Patent Issued for Generating an expected prescriptions model using graphical models (USPTO 11132615): International Business Machines Corporation

2021 OCT 18 (NewsRx) -- By a News Reporter-Staff News Editor at Insurance Daily News -- From Alexandria, Virginia, NewsRx journalists report that a patent by the inventors Natarajan, Ramesh (Pleasantville, NY, US), Olsen, Peder A. (New York, NY, US), Weiss, Sholom M. (New York, NY, US), filed on March 10, 2015, was published online on September 28, 2021.

The patent’s assignee for patent number 11132615 is International Business Machines Corporation (Armonk, New York, United States).

News editors obtained the following quote from the background information supplied by the inventors: “The present invention relates generally to the field of healthcare, and more particularly to detecting fraud and/or abuse in medical treatment activity.

“Health care (or “healthcare”) is widely and generally known as the diagnosis, treatment, and prevention of physical and mental impairments in human beings. Likewise, health insurance (or “healthcare insurance”) is insurance against the risk of incurring healthcare expenses. Typically, the cost (to the insured) of health insurance is associated with the overall risk of healthcare expenses for the insured. An important factor known to lead to increases in healthcare expenses (and therefore health insurance costs) is fraud and abuse in medical treatment activity. For example, fraudulent health insurance claims for medically unnecessary medical procedures and/or prescriptions increase the cost of covering healthcare expenses and therefore can lead to an increase in health insurance premiums (as well as significant health damages to those receiving the medically unnecessary medical procedures and/or prescriptions). As such, fraud and abuse detection is an important focus in the healthcare industry.

“Graphical models are known. A graphical model is a probabilistic model for which a graph demonstrates a conditional independence structure between variables. Generally speaking, graphical models use a graph-based representation as the basis for encoding a complete probability distribution over a multi-dimensional space. Known types of graphical models include, for example, Bayesian networks and Markov networks. Graphical models can be used in combination with sets of data to identify predictive relationships between variables. Some sets, called “training sets,” are used to discover potentially predictive relationships, while other sets, called “test sets” are used to assess the strength and/or utility of those potentially predictive relationships.

“Dynamic programming is known. Dynamic programming is a method for solving complex problems by breaking them down into collections of simpler subproblems. Generally speaking, dynamic programming algorithms examine previously-solved subproblems and combine their solutions to give the best solution to a given problem. One known dynamic programming algorithm, which finds the most likely sequence of hidden states that result in a sequence of observed events, is called the Viterbi algorithm.”

As a supplement to the background information on this patent, NewsRx correspondents also obtained the inventors’ summary information for this patent: “According to an aspect of the present invention, there is a method, computer program product and/or system that performs the following steps (not necessarily in the following order): (i) receiving, by one or more computer processors, a first set of observed data including data from a first database and data from a second database, the first database and the second database being independent databases that do not directly reference each other, (ii) identifying, by one or more computer processors, from the first set of observed data, a training subset and a test subset, the training subset and the test subset each including data from both the first database and the second database; (iii) generating, by one or more computer processors, a first graphical model using data from the training subset, the first graphical model representing probabilities of receiving certain values of a first variable of the first database in combination with certain values of a second variable of the second database; (iv) generating, by one or more computer processors, a second graphical model using data from the training subset, the second graphical model representing probabilities of receiving certain value of the first variable of the first database in combination with certain values of a third variable of the second database; (v) determining, by one or more computer processors, respective weights for the first graphical model and the second graphical model by using an expectation maximization method on the data from the test subset; (vi) generating, by one or more computer processors, a third graphical model by interpolating at least the first graphical model and the second graphical model using the respectively determined weights for the first graphical model and the second graphical model; and (vii) defining, by one or more computer processors, one or more links between the data from the first database and the data from the second database using the third graphical model.”

The claims supplied by the inventors are:

“1. A computer-implemented method comprising: receiving, by one or more computer processors, a set of observed data including (i) prescription data from a first database, and (ii) patient care event data from a second database; estimating, by one or more computer processors, a set of latent variables from the set of observed data using an expectation maximization method; generating, by one or more computer processors, a set of graphical models utilizing the set of observed data and the set of latent variables, the set of graphical models including at least: (i) a first graphical model that models relationships between prescription values of the first database and diagnosis values of the second database, and (ii) a second graphical model that models relationships between the prescription values of the first database and procedure values of the second database; interpolating, by one or more computer processors, at least the first graphical model and the second graphical model to generate an expected prescriptions model; and utilizing, by one or more computer processors, the expected prescriptions model to verify viability of a set of prescriptions written by a provider, the set of prescriptions including at least: (i) a first prescription having a date that does not directly correlate to a date of a diagnosis or a date of a procedure in the set of patient care event data, and (ii) a second prescription that does not directly correlate to any diagnoses in the set of patient care event data.

“2. The computer-implemented method of claim 1, further comprising computing, by one or more computer processors, expected prescriptions for a new set of patient care event data using the expected prescriptions model.

“3. The computer-implemented method of claim 2, further comprising identifying, by one or more computer processors, outlier prescription behavior in a new set of prescription data based, at least in part, on the expected prescriptions computed using the expected prescriptions model.

“4. The computer-implemented method of claim 3, wherein identifying the outlier prescription behavior further comprises: computing, by one or more computer processors, using one or more dynamic programming algorithms, a probability that an actual number of prescriptions, from the new set of prescription data, having prescription values of a certain class would exceed an expected number of prescriptions, from the computed expected prescriptions, having prescription values of the certain class.

“5. The computer-implemented method of claim 3, further comprising, in response to identifying the outlier prescription behavior, triggering, by one or more computer processors, an audit process to investigate for insurance fraud.

“6. The computer-implemented method of claim 5, wherein the second database is a medical claims database from a health insurance provider.

“7. The computer-implemented method of claim 3, further comprising: in response to identifying the outlier prescription behavior, generating, by one or more computer processors, a report indicating expected prescription behavior compared to actual prescription behavior for a particular prescriber; and providing, by one or more computer processors, the report to the particular prescriber.

“8. The computer-implemented method of claim 7, further comprising normalizing, by one or more computer processors, the expected prescription behavior compared to the actual prescription behavior in the report according to peer prescription behavior.

“9. The computer-implemented method of claim 7, wherein the report flags outlier prescription activity for the particular prescriber.

“10. The computer-implemented method of claim 1, wherein: the first graphical model further models relationships between the prescription values of the first database and diagnosis value profiles of the second database, the diagnosis value profiles including diagnosis values as well as patient profiles and prescriber profiles; and the second graphical model further models relationships between the prescription values of the first database and procedure value profiles of the second database, the procedure value profiles including procedure values as well as the patient profiles and the prescriber profiles.

“11. A computer program product comprising one or more computer readable storage media and program instructions collectively stored on the one or more computer readable storage media, the stored program instructions executable by one or more processors, the stored program instructions comprising: program instructions to receive a set of observed data including (i) prescription data from a first database, and (ii) patient care event data from a second database; program instructions to estimate a set of latent variables from the set of observed data using an expectation maximization method; program instructions to generate a set of graphical models utilizing the set of observed data and the set of latent variables, the set of graphical models including at least: (i) a first graphical model that models relationships between prescription values of the first database and diagnosis values of the second database, and (ii) a second graphical model that models relationships between the prescription values of the first database and procedure values of the second database; program instructions to interpolate at least the first graphical model and the second graphical model to generate an expected prescriptions model; and program instructions to utilize the expected prescriptions model to verify viability of a set of prescriptions written by a provider, the set of prescriptions including at least: (i) a first prescription having a date that does not directly correlate to a date of a diagnosis or a date of a procedure in the set of patient care event data, and (ii) a second prescription that does not directly correlate to any diagnoses in the set of patient care event data.

“12. The computer program product of claim 11, the stored program instructions further comprising program instructions to compute expected prescriptions for a new set of patient care event data using the expected prescriptions model.

“13. The computer program product of claim 12, the stored program instructions further comprising program instructions to identify outlier prescription behavior in a new set of prescription data based, at least in part, on the expected prescriptions computed using the expected prescriptions model.

“14. The computer program product of claim 13, wherein the program instructions to identify the outlier prescription behavior further comprise: program instructions to compute, using one or more dynamic programming algorithms, a probability that an actual number of prescriptions, from the new set of prescription data, having prescription values of a certain class would exceed an expected number of prescriptions, from the computed expected prescriptions, having prescription values of the certain class.

“15. The computer program product of claim 13, the stored program instructions further comprising program instructions to, in response to identifying the outlier prescription behavior, trigger an audit process to investigate for insurance fraud.

“16. A computer system comprising a processor set and a computer readable storage medium, wherein: the processor set is structured, located, connected and/or programmed to run program instructions stored on the computer readable storage medium; and the stored program instructions include: program instructions to receive a set of observed data including (i) prescription data from a first database, and (ii) patient care event data from a second database; program instructions to estimate a set of latent variables from the set of observed data using an expectation maximization method; program instructions to generate a set of graphical models utilizing the set of observed data and the set of latent variables, the set of graphical models including at least: (i) a first graphical model that models relationships between prescription values of the first database and diagnosis values of the second database, and (ii) a second graphical model that models relationships between the prescription values of the first database and procedure values of the second database; program instructions to interpolate at least the first graphical model and the second graphical model to generate an expected prescriptions model; and program instructions to utilize the expected prescriptions model to verify viability of a set of prescriptions written by a provider, the set of prescriptions including at least: (i) a first prescription having a date that does not directly correlate to a date of a diagnosis or a date of a procedure in the set of patient care event data, and (ii) a second prescription that does not directly correlate to any diagnoses in the set of patient care event data.

“17. The computer system of claim 16, the stored program instructions further comprising program instructions to compute expected prescriptions for a new set of patient care event data using the expected prescriptions model.

“18. The computer system of claim 17, the stored program instructions further comprising program instructions to identify outlier prescription behavior in a new set of prescription data based, at least in part, on the expected prescriptions computed using the expected prescriptions model.

“19. The computer system of claim 18, wherein the program instructions to identify the outlier prescription behavior further comprise: program instructions to compute, using one or more dynamic programming algorithms, a probability that an actual number of prescriptions, from the new set of prescription data, having prescription values of a certain class would exceed an expected number of prescriptions, from the computed expected prescriptions, having prescription values of the certain class.”

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

For additional information on this patent, see: Natarajan, Ramesh. Generating an expected prescriptions model using graphical models. U.S. Patent Number 11132615, filed March 10, 2015, and published online on September 28, 2021. Patent URL: http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PALL&p=1&u=%2Fnetahtml%2FPTO%2Fsrchnum.htm&r=1&f=G&l=50&s1=11132615.PN.&OS=PN/11132615RS=PN/11132615

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