Granted Patent B2

US 12,217,839 · App. 18/063,834 · Granted Feb 4, 2025

Inventors: Vishnuvyas Sethumadhavan (Mountain View, CA); John O. Schneider (Los Gatos, CA); Darren Matthew Schulte (San Francisco, CA); Robert Derward Rogers (Pleasanton, CA)

Assignee: Apixio, LLC

G16H10/60G06F16/313G06F40/134G06F40/169G16H70/00

Abstract

Claims (57)

1. A health information computer system comprising:

at least one memory with instructions stored thereon; and

at least one processor in communication with the at least one memory, wherein the instructions, when executed by the at least one processor, cause the at least one processor to:

receive a plurality of medical record documents;

select a model and model weights based on an annotation activity and a user, wherein the selecting includes modifying a dynamic confidence threshold based at least in part on the annotation activity, wherein the dynamic confidence threshold is applied to at least some documents of the plurality of medical record documents;

identify a finding using the model;

transform the plurality of medical record documents into a structured data set;

select a number of the plurality of medical record documents associated with the finding;

embed links into the structured data set that reference at least one medical record source document of the number of medical record documents;

identify a set of proficient coders by filtering a plurality of coders for proficiency, wherein proficiency is a metric including accuracy in an area associated with the finding and time taken to perform coding in the area:

measure a number of risk adjustment factors each proficient coder of the set of proficient coders completes per hour on average, a rate of false positive coding measured as a rate of annotation disagreement with a supervisor, and an accuracy measure comprising a degree that a given coder agrees with other coders when presented the same task;

route the structured dataset to a coder of the set of proficient coders based at least in part upon the number of risk adjustment factors; and

submit the finding for reimbursement.

2. The health information computer system of claim 1 , wherein the instructions further cause the at least one processor to rank the set of proficient coders in a ranked coder list by at least one of a risk adjustment factor of the number of risk adjustment factors, the accuracy measure, or the rate of false positive coding.

3. The health information computer system of claim 2 , wherein the instructions further cause the at least one processor to:

filter the ranked coder list to identify a best suited coder associated with the finding; and

route the structured data set to the coder, wherein the coder comprises the best suited coder.

4. The health information computer system of claim 1 , wherein the instructions further cause the at least one processor to identify the finding based on evidence from natural language processing.

5. The health information computer system of claim 1 , wherein the instructions further cause the at least one processor to:

receive a selection of one of the number of medical record documents; and

cause display of the one of the number of medical record documents.

6. The health information computer system of claim 5 , wherein the instructions further cause the at least one processor to cause display of an annotated version of the one of the number of medical record documents.

7. At least one non-transitory computer-readable storage medium with instructions stored thereon that, in response to execution by at least one processor, cause the at least one processor to:

receive a plurality of medical record documents;

select a model and model weights based on an annotation activity and a user, wherein the selecting includes modifying a dynamic confidence threshold based at least in part on the annotation activity, wherein the dynamic confidence threshold is applied to at least a portion of the plurality of medical record documents;

identify a finding using the model;

transform the plurality of medical record documents into a structured data set;

select a number of the plurality of medical record documents associated with the finding;

embed links into the structured data set that reference at least one medical record source document of the number of medical record documents;

identify a set of proficient coders by filtering a plurality of coders for proficiency, wherein proficiency is a metric including accuracy in an area associated with the finding and time taken to perform coding in the area;

measure a number of risk adjustment factors each proficient coder of the set of proficient coders completes per hour on average, a rate of false positive coding measured as a rate of annotation disagreement with a supervisor, and an accuracy measure comprising a degree that a given coder agrees with other coders when presented the same task;

route the structured data set to a coder of the set of proficient coders based at least in part upon the number of risk adjustment factors; and

submit the finding for reimbursement.

8. The at least one non-transitory computer-readable storage medium of claim 7 , wherein the instructions further cause the at least one processor to rank the set of proficient coders in a ranked coder list by at least one of a risk adjustment factor of the number of risk adjustment factors, the accuracy measure, or the rate of false positive coding.

9. The at least one non-transitory computer-readable storage medium of claim 8 , wherein the instructions further cause the at least one processor to:

filter the ranked coder list to identify a best suited coder associated with the finding; and

route the structured data set to the coder, wherein the coder comprises the best suited coder.

10. The at least one non-transitory computer-readable storage medium of claim 7 , wherein the instructions further cause the at least one processor to identify the finding based on evidence from natural language processing.

11. The at least one non-transitory computer-readable storage medium of claim 7 , wherein the instructions further cause the at least one processor to:

receive a selection of one of the number of medical record documents; and

cause display of the one of the number of medical record documents.

12. The at least one non-transitory computer-readable storage medium of claim 7 , wherein the instructions further cause the at least one processor to cause display of an annotated version of the one of the number of medical record documents.

13. A method of medical data management implemented by at least one processor in communication with at least one memory, the method comprising:

receiving a plurality of medical record documents;

selecting a model and model weights based on an annotation activity, wherein the selecting includes modifying a dynamic confidence threshold based at least in part on the annotation activity, wherein the dynamic confidence threshold is applied to at least one document of the plurality of medical record documents;

identifying a finding using the model;

transforming the plurality of medical record documents into a structured data set;

selecting a number of the plurality of medical record documents associated with the finding;

embedding links into the structured data set that reference at least one medical record source document of the number of medical record documents;

identifying a set of proficient coders by filtering a plurality of coders for proficiency, wherein proficiency is a metric including accuracy in an area associated with the finding and time taken to perform coding in the area;

measuring a number of risk adjustment factors each proficient coder of the set of proficient coders completes per hour on average, a rate of false positive coding measured as a rate of annotation disagreement with a supervisor, and an accuracy measure comprising a degree that a given coder agrees with other coders when presented the same task;

routing the structured data set to a coder; and

submitting the finding for reimbursement.

14. The method of claim 13 , further comprising:

ranking the set of proficient coders in a ranked coder list by at least one of a risk adjustment factor of the number of risk adjustment factors, the accuracy measure, or the rate of false positive coding;

filtering the ranked coder list to identify a best suited coder associated with the finding; and

routing the structured data set to the coder, wherein the coder comprises the best suited coder.

Assignments (4)

Continuity (9)

References Cited (108)