US 11,538,112 B1 · Machine learning systems and methods for processing data for healthcare applications

Abstract

A method of predicting an outcome of a prior-authorization, claim, or appeal includes receiving, at a server, a natural language data file representing doctors notes from a provider visit related to a service instance; receiving, at the server, a structured data set including patient profile data, diagnosis and procedure codes, and quantitative data related to a payment requested; processing at least the natural language data file using a medical dictionary to output a set of key medical terms; processing, using a supervised machine learning algorithm, the structured data set and the set of key medical terms to predict an outcome of the payment requested; and outputting an indication of the predicted outcome of the payment requested.

Drawings

Drawings are temporarily unavailable. Open the PDF at USPTO

Drawings

Claims20

Dependent claims

A method of prioritizing denied insurance claims for appeal to one or more payers, the method comprising:

receiving, from a first payer, a denial for a first insurance claim;
receiving, from a second payer, a denial for a second insurance claim;
training a first machine learning algorithm to determine a percentage likelihood that an appeal of an insurance claim to a first payer will result in a paid insurance claim, wherein the first machine learning algorithm is associated with the first payer;
training a second machine learning algorithm to determine a percentage likelihood that an appeal of an insurance claim to a second payer will result in a paid insurance claim, wherein the second machine learning algorithm is associated with the second payer;
receiving a first set of data associated with the first insurance claim denied by the first payer;
receiving a second set of data associated with the second insurance claim denied by the second payer;
inputting at least a portion of the first set of data into the first machine learning algorithm;
using the first machine learning algorithm, determining a first percentage likelihood that an appeal of the first insurance claim to the first payer will result in a first paid claim;
inputting at least a portion of the second set of data into the second machine learning algorithm;
using the second machine learning algorithm, determining a second percentage likelihood that an appeal of the second insurance claim to the second payer will result in a second paid claim;
generating, for the first insurance claim, an appeal document, wherein the appeal document includes an appeal content field, and wherein the generating the appeal document includes automatically populating the appeal content field based on sets of one or more of key words and phrases that are associated with granted appeals as determined by the first machine learning algorithm;
and
submitting, to the first payer, the appeal document.

10 more elements

A method of prioritizing denied insurance claims for appeal to one or more payers, the method comprising:

receiving, from a first payer, a denial of a first insurance claim;
receiving, from a second payer, a denial of a second insurance claim;
training a first machine learning algorithm to predict outcomes of insurance claim appeals to a first payer, wherein the first machine learning algorithm is associated with the first payer;
training a second machine learning algorithm to predict outcomes of insurance claim appeals to a second payer different from the first payer, wherein the second machine learning algorithm is associated with the second payer;
receiving a first set of data associated with the first insurance claim denied by the first payer;
receiving a second set of data associated with the second insurance claim denied by the second payer;
inputting at least a portion of the first set of data into the first machine learning algorithm;
using the first machine learning algorithm, predicting an outcome of an appeal of the first insurance claim to the first payer;
inputting at least a portion of the second set of data into the second machine learning algorithm;
using the second machine learning algorithm, predicting an outcome of an appeal of the second insurance claim to the second payer;
displaying the predicted outcomes on a display device;
generating, for the first insurance claim, an appeal document, wherein the appeal document includes an appeal content field, and wherein the generating the appeal document includes automatically populating the appeal content field based on sets of one or more of key words and phrases that are associated with granted appeals as determined by the first machine learning algorithm;
and
submitting, to the first payer, the appeal document.

11 more elements

A method of prioritizing denied insurance claims for appeal to one or more payers, the method comprising:

receiving, from a plurality of payers, denials for a plurality of insurance claims;
training a plurality of machine learning algorithms to predict outcomes of insurance claim appeals, each of the plurality of machine learning algorithms being associated with a respective one of the plurality of payers;
receiving a set of data associated with the plurality of denied insurance claims, each of the denied insurance claims being denied by one of the plurality of payers;
dividing the set of data into groups of data, each of the groups of data being associated with insurance claims denied by a respective one of the plurality of payers;
for each of the groups of data associated with insurance claims denied by a respective one of the plurality of payers, inputting at least a portion of the received set of data into a respective one of the plurality of machine learning algorithms that is associated with a respective one of the plurality of payers;
using the respective one of the plurality of machine learning algorithms, predicting an outcome of an appeal to the respective one of the plurality of payers;
grouping the plurality of denied insurance claims into a first group associated with a predicted approved outcome and into a second group associated with a predicted denial outcome;
displaying at least the first group associated with a predicted approved outcome on a display device;
generating, for one of the plurality of insurance claims denied by one of the plurality of payers, an appeal document, wherein the appeal document includes an appeal content field, and wherein the generating the appeal document includes automatically populating the appeal content field based on sets of one or more key words and phrases that are associated with granted appeal as determined by at least one of the plurality of machine learning algorithms;
and
submitting, to the one of the plurality of payers, the appeal document.

8 more elements

Dependencies are inferred from the claim text.

Description13,740 words

Summary2,609 words

#The present disclosure relates generally to machine learning algorithms and methods for processing structured and natural language data to output likelihoods of success and automatically populate letter templates.

#Data sources in the healthcare industry are generally output and stored by a variety of systems and in a variety of formats. For instance, before or after a healthcare provider (e.g., a hospital, a physician's office, etc.) provides a service to a patient, the provider often submits an insurance claim to a payer (e.g., the patient's health insurance company) for payment for the service(s). In some cases, the insurance claim is denied by the payer or the payer only pays a portion of the billed amount (often referred to as an adjustment). The provider may appeal the denial or adjustment of the insurance claim to the payer or, in some cases, a third party (e.g., an arbitrator).

#The insurance claim or appeal may contain data from a variety of sources, including data from the electronic health records “EHR” system (e.g. doctor notes, diagnosis and service codes, test results, etc.), and data from the provider's revenue cycle management (“RCM”) that includes medical and billing software, to track patient care episodes from registration to appointment scheduling and payment. The providers submit this information as prior-authorizations, claims and appeals in a variety of different electronic and paper formats. For instance, some payers allow these forms to be submitted electronically, but most do not and require paper submission. The formats may include structured data fields including the procedure and diagnosis codes (e.g., CPT codes), data from .pdf or other document types in the form of letters regarding the claims or appeals, natural language data in the form of text representing doctor's notes, numerical fields relating the billing and payment history, profile information of payments as structured data and others.

Brief description of the drawings7 figures

#FIG. 1 is a process flow diagram of a method for training one or more machine learning algorithms according to some implementations of the present disclosure;

#FIG. 2A is a process flow diagram of a method for prioritizing denied insurance claims for appeal to one or more payers according to some implementations of the present disclosure;

#FIG. 2B is an exemplary ordered database of appeal candidate entries according to some implementations of the present disclosure;

Detailed description10,952 words

#While the disclosure is susceptible to various modifications and alternative forms, specific embodiments thereof have been shown by way of example in the drawings and will herein be described in detail. It should be understood, however, that it is not intended to limit the invention to the particular forms disclosed, but on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention as defined by the appended claims.

#Overview

#Data sources in the healthcare industry are generally output and stored by a variety of systems and in a variety of formats. When a provider (e.g., a healthcare provider, hospital, physician's office, etc.) provides a service to a patient (e.g., an exam, surgery, a diagnostic, a test, treatment, therapy, etc.), the provider often submits an insurance claim to a payer (e.g., a health insurance company) for at least partial payment for the services rendered (or, for services to be rendered in the future). In some cases, the provider may also submit a prior-authorization request before performing the services. Before submitting a claim to the payer, the provider inputs information from an EHR that contains, for example, procedure codes, diagnosis codes, and other information relating to the rendered service and information relating to the patient. The procedure and/or diagnosis codes contained in the electronic health record may be extracted from, for example, doctor notes and/or charts, diagnosis information from exams, lab tests, or notes from radiologists, etc.

Facts

Application

16/440,972Filed 13 Jun 2019 · us provisional application

Priority

15 Jun 2018

Granted

27 Dec 2022

Term ends

15 Oct 2039

Estimated · how

20 years from earliest known nonprovisional domestic filing plus recorded patent term adjustment, including 124 days of adjustment. Maintenance-fee lapses, terminal disclaimers and statutory extensions have not been checked. The continuity chain may be incomplete; this is not an enforceability determination.

Classification5 codes

G06Q 40/08
Finance; Insurance; Tax strategies; Processing of corporate or income taxes: insurance
G06N 20/20
Machine learning: ensemble learning
G06Q 10/10
Administration; Management: office automation; Time management
G06Q
Information and communication technology [ict] specially adapted for administrative, commercial, financial, managerial

Timeline

15 Jun 2018Priority claimedParent application 62/685,678
13 Jun 2019Application filedUS 16/440,972
24 Sep 2020Assigned to Docvocate, Inc.assignment of assignor's interest
27 Dec 2022GrantedUS 11,538,112 B1
12 Oct 2026TodayWithin its term
15 Oct 2039Term endsEstimated, includes 124 days of adjustment

Machine learning systems and methods for processing data for healthcare applications

Life of this patent

Machine learning systems and methods for processing data for healthcare applications

Life of this patent

Life of this patent

Abstract

Drawings

Drawings

Claims20

Description13,740 words

Drawings

Classification5 codes

Life of this patent

Description13,740 words