Patent Issued for Application of bayesian networks to patient screening and treatment (USPTO 11562323): Decisionq Corporation

2023 FEB 10 (NewsRx) -- By a News Reporter-Staff News Editor at Insurance Daily News -- According to news reporting originating from Alexandria, Virginia, by NewsRx journalists, a patent by the inventors Eberhardt, III, John S. (Jeffersonton, VA, US), Kalina, Philip M. (Reston, VA, US), Radano, Todd A. (Jeffersonton, VA, US), filed on September 30, 2010, was published online on January 24, 2023.

The assignee for this patent, patent number 11562323, is Decisionq Corporation (Washington, District of Columbia, United States).

Reporters obtained the following quote from the background information supplied by the inventors: “A Bayesian belief network (BBN) is a directed graph and an associated set of probability tables. The graph consists of nodes and arcs. The nodes represent variables, input data for which can be discrete or continuous; however the BBN must segment continuous data into parameterized ranges. The arcs represent causal or influential relationships between variables. More specifically, a BBN is a probabilistic graphical model that represents a set of random variables and their conditional independencies. It is a way of describing complex, probabilistic reasoning.

“Machine learning is a field of computer science that uses intelligent algorithms to allow a computer to mimic the process of human learning. The machine learning algorithm allows the computer to learn information structure dynamically from the data that resides in the data warehouse. The machine learning algorithms automatically detect and promote significant relationships between variables, without the need for human interaction. This allows for the processing of vast amounts of complex data quickly and easily.

“The machine learning models can be scored in different ways: Minimum Description Length (MDL), also known as the Bayesian Information Criterion (BIC), as well as Bayesian Scoring (BDe). Minimum Description Length scoring provides a measure of the quality of a model. It trades off between goodness of fit and model complexity (parsimonious scoring). Goodness of fit is measured as the likelihood of the data given the model. Model complexity equals the amount of information required to store the model, subject to an inflator/deflator set by the user. The BBN networks and/or machine learning models have not been previously utilized in policy decision making processes of insurance plans or in selection of enrollees for disease management or care interventions.”

In addition to obtaining background information on this patent, NewsRx editors also obtained the inventors’ summary information for this patent: “SUMMARY OF THE DESCRIPTION

“According to one aspect of the invention, health insurance claim data for a first group of individuals is obtained to generate a training corpus, including a training set of claim data and a holdout set of claim data. The first group of individuals represents enrollees of one or more health insurance plans and the health insurance claim data represents historic insurance claim information for each individual in the first group. A Bayesian belief network (BBN) model is created by training a BBN classifier using the training set of claim data using predetermined machine learning algorithms. K-fold cross-validation may also be used during the model training process to assess model robustness and feature selection. The BBN model is finally validated using the holdout set of claim data. A successfully validated BBN model can then be used to identify individual-specific risk of disorder and individual-specific likelihood of benefit from intervention and successful treatment for the disorder.

“According to another aspect of the invention, a first set of claim data is received from a client, where the first set of claim data is associated with a first group of individuals representing enrollees of a first health insurance plan. A screening operation is performed using at least one screening Bayesian belief network (BBN) model based on the first set of claim data to identify a subset of individuals in the first group having risk characteristics associated with a disorder. Cost estimation is performed on the subset of individuals using at least one cost BBN model to produce enrollee specific cost estimates. These BBN models can consist of at least one screening BBN model and at least one cost BBN model that were trained using a predetermined machine learning algorithm based on a second, separate set of claim data associated with a second group of individuals of a second health insurance plan. The second set of claim data represents historic insurance claim information for each individual in the second group. The enrollee-specific estimates of risk and cost as developed using a BBN model trained on the second set are then transmitted to the client.

“Other features of the present invention will be apparent from the accompanying drawings and from the detailed description which follows.”

The claims supplied by the inventors are:

“1. A computer-implemented method for evaluating enrollees of a health insurance plan, the method comprising: receiving, by a training module executed within a server, a first set of health insurance claim data for a first group of individuals from a client device over a network, the first group of individuals representing enrollees of one or more first health insurance plans and the first set of health insurance claim data representing historic insurance claim information for each individual in the first group of individuals, wherein the first set of historic insurance claim information includes co-morbid conditions; separating, by the training module, the received first set of health insurance claim data into a training set of claim data and a holdout set of claim data; creating, by the training module, a set of Bayesian belief network (BBN) model candidates, wherein creating the set of BBN model candidates comprises: a) calculating distributions of discrete states for variables of the BBN network, b) generating a first BBN model by performing a preliminary modeling operation on the variables to identify appropriate machine learning parameters, c) generating a second BBN model by performing a global modeling operation to set the appropriate machine learning parameters and to observe variable relationships, d) generating a third BBN model by performing a naive modeling operation to identify a contribution of each of the variables, and e) generating a fourth BBN model by performing a focused modeling operation on a subset of the variables identified in at least one of the preliminary modeling operation, the global modeling operation, and the naive model operation, wherein the set of BBN model candidates comprises at least the first BBN model, the second BBN model, the third BBN model, and the fourth BBN model; assigning, by the training module, a score to each BBN model candidate of the set of BBN model candidates using a predetermined scoring method; selecting, by the training module, a BBN model from the set of BBN model candidates, the selected BBN model having the highest score among the set of BBN model candidates, wherein the selected BBN model is a screening BBN model to identify individuals having risk characteristics associated with a disorder or a cost BBN model to produce cost estimates with and without intervention or treatment of the disorder; validating, by the training module, the selected BBN model using the holdout set of claim data by generating a receiver operating characteristic (ROC) curve based on an outcome of the selected BBN model operated on the holdout set of claim data, and calculating an area under the curve (AUC) based on the ROC curve to evaluate predictive correctness of the selected BBN model given the holdout set of claim data; receiving, by a diagnostic module executed within the server, a second set of health insurance claim data for a second group of individuals from the client device over the network, the second group of individuals representing enrollees of one or more second health insurance plans; identifying, by the diagnostic module, a subset of individuals in the second group of individuals having the risk characteristics associated with the disorder by performing a screening operation on the received second set of health insurance claim data using the screening BBN model; generating a health insurance policy for one or more individuals in the identified subset of individuals, wherein the health insurance policy is based on the screening operation; and transmitting the health insurance policy to the client device over the network.

“2. The method of claim 1, wherein the first set of historic insurance claim information includes at least two of the group consisting of: hospitalizations, outpatient services, procedure codes, diagnostic codes, diagnostic-related group codes, drug prescribed, chronic conditions, an interval between encounters, historic cost and utilization trends, enrollee industry, employment status, sex, age, region of country, and urban or rural setting.

“3. The method of claim 1, wherein the predetermined scoring method comprises a minimum description length (MDL) method.

“4. The method of claim 1, further comprising retraining the screening BBN model and the cost BBN model based on the second set of health insurance claim data and the cost estimates for future use.

“5. A non-transitory computer-readable storage medium having computer instructions stored therein, which when executed by a computer, cause the computer to perform a method for evaluating enrollees of a health insurance plan, the method comprising: receiving, from a client device over a network, a first set of health insurance claim data for a first group of individuals, the first group of individuals representing enrollees of one or more first health insurance plans and the first set of health insurance claim data representing historic insurance claim information for each individual in the first group of individuals, wherein the first set of historic insurance claim information includes co-morbid conditions; separating the received first set of health insurance claim data into a training set of claim data and a holdout set of claim data; creating a set of Bayesian belief network (BBN) model candidates, wherein creating the set of BBN model candidates comprises: a) calculating distributions of discrete states for variables of the BBN network, b) generating a first BBN model by performing a preliminary modeling operation on the variables to identify appropriate machine learning parameters, c) generating a second BBN model by performing a global modeling operation to set the appropriate machine learning parameters and to observe variable relationships, d) generating a third BBN model by performing a naive modeling operation to identify a contribution of each of the variables, and e) generating a fourth BBN model by performing a focused modeling operation on a subset of the variables identified in at least one of the preliminary modeling operation, the global modeling operation, and the naive model operation, wherein the set of BBN model candidates comprises at least the first BBN model, the second BBN model, the third BBN model, and the fourth BBN model; assigning a score to each BBN model candidate of the set of BBN model candidates using a predetermined scoring method; selecting a BBN model from the set of BBN model candidates, the selected BBN model having the highest score among the set of BBN model candidates, wherein the selected BBN model is a screening BBN model to identify individuals having risk characteristics associated with a disorder or a cost BBN model to produce cost estimates with and without intervention or treatment of the disorder; validating the selected BBN model using the holdout set of claim data by generating a receiver operating characteristic (ROC) curve based on an outcome of the selected BBN model operated on the holdout set of claim data, and calculating an area under the curve (AUC) based on the ROC curve to evaluate predictive correctness of the selected BBN model given the holdout set of claim data; receiving a second set of health insurance claim data for a second group of individuals from the client device over the network, the second group of individuals representing enrollees of one or more second health insurance plans; identifying a subset of individuals in the second group having the risk characteristics associated with the disorder by performing a screening operation on the received second set of health insurance claim data using the screening BBN model; generating a health insurance policy for one or more individuals in the identified subset of individuals, wherein the health insurance policy is based on the screening operation; and transmitting the health insurance policy to the client device over the network.

“6. The computer-readable storage medium of claim 5, wherein the first set of historic insurance claim information includes at least two of the group consisting of: hospitalizations, outpatient services, procedure codes, diagnostic codes, diagnostic-related group codes, drug prescribed, chronic conditions, an interval between encounters, historic cost and utilization trends, enrollee industry, employment status, sex, age, region of country, and urban or rural setting.

“7. The computer-readable storage medium of claim 5, wherein the predetermined scoring method comprises a minimum description length (MDL) method.

“8. The computer-readable storage medium of claim 5, wherein the method further comprises retraining the screening BBN model and the cost BBN model based on the second set of health insurance claim data and the cost estimates for future use.”

There are additional claims. Please visit full patent to read further.

For more information, see this patent: Eberhardt, III, John S. Application of bayesian networks to patient screening and treatment. U.S. Patent Number 11562323, filed September 30, 2010, and published online on January 24, 2023. Patent URL (for desktop use only): https://ppubs.uspto.gov/pubwebapp/external.html?q=(11562323)&db=USPAT&type=ids

(Our reports deliver fact-based news of research and discoveries from around the world.)