Domain-transferred health-related predictive data analysis

Patent metadata

Field	Value
Publication number	US-2021383927-A1
Application number	US-202016895424-A
Country	US
Kind code	A1
Filing date	Jun 8, 2020
Priority date	Jun 8, 2020
Publication date	Dec 9, 2021
Grant date	—

Abstract

There is a need for more effective and efficient health-related predictive data analysis. This need can be addressed by, for example, solutions for performing domain-transferred health-related predictive data analysis. In one example, a method includes identifying an initial risk scoring model, generating a cross-domain mapping of the initial risk scoring model that maps initial risk categories of the initial risk scoring model to inferred risk categories, generating a weighted risk category value for each inferred risk category, generating a health-related risk prediction based on each weighted risk category value, and performing prediction-based actions based on the health-related risk prediction.

First claim

1 . A computer-implemented method for performing health-related predictive data analysis for a target individual with respect to a target condition, the computer-implemented method comprising:
identifying an initial risk scoring model, wherein the initial risk scoring model is associated with a plurality of initial risk categories; generating a cross-domain mapping of the initial risk scoring model, wherein: (i) the cross-domain mapping maps each initial risk category of the plurality of initial risk categories to an inferred risk category of a plurality of inferred risk categories, and (ii) each inferred risk category of the plurality of inferred risk categories is associated with one or more observed input variables for the target individual; for each inferred risk category of the plurality of inferred risk categories:
determining an inferred risk category value for the inferred risk category based on the one or more observed input variables for the inferred risk category,
determining a per-category weight value for the inferred risk category value, and
determining a weighted risk category value for the inferred risk category based on the inferred risk category value for the inferred risk category and the per-category weight value for the inferred risk category;
processing each weighted risk category value for an inferred risk category of the plurality of inferred risk categories using the initial risk scoring model and in accordance with the cross-domain mapping in order to generate a health-related risk prediction for the target individual with respect to the target condition; and performing one or more prediction-based actions based on the health-related risk prediction.

2 . The computer-implemented method of claim 1 , wherein:
the plurality of initial risk categories comprise a compliance history category, the plurality of inferred risk categories comprise a medical history category, and the cross-domain mapping maps the compliance history category to the medical history category.

3 . The computer-implemented method of claim 1 , wherein:
the plurality of initial risk categories comprise a record magnitude category, the plurality of inferred risk categories comprise a current phenotype category, and the cross-domain mapping maps the record magnitude category to the current phenotype category.

4 . The computer-implemented method of claim 1 , wherein:
the plurality of initial risk categories comprise a record history length category, the plurality of inferred risk categories comprise a target condition onset delay category, and the cross-domain mapping maps the record history length to the target condition onset delay category.

5 . The computer-implemented method of claim 1 , wherein:
the plurality of initial risk categories comprise a record diversity category, the plurality of inferred risk categories comprise a current therapeutic management category, and the cross-domain mapping maps the record diversity category to current therapeutic management category.

6 . The computer-implemented method of claim 1 , wherein:
the plurality of initial risk categories comprise a query frequency category, the plurality of inferred risk categories comprise a genetic variance category, and the cross-domain mapping maps the query frequency category to current genetic variance category.

7 . The computer-implemented method of claim 1 , wherein generating each inferred risk category value for an inferred risk category of the plurality of inferred risk categories comprises:
processing the one or more observed input variables associated with the inferred risk category using a trained machine learning model associated with the inferred risk category to generate the inferred risk category value.

8 . The computer-implemented method of claim 1 , wherein:
the initial risk scoring model defines an initial weight for each initial risk category of the plurality of initial risk categories, and each per-category weight value for an inferred risk category of the plurality of inferred risk categories is determined based on the initial weight value for the initial risk category that is mapped to the inferred risk category according to the cross-domain mapping.

9 . The computer-implemented method of claim 1 , wherein each per-category weight value for an inferred risk category of the plurality of inferred risk categories is determined in accordance with an optimization-based training technique and based on ground-truth health-related risk predictions for a group of training individual-condition pairs.

10 . The computer-implemented method of claim 1 , wherein the health-related risk prediction is updated in accordance with a Polygenic Risk Score (PRS) for the target individual with respect to the target condition.

11 . An apparatus for performing health-related predictive data analysis for a target individual with respect to a target condition, the apparatus comprising at least one processor and at least one memory including program code, the at least one memory and the program code configured to, with the processor, cause the apparatus to at least:
identify an initial risk scoring model, wherein the initial risk scoring model is associated with a plurality of initial risk categories; generate a cross-domain mapping of the initial risk scoring model, wherein: (i) the cross-domain mapping maps each initial risk category of the plurality of initial risk categories to an inferred risk category of a plurality of inferred risk categories, and (ii) each inferred risk category of the plurality of inferred risk categories is associated with one or more observed input variables for the target individual; for each inferred risk category of the plurality of inferred risk categories:
determine an inferred risk category value for the inferred risk category based on the one or more observed input variables for the inferred risk category,
determine a per-category weight value for the inferred risk category value, and
determine a weighted risk category value for the inferred risk category based on the inferred risk category value for the inferred risk category and the per-category weight value for the inferred risk category;
process each weighted risk category value for an inferred risk category of the plurality of inferred risk categories using the initial risk scoring model and in accordance with the cross-domain mapping in order to generate a health-related risk prediction for the target individual with respect to the target condition; and perform one or more prediction-based actions based on the health-related risk prediction.

12 . The apparatus of claim 11 , wherein:
the plurality of initial risk categories comprise a compliance history category, the plurality of inferred risk categories comprise a medical history category, and the cross-domain mapping maps the compliance history category to the medical history category.

13 . The apparatus of claim 11 , wherein:
the plurality of initial risk categories comprise a record magnitude category, the plurality of inferred risk categories comprise a current phenotype category, and the cross-domain mapping maps the record magnitude category to the current phenotype category.

14 . The apparatus of claim 11 , wherein:
the plurality of initial risk categories comprise a record history length category, the plurality of inferred risk categories comprise a target condition onset delay category, and the cross-domain mapping maps the record history length to the target condition onset delay category.

15 . The apparatus of claim 11 , wherein:
t

Assignees

• Optum Inc

Inventors

• Godden Paul J
• Omosaiye Olusola
• Mccandless Sarah
• Boss Gregory J

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N20/00

  Machine learning · CPC title

• G16H50/20

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

• G16H50/70

  for mining of medical data, e.g. analysing previous cases of other patients · CPC title