Computer network architecture with machine learning and artificial intelligence and automated insight generation

• US 10,643,749 B1
• Filed: 09/30/2019
• Issued: 05/05/2020
• Est. Priority Date: 09/30/2019
• Status: Active Grant

• Current Assignee: Clarify Health Solutions, Inc.
• Original Assignee: Clarify Health Solutions, Inc.
• Inventors: Warner, Justin, Drouin, Jean P., Gottula, Todd, Sun, Emmet
• Primary Examiner(s): Nilsson, Eric
• Application Number:
  US16/587,780
• Time in Patent Office: 218 Days
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/041 Abduction
• CPC Class Codes: G06Q 40/08 Insurance
• CPC Class Codes: G16H 40/20 for the management or administration of healthcare resources or facilities, e.g. managing hospital staff or surgery rooms
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: G16H 50/30 for calculating health indices; for individual health risk assessment
• CPC Class Codes: H04L 67/54 Presence management, e.g. monitoring or registration for receipt of user log-on information, or the connection status of the users

Abstract

Embodiments in the present disclosure relate generally to computer network architectures for machine learning, artificial intelligence, and automated insight generation. Embodiments of computer network architecture automatically identify, measure, and generate insight reports of underperformance and over performance in healthcare practices. Embodiments may generate the insight reports of performance either occasionally on demand, or periodically, or as triggered by events such as an update of available data. Embodiments may include a combination of system databases with data provided by system users, and third-party databases to generate the insight reports, including social media data, financial data, socio-economic data, medical data, search engine data, e-commerce site data, and other databases.

Claims

1. A computer network architecture with artificial intelligence

and machine learning, comprising:

a prediction module with a prediction generator and an updated database,

a learning module with a training submodule, in electronic communication with the prediction module and the updated database, and

an automated insight generation web application (AIGWA) in electronic communication with both the prediction module, and a user device,

wherein, the AIGWA is configured to execute the steps of;

a. log a user into an Automated Insight Generation Web Application (AIGWA) embodiment,

b. receive a request from the user that automated insight reports be generated and transmitted, for specified topics and requested subgroups of data,

c. log out the user from the AIGWA,

d. the AIGWA selects a metric to measure the performance insight requested,

e. the AIGWA selects benchmarking data for the selected metrics, which may be either (1) available empirical data averages from sources available to the AIGWA, or (2) the forecasted outcomes for the selected metrics from the system'"'"'s patient risk scoring web application,

f. the AIGWA selects an estimate of error for each metric,

g. the AIGWA searches the data available to the AIGWA in the system for combinations of metrics for subgroups of data, and computes statistics summarizing each subgroup,

h. the AIGWA automatically generates a list of the subgroups where the selected metrics differ from the benchmarking data by a statistically significant amount of underperformance or over performance,

i. the AIGWA filters and ranks the insights by impact on the outcomes and selected metrics, and

j. the AIGWA automatically generates an insight report of the underperformance and over performance, and their metrics, and transmits the report to the user, occasionally as requested, or periodically, or when triggered by an event,

wherein, the learning module is configured to;

receive a list of algorithm definitions and datasets for patient risk scoring,

automatically calibrate one or more defined algorithms with the database,

test the calibrated algorithms with a plurality of evaluation metrics,

store the calibrated algorithms and evaluation metrics in a library,

automatically select an algorithm for patient risk scoring based on the evaluation metrics,

update further the database with third party data, and with user episode data, and

re-execute the calibrate, test, store, and select steps after the update of the database step,

wherein, the prediction generator is configured to;

receive a user prediction request for patient risk scoring, including episode data and a client model,

run the currently selected algorithm corresponding to the user of the episode data, and generate patient risk scoring prediction output,

generate a patient risk scoring prediction report based on the algorithm output, and

transmit the patient risk scoring prediction report to the user,

wherein, the algorithm definitions are of types that are members of the group comprising;

multi-level models,

random forest regression,

logistical regression,

gamma-distributed regression, and

linear regression;

the third-party data is from a party that is a member of the group comprising;

hospitals,

medical practices,

insurance companies,

credit reporting agencies, and

credit rating agencies;

the database includes patient medical data, patient personal data, patient outcome data, and medical treatment data;

the episode data includes individual patient medical data and personal data; and

the user is a member of the group comprising;

hospitals,

medical practices, and

insurance companies,

wherein the user device is remote from the prediction module, and

the user device is a member of the group comprising;

a computer,

a desktop PC,

a laptop PC,

a smart phone,

a tablet computer, and

a personal wearable computing device,

wherein the web application communicates with the user device by the Internet, or an extranet, or a VPN, or other network, and the web application is generic for any user, or customized for a specific user, or class of user,

wherein, the user prediction request requests calibration of the correlation of demographic, social and medical attributes of the patient, to the outcome of a specific patient clinical episode type, and

wherein the updated database includes data from at least one third party, containing data of one or more types from the group consisting of;

medical claims data, prescription refill data, publicly available social media data, socio-economic data, credit agency data, marketing data, travel website data, e-commerce website data, search engine data, credit card data, credit score and credit history data, lending data, mortgage data, financial data, travel data, geolocation data, telecommunications usage data, and other third-party databases.

2. A computer network architecture with a web application for artificial intelligence and machine learning and automated insight generation comprising:

an automated insight generation web application (AIGWA) in electronic communication with a prediction module, a patient risk scoring web application, and a user device, and

wherein, the AIGWA is configured to;

a. log a user into an Automated Insight Generation Web Application (AIGWA) embodiment,

b. receive a request from the user that automated insight reports be generated and transmitted, for specified topics and requested subgroups of data,

c. log the user out from the AIGWA,

d. the AIGWA selects a metric to measure the performance insight requested,

e. the AIGWA selects benchmarking data for the selected metrics, which may be either (1) available empirical data averages from sources available to the AIGWA, or (2) the forecasted outcomes for the selected metrics from the system'"'"'s patient risk scoring web application,

f. the AIGWA selects an estimate of error for each metric,

g. the AIGWA searches the data available to the AIGWA in the system for combinations of metrics for subgroups of data, and computes statistics summarizing each subgroup,

h. the AIGWA automatically generates a list of the subgroups where the selected metrics differ from the benchmarking data by a statistically significant amount of underperformance or over performance,

i. the AIGWA filters and ranks the insights by impact on the outcomes and selected metrics, and

j. the AIGWA automatically generates an insight report of the underperformance and over performance, and their metrics, and transmits the report to the user, occasionally as requested, or periodically, or when triggered by an event,

wherein, the learning module is configured to;

receive a list of algorithm definitions and datasets for patient risk scoring,

automatically calibrate one or more defined algorithms with the database,

test the calibrated algorithms with a plurality of evaluation metrics,

store the calibrated algorithms and evaluation metrics in a library,

automatically select an algorithm for patient risk scoring based on the evaluation metrics,

update further the database with third party data, and with user episode data, and

re-execute the calibrate, test, store, and select steps after the update of the database step,

wherein, the prediction generator is configured to;

receive a user prediction request for patient risk scoring, including episode data and a client model,

run the currently selected algorithm corresponding to the user of the episode data, and generate patient risk scoring prediction output,

generate a patient risk scoring prediction report based on the algorithm output, and

transmit the patient risk scoring prediction report to the user,

wherein, the algorithm definitions are of types that are members of the group comprising;

multi-level models,

random forest regression,

logistical regression,

gamma-distributed regression, and

linear regression;

the third-party data is from a party that is a member of the group comprising;

hospitals,

medical practices,

insurance companies,

credit reporting agencies, and

credit rating agencies;

the database includes patient medical data, patient personal data, patient outcome data, and medical treatment data;

the episode data includes individual patient medical data and personal data; and

the user is a member of the group comprising;

hospitals,

medical practices, and

insurance companies,

wherein, the user device is remote from the prediction module, and

the user device is a member of the group comprising;

a computer,

a desktop PC,

a laptop PC,

a smart phone,

a tablet computer, and

a personal wearable computing device,

wherein the web application communicates with the user device by the Internet, or an extranet, or a VPN, or other network, and the web application is generic for any user, or customized for a specific user, or class of user, and

wherein, the user prediction request requests calibration of the correlation of demographic, social and medical attributes of the patient, to the outcome of a specific patient clinical episode type.

3. The computer architecture of claim 2, wherein the updated database includes data from at least one third party, containing data of one or more types from the group consisting of:

medical claims data, prescription refill data, publicly available social media data, socio-economic data, credit agency data, marketing data, travel web site data, e-commerce web site data, search engine data, credit card data, credit score and credit history data, lending data, mortgage data, financial data, travel data, geolocation data, telecommunications usage data, and other third-party databases.

4. A computer network architecture with an automated insight generation web application with artificial intelligence and machine learning, comprising:

an automated insight generation web application (AIGWA) in electronic communication with a prediction module, a patient risk scoring web application and a user device, and

wherein, the AIGWA web application is configured to;

a. log a user into an Automated Insight Generation Web Application (AIGWA) embodiment,

b. receive a request from the user that automated insight reports be generated and transmitted, for specified topics and requested subgroups of data,

c. log the user out from the AIGWA,

d. the AIGWA selects a metric to measure the performance insight requested,

e. the AIGWA selects benchmarking data for the selected metrics, which may be either (1) available empirical data averages from sources available to the AIGWA, or (2) the forecasted outcomes for the selected metrics from the system'"'"'s patient risk scoring web application,

f. the AIGWA selects an estimate of error for each metric,

g. the AIGWA searches the data available to the AIGWA in the system for combinations of metrics for subgroups of data, and computes statistics summarizing each subgroup,

h. the AIGWA automatically generates a list of the subgroups where the selected metrics differ from the benchmarking data by a statistically significant amount of underperformance or over performance,

i. the AIGWA filters and ranks the insights by impact on the outcomes and selected metrics, and

j. the AIGWA automatically generates an insight report of the underperformance and over performance, and their metrics, and transmits the report to the user, occasionally as requested, or periodically, or when triggered by an event.

5. The computer network architecture of claim 4:

wherein, the learning module is configured to;

receive a list of algorithm definitions and datasets for patient risk scoring,

automatically calibrate one or more defined algorithms with the database,

test the calibrated algorithms with a plurality of evaluation metrics,

store the calibrated algorithms and evaluation metrics in a library,

automatically select an algorithm for patient risk scoring based on the evaluation metrics,

update further the database with third party data, and with user episode data, and

re-execute the calibrate, test, store, and select steps after the update of the database step.

6. The computer network architecture of claim 4, wherein:

the prediction generator is configured to;

receive a user prediction request for patient risk scoring, including episode data and a client model,

run the currently selected algorithm corresponding to the user of the episode data, and generate patient risk scoring prediction output,

generate a patient risk scoring prediction report based on the algorithm output, and

transmit the patient risk scoring prediction report to the user.

7. The computer network architecture of claim 4, wherein:

the algorithm definitions are of types that are members of the group comprising;

multi-level models,

random forest regression,

logistical regression,

gamma-distributed regression, and

linear regression;

the third-party data is from a party that is a member of the group comprising;

hospitals,

medical practices,

insurance companies,

credit reporting agencies, and

credit rating agencies;

the database includes patient medical data, patient personal data, patient outcome data, and medical treatment data;

the episode data includes individual patient medical data and personal data; and

the user is a member of the group comprising;

hospitals,

medical practices, and

insurance companies.

8. The computer network architecture of claim 4, wherein the user device is remote from the prediction module, and

the user device is a member of the group comprising:

a computer,

a desktop PC,

a laptop PC,

a smart phone,

a tablet computer, and

a personal wearable computing device.

9. The computer network architecture of claim 4, wherein the web application communicates with the user device by the Internet, or an extranet, or a VPN, or other network, and the web application is generic for any user, or customized for a specific user, or class of user, and

wherein, the user prediction request requests calibration of the correlation of demographic, social and medical attributes of the patient, to the outcome of a specific patient clinical episode type.

10. The computer architecture of claim 4, wherein the updated database includes data from at least one third party, containing data of one or more types from the group consisting of:

medical claims data, prescription refill data, publicly available social media data, socio-economic data, credit agency data, marketing data, travel web site data, e-commerce web site data, search engine data, credit card data, credit score and credit history data, lending data, mortgage data, financial data, travel data, geolocation data, telecommunications usage data, and other third-party databases.