Identifying group and individual-level risk factors via risk-driven patient stratification

• US 9,996,889 B2
• Filed: 10/01/2012
• Issued: 06/12/2018
• Est. Priority Date: 10/01/2012
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Gotz, David H., Hsueh, Pei-Yun S., Hu, Jianying, Sun, Jimeng
• Primary Examiner(s): NGUYEN, HIEP VAN
• Application Number:
  US13/632,659
• Publication Number:
  US 20140095184A1
• Time in Patent Office: 2,080 Days
• Field of Search: 705 2
• CPC Class Codes: G06Q 10/0635 Risk analysis of enterprise or organisation activities
• CPC Class Codes: G06Q 10/0637 Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
• CPC Class Codes: G16H 50/30 for calculating health indices; for individual health risk assessment
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients

Abstract

Systems and methods for individual risk factor identification include identifying common risk factors for one or more risk targets from population data. Individuals are stratified into clusters based upon the common risk factors. A discriminability of each of the common risk factors is determined, using a processor, for a target cluster using individual data of the target cluster to provide re-ranked common risk factors as individual risk factors for the target cluster, such that the discriminability is a measure of how a risk factor discriminates its cluster from other clusters.

Claims

1. A method for individual risk factor identification, comprising:

identifying common risk factors for one or more risk targets from population data;

stratifying individuals into clusters based upon the common risk factors, wherein a distance between each cluster represents a similarity based upon the common risk factors, and wherein a closest pair of clusters is iteratively merged into a single cluster until a threshold condition is reached;

determining, using a processor, a discriminability of each of the common risk factors for a target cluster using individual data of the target cluster to provide re-ranked common risk factors as individual risk factors for the target cluster, such that the discriminability is a quantitative measure of how a risk factor discriminates its cluster from other clusters; and

customizing and performing, using a personalized user interface and dashboard display operatively coupled to at least one hardware based care-management machine, a personalized hardware-based care management process by controlling the at least one hardware based care-management machine for personalized treatment of a particular individual, the personalized user interface and dashboard display comprising a customized graphical user interface (GUI) configured for customizing healthcare plans, tailored to the particular individual based on the individual risk factors determined by the identifying common risk factors, the stratifying individuals into clusters and the re-ranked common risk factors, and for providing real-time clinical decision support at a point-of-care for the particular individual.

2. The method as recited in claim 1, further comprising identifying the clusters as one of a plurality of risk levels.

3. The method as recited in claim 2, wherein the plurality of risk levels include at least one low-risk cluster and at least one high-risk cluster.

4. The method as recited in claim 3, wherein determining the discriminability for each risk factor for the target cluster includes comparing risk factors for the target cluster with risk factors for at least one of:

other high-risk clusters, low-risk clusters, and a general population.

5. The method as recited in claim 2, wherein identifying the clusters includes identifying the clusters as one of a plurality of risk levels based upon a proportion of at-risk individuals in each cluster.

6. The method as recited in claim 5, wherein at-risk individuals are identified based upon a risk score.

7. The method as recited in claim 6, wherein the risk score is assigned using a classifier.

8. The method as recited in claim 1, wherein determining the discriminability includes determining an amount each risk factor contributes to training a classifier for a particular comparison configuration while accounting for a bias introduced by the classifier.

9. The method as recited in claim 1, wherein determining the discriminability includes determining a difference in a frequency count-based distribution between each risk factor in the target cluster and the other clusters.

10. The method as recited in claim 1, further comprising validating each of the individual risk factors using the individual data by filtering out the common risk factors that do not indicate actual risk.

11. The method as recited in claim 1, wherein the individual data includes one or more of:

diagnosis, lab results, medication, hospitalization records, questionnaire data and genetic information for the target cluster.

12. A method for individual risk factor identification, comprising:

identifying common risk factors for one or more risk targets from population data;

stratifying individuals into clusters based upon the common risk factors, wherein a distance between each cluster represents a similarity based upon the common risk factors, and wherein a closest pair of clusters is iteratively merged into a single cluster until a threshold condition is reached;

identifying the clusters as one of a plurality of risk levels including at least one high-risk cluster and at least one low-risk cluster;

determining, using a processor, a discriminability of each of the common risk factors for a target cluster using individual data of the target cluster to provide re-ranked common risk factors as individual risk factors for the target cluster, such that the discriminability is a quantitative measure of how a risk factor discriminates its cluster from other clusters, the other clusters including at least one of other high-risk clusters, low-risk clusters and a general population; and

customizing and performing, using a personalized user interface and dashboard display operatively coupled to at least one hardware based care-management machine, a personalized hardware-based care management process by controlling the at least one hardware based care-management machine for personalized treatment of a particular individual, the personalized user interface and dashboard display comprising a customized graphical user interface (GUI) configured for customizing healthcare plans, tailored to the particular individual based on the individual risk factors determined by the identifying common risk factors the stratifying individuals into clusters, and the re-ranked common risk factors, and for providing real-time clinical decision support at a point-of-care for the particular individual.

13. The method as recited in claim 12, wherein identifying the clusters includes identifying the clusters as one of a plurality of risk levels based upon a proportion of at-risk individuals in each cluster.

14. The method as recited in claim 13, wherein at-risk individuals are identified based upon a risk score.

15. The method as recited in claim 14, wherein the risk score is assigned using a classifier.

16. The method as recited in claim 12, wherein determining the discriminability includes determining an amount each risk factor contributes to training a classifier for a particular comparison configuration while accounting for a bias introduced by the classifier.

17. The method as recited in claim 12, wherein determining the discriminability includes determining a difference in a frequency count-based distribution between each risk factor in the target cluster and the other clusters.

18. The method as recited in claim 12, further comprising validating each of the individual risk factors using the individual data by filtering out the common risk factors that do not indicate actual risk.

19. The method as recited in claim 12, wherein the individual data includes one or more of:

diagnosis, lab results, medication, hospitalization records, questionnaire data and genetic information for the target cluster.

20. A method for individual risk factor identification, comprising:

identifying common risk factors for one or more risk targets from population data;

stratifying individuals into clusters based upon the common risk factors, wherein a distance between each cluster represents a similarity based upon the common risk factors, and wherein a closest pair of clusters is iteratively merged into a single cluster a threshold condition is reached;

identifying the clusters as one of a plurality of risk levels including at least one high-risk cluster and at least one low-risk cluster;

determining, using a processor, a discriminability of each of the common risk factors for a target cluster using individual data of the target cluster to provide re-ranked common risk factors, such that the discriminability is a quantitative measure of how a risk factor discriminates its cluster from other clusters, the other clusters including at least one of other high-risk clusters, low-risk clusters, and a general population;

validating each of the re-ranked common risk factors using the individual data to provide individual risk factors for the target cluster by filtering out the common risk factors that do not indicate actual risk; and

customizing and performing, using a personalized user interface and dashboard display operatively coupled to at least one hardware based care-management machine, a personalized hardware-based care management process by controlling the at least one hardware based care-management machine for personalized treatment of a particular individual, the personalized user interface and dashboard display comprising a customized graphical user interface (GUI) configured for customizing healthcare plans, tailored to the particular individual based on the individual risk factors determined by the identifying common risk factors, the stratifying individuals into clusters, and the re-ranked common risk factors, and for providing real-time clinical decision support at a point-of-care for the particular individual.