processors to:
receive initial patient data associated with a patient;
interpret, using natural language processing, the initial patient data;
store the interpreted initial patient data in the raw patient database;
extract, from the interpreted initial patient data, one or more conditions of the patient upon arrival to a first unit of the healthcare facility;
categorize the extracted one or more conditions;
store the categorized extracted one or more conditions as initial categorized patient data in the normalized and categorized patient database;
generate, from the initial categorized patient data, a first prediction comprising a predicted characteristic of a patient stay at the first unit, wherein the first prediction is generated based on a selected decision of a plurality of decisions output from one or more data mining models, the selected decision being obtained using a threshold determined from a Gaussian denormalization of historical data, a calculation of confidence intervals for each of the plurality of decisions, and one or more clinical inputs;
store the first prediction in the prediction database;
receive additional patient data based on the patient stay at the first unit;
interpret, using natural language processing, the additional patient data;
store the interpreted additional patient data in the raw patient database;
extract, from the interpreted additional patient data, one or more additional conditions of the patient upon discharge from the first unit;
categorize the extracted additional one or more conditions;
store the categorized extracted additional one or more conditions as additional categorized patient data in the normalized and categorized patient database;
segment portions of the initial categorized patient data and the additional categorized patient data into segmented data, wherein the segmented data comprises a plurality of time series, each of the plurality of time series comprises a plurality of indicators for a condition of the patient across a period of time;
build a second data mining model comprising a plurality of patterns based on the normalized and categorized data, each pattern defining a decision of a patient stay at a second unit of the healthcare facility as a function of one or more variables of the normalized and categorized data satisfying a given decision condition;
generate, as a function of the segmented data and the second data mining model, a second prediction comprising a predicted characteristic of the patient stay at the second unit; and
store the second prediction in the prediction database.

2. The system of claim 1 , wherein generating the second prediction comprises comparing the segmented data to the plurality of patterns of the second data mining model.

3. The system of claim 1 , wherein the characteristic of the patient stay at the first unit comprises an admit to the second unit prediction, a discharge prediction, a length-of-stay in the first unit prediction.

4. The system of claim 1 , wherein the characteristic of the patient stay at the second unit comprises a length-of-stay in the second unit prediction, and/or a final disposition prediction.

5. The system of claim 1 , wherein the condition of the patient comprises arrival location, clinical condition, and/or demographics data.

6. The system of claim 1 , wherein generating the second prediction comprises:
comparing the segmented data to patterns of a plurality of predictive models; and
combining results of the comparing with majority or weighted voting or statistical combination.

7. The system of claim 1 , wherein the instructions, when executed by one or more processors, cause the one or more processors to:
after receiving the initial data, receive temporal data comprising variables relating to the patient;
transform the temporal data into transformed data; and
store the transformed data in one or more of a plurality of pattern categories.

8. The system of claim 7 , wherein the transforming comprises:
converting a variable of the temporal data into a categorical variable;
changing a scale or orientation of the variable;
combining multiple variables of the temporal data; and
transforming a coordinate system of the temporal data.

9. A method for predicting patient admission and stay at a healthcare facility, the method comprising:
receiving, by a device comprising at least one processor and at least one memory in communication with the at least one processor, initial patient data;
interpreting, by the device and using natural language processing, the initial patient data;
storing, by the device, the interpreted initial patient data associated with a patient in a raw patient database;
extracting, by the device from the interpreted initial patient data, one or more conditions of the patient upon arrival to a first unit of the healthcare facility;
categorizing, by the device, the extracted one or more conditions;
storing, by the device, the categorized extracted one or more conditions as initial categorized patient data in a normalized and categorized patient database, wherein the normalized and categorized patient data stores normalized and categorized data associated with each of a plurality of patients of the healthcare facility;
generating, by the device and from the initial categorized patient data, a first prediction comprising a predicted characteristic of a patient stay at the first unit, wherein the first prediction is generated based on a selected decision of a plurality of decisions output from one ro more data mining models, the selected decision being obtained using a threshold determined from a Gaussian denormalization of historical data, a calculation of confidence intervals for each of the plurality of decisions, and one or more clinical inputs;
storing, by the device, the first prediction in a prediction database;
receiving, by the device, additional patient data based on the patient stay at the first unit;
interpreting, by the device and using natural language processing, the additional patient data:
storing, by the device, the interpreted additional patient data in the raw patient database;
extracting, by the device from the interpreted additional patient data, one or more conditions of the patient upon discharge from the first unit;
categorizing, by the device, the extracted additional one or more conditions;
storing, by the device, the categorized extracted additional one or more conditions as additional categorized patient data in a normalized and categorized patient database;
segmenting, by the device, portions of the initial categorized patient data and the additional categorized patient data into segmented data, wherein the segmented data comprises a plurality of time series, each of the plurality of time series comprises a plurality of indicators for a condition of the patient across a period of time;
building a second data mining model comprising a plurality of patterns based on the normalized and categorized data, each pattern defining a decision of a patient stay at a second unit of the healthcare facility as a function of one or more variables of the normalized and categorized data satisfying a given de

Assignees

• Univ Rush Medical Center

Inventors

• Rumoro Dino Peter

Classifications

• Y02A90/10

  Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation · CPC title

• G06F16/2365

  Ensuring data consistency and integrity · CPC title

• G16H40/63 Primary

  for local operation · CPC title

• G16H50/30

  for calculating health indices; for individual health risk assessment · CPC title

• G16H50/20

  for computer-aided diagnosis, e.g. based on medical expert systems · CPC title