# US 2021/0279612 A1 · Computerized System and Method of Open Account Processing

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0279612 A1
- Document ID: US20210279612A1
- Status: Published application
- Kind: A1
- Application number: 17/328,065
- Filed: 2021-05-24
- Published: 2021-09-09
- Earliest claimed: 2016-03-07
- Assignees: Crowe LLP
- Applicants: Crowe LLP
- Inventors: Kevin Jarrett; Eric Boggs; Cory Herendeen; Justin Bass
- Web page: https://patentgrep.com/patent/US20210279612A1

## CPC

- G06N 5/04: Inference or reasoning models
- G06N 3/042: Knowledge-based neural networks; Logical representations of neural networks
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06N 20/00: Machine learning
- G16H 40/20: for the management or administration of healthcare resources or facilities, e.g. managing hospital staff or surgery rooms
- G16H 50/70: for mining of medical data, e.g. analysing previous cases of other patients
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA

## Abstract

A computerized system and method for health care facilities to reduce manual handling of at least some open account issues. The system provides healthcare facilities with the ability to resolve current open patient account issues by utilizing the data patterns from a facility's historical patient account transaction activity, to create a machine learning model that can predict resolutions to the open accounts. These patterns are then applied to a facility's current transaction data providing next step resolution to each patient account.

## Claims

### Claim 1

One or more non-transitory, computer-readable storage media comprising a plurality of instructions stored thereon that, in response to being executed, cause a computing device to:
create a plurality of machine learning models comprising at least a credit model and a denial model for predicting resolutions to open accounts by:
obtaining training data representative of historical account transactions between a plurality of patients and a healthcare facility, wherein at least a portion of the training data is targeted specifically for each of the credit model and the denial model;
analyzing the training data to create the plurality of machine learning models configured to make predictions representative of resolutions of open account transactions, wherein the credit model and the denial model are configured to make predictions based on respective open account types of credit balances in which the credit model is to predict an automated resolution based on identifying one or more of problem payors, timing issues, transaction code issues, transaction type issues, or transaction sequence issues and denial of benefits in which the denial model is to predict an automated resolution based on identifying one or more of denial rates, service line issues, healthcare provider specific issues, revenue impact, or management reports, wherein the open account transactions are transactions for accounts that have previously been billed and remain unpaid;
training and/or retraining at least a portion of the plurality of machine learning models by making predictions regarding open account transactions,
wherein in response to at least one of the plurality of machine learning models making a prediction at a confidence level exceeding a threshold confidence level regarding an open account transaction: (i) assigning the prediction to the open account transaction; (ii) flagging the prediction for a holding area when the prediction is a different outcome from a human decision for the open account transaction; and (iii) periodically retraining the plurality of machine learning models based on open account transactions associated with the holding area;
wherein in response to the confidence level associated with the prediction being less than the threshold confidence level regarding the open account transaction, performing a deep learning process by (i) making predictions regarding the open account transaction with a plurality of different machine learning models that are different levels to each other in that a machine learning model of a subsequent level includes output from a machine learning model of a previous level; and (ii) retraining the machine learning model that made the prediction based on the deep learning process;
applying the machine learning model to predict resolutions of a plurality of open account transactions as a function of open account type by:
making predictions on a level of human interaction needed to resolve the plurality of open account transactions using the plurality of machine learning models; and
electronically resolving at least a portion of the open accounts based on predictions made by the plurality of machine learning models.

### Claim 2 (depends on 1)

The non-transitory, computer-readable storage media of claim 1, wherein at least a portion of the plurality of machine learning models are configured to categorize at least a portion of the open account transactions into a category of at least two categories based on a level of human interaction needed to resolve each respective open account transaction.

### Claim 3 (depends on 2)

The non-transitory, computer-readable storage media of claim 2, wherein at least a portion of the plurality of machine learning models are configured to categorize at least a portion of the open account transactions into: (1) a first category if an exception can be resolved by an automated solution without any human involvement, (2) a second category if an exception can be resolved by an automated repeatable work list with some human involvement, and (3) a third category if an exception needs to be resolved with a higher degree of human involvement.

### Claim 4 (depends on 3)

The non-transitory, computer-readable storage media of claim 3, wherein at least a portion of the plurality of machine learning models are configured to categorize into the first, second and third categories based on a predicted confidence level on an amount of human involvement needed to resolve an exception.

### Claim 5 (depends on 4)

The non-transitory, computer-readable storage media of claim 4, wherein at least a portion of the plurality of machine learning models are configured to categorize into the first category if the predicted confidence level is above a first threshold confidence level, categorize into the second category if the predicted confidence level is above a second threshold confidence level, categorize into the third category if the predicted confidence level is above a third threshold confidence level.

### Claim 6 (depends on 5)

The non-transitory, computer-readable storage media of claim 5, wherein the third threshold confidence level is less than the first threshold confidence level and the second threshold confidence level, wherein the second threshold confidence level is less than the first threshold confidence level.

### Claim 7 (depends on 4)

The non-transitory, computer-readable storage media of claim 4, wherein at least a portion of the plurality of machine learning models are configured to categorize an exception in the first category if the predicted confidence level is above ninety percent.

### Claim 8 (depends on 4)

The non-transitory, computer-readable storage media of claim 4, wherein at least a portion of the plurality of machine learning models is configured to categorize an exception in the second category if the predicted confidence level is less than ninety percent, but above fifty percent.

### Claim 9 (depends on 4)

The non-transitory, computer-readable storage media of claim 4, wherein at least a portion of the plurality of machine learning models are configured to categorize an exception in the third category if the predicted confidence level is less than fifty percent.

### Claim 10 (depends on 9)

The non-transitory, computer-readable storage media of claim 9, wherein at least a portion of the plurality of machine learning models are configured to flag any exceptions categorized in the third category for a holding area for additional attention for resolution.

### Claim 11 (depends on 10)

The non-transitory, computer-readable storage media of claim 10, wherein at least a portion of the plurality of machine learning models are periodically updated by obtaining current transactional data and analyzing both the training data and the current transactional data to update the plurality of machine learning models on how to make predictions representative of resolutions of open account transactions.

### Claim 12 (depends on 10)

The non-transitory, computer-readable storage media of claim 10, wherein at least a portion of the plurality of machine learning models are periodically updated with at least a portion of the transactions in the holding area for which the machine plurality of learning models predicted a different outcome from a human decision.

### Claim 13 (depends on 12)

The non-transitory, computer-readable storage media of claim 12, wherein at least a portion of the plurality of machine learning models is periodically updated with the portion of the transactions in the holding area for which the plurality of machine learning models predicted a different outcome from a human decision and a sample of the training to update the plurality of machine learning models on how to make predictions representative of resolutions of open account transactions.

### Claim 14 (depends on 13)

The non-transitory, computer-readable storage media of claim 13, wherein upon creating an updated machine learning model based on the transactions from the holding area and sample of training data, a comparison is made whether the updated machine learning model makes more accurate predictions on resolutions of open account transactions than an existing machine learning model.

### Claim 15 (depends on 14)

The non-transitory, computer-readable storage media of claim 14, wherein if the updated machine learning model is determined to make more accurate predictions on resolutions of open account transactions than the existing machine learning model, the updated machine learning model replaces the existing machine learning model.

### Claim 16 (depends on 14)

The non-transitory, computer-readable storage media of claim 14, wherein a determination of whether the updated machine learning model makes more accurate predictions than the existing machine learning model is based on a score from how each respective machine learning model performs against a validation data set.

### Claim 17 (depends on 16)

The non-transitory, computer-readable storage media of claim 16, wherein the scoring of how the respective machine learning models performs includes a weight for determining performance based on one or more of a maximum time allowed for the machine learning model to analyze the validation data set and/or an accuracy of predicting resolutions of open account transactions.

### Claim 18

One or more non-transitory, computer-readable storage media comprising a plurality of instructions stored thereon that, in response to being executed, cause a computing device to:
create a plurality of machine learning models including a credit model for predicting resolutions to open accounts by:
obtaining training data representative of historical account transactions between a plurality of customers and an entity, wherein at least a portion of the training data is targeted specifically for the credit model;
analyzing the training data to create the plurality of machine learning models configured to make predictions representative of resolutions of open account transactions, wherein the credit model is configured to make predictions based on respective open account types of credit balances in which the credit model is to predict an automated resolution based on identifying one or more of timing issues, transaction code issues, transaction type issues, or transaction sequence issues, wherein the open account transactions are transactions for accounts that have previously been billed and remain unpaid;
training and/or retraining at least a portion of the plurality of machine learning models by making predictions regarding open account transactions,
wherein in response to at least one of the plurality of machine learning models making a prediction at a confidence level exceeding a threshold confidence level regarding an open account transaction: (i) assigning the prediction to the open account transaction; (ii) flagging the prediction for a holding area when the prediction is a different outcome from a human decision for the open account transaction; and (iii) periodically retraining the plurality of machine learning models based on open account transactions associated with the holding area;
wherein in response to the confidence level associated with the prediction being less than the threshold confidence level regarding the open account transaction, performing a deep learning process by (i) making predictions regarding the open account transaction with a plurality of different machine learning models that are different levels to each other in that a machine learning model of a subsequent level includes output from a machine learning model of a previous level; and (ii) retraining the machine learning model that made the prediction based on the deep learning process;
applying one or more of the machine learning models to predict resolutions of a plurality of open account transactions by:
making predictions on a level of human interaction needed to resolve the plurality of open account transactions using the plurality of machine learning models; and
electronically resolving at least a portion of the open accounts based on predictions made by the plurality of machine learning models.

### Claim 19 (depends on 18)

The non-transitory, computer-readable storage media of claim 18, wherein at least a portion of the plurality of machine learning models are configured to categorize at least a portion of the open account transactions into a category of at least two categories based on a level of human interaction needed to resolve each respective open account transaction.

### Claim 20 (depends on 19)

The non-transitory, computer-readable storage media of claim 19, wherein at least a portion of the plurality of machine learning models are configured to categorize at least a portion of the open account transactions into: (1) a first category if an exception can be resolved by an automated solution without any human involvement, (2) a second category if an exception can be resolved by an automated repeatable work list with some human involvement, and (3) a third category if an exception needs to be resolved with a higher degree of human involvement.
