Express Scripts Strategic Development, Inc.

Evernorth Strategic Development, Inc.

Christopher G Lehmuth

Pritesh J Shah

Christopher R Markson

Amit K Bothra

Bradley D Flynn

Alexi E Makarkin

David Alfredo Tomala

Stanislav Samarin

Logan R Meltabarger

Robert Monzyk

Bose S Daggubati

Gang Wu

Varun Tandra

Adithya Chowdary Boppana

Snezana Mahon

Zachary A Goodman

Amy C Gross

Kelcey A Blair

Jackie L Ahlstrom

Jeremy A Rower

Rochelle R Henderson

Callie S Carter

Rahul Rajendran

Systems and methods herein describe a mental health management system. The mental health management system accesses patient data associated with a patient from a database, determines that the patient is associated with a trigger event, identifies a category associated with the trigger event, generates a notification based on the trigger event and the identified category, validates the notification based on a notification history and transmits the notification to a computing device.

Mental health predictive model management system

A modeling computing device for predicting relative patient hazards based on predicted patient pharmaceutical adherence is provided. The modeling computing device includes a processor configured to receive a set of adherence data associated with adherence to a pharmaceutical prescription by a patient. The processor is also configured to determine a set of predictive adherence scores and determine a variability score associated with the set of predictive adherence scores. The processor is further configured to integrate the set of predictive adherence scores and the variability score into a feature value. The processor is also configured to identify a hazard regression model and to apply the feature value to the hazard regression model to generate an integrated patient hazard model. The processor is also configured to process the integrated patient hazard model to determine a risk level associated with the patient, and transmit an alert based on the risk level.

Systems and methods for predicting relative patient hazards using pharmaceutical adherence predictive models

A method includes receiving historical data collected from a client associated with members. The historical data includes per-member metrics for the client and demographic information for the members. The method includes identifying therapeutic classes for the client based on the per-member metrics and the demographic information. The method includes segmenting the historical data into a data set for each therapeutic class. The method includes, for each therapeutic class of the set of therapeutic classes, determining a pattern for the per-member metrics corresponding to the respective therapeutic class, generating a respective predictive model based on the pattern, and training a neural network of the respective predictive model using a two-stage training process. The predictive model is configured to generate, as output for the therapeutic class, a per-member metric prediction for an input period of future time. The method includes generating predictions for the therapeutic classes using the predictive models.

Neural network artificial intelligence system for class-based modeling

A method includes storing a parameter related to a user, storing descriptive data for multiple identifiers, and indexing multiple events. Each event corresponds to a physical object supplied to a user on behalf of an entity. The method includes identifying a first set of identifiers based on commonality among the descriptive data. The method includes training a machine learning model for the first set of identifiers based on event data from within a predetermined epoch. The method includes receiving an indication of a selected identifier and determining a first intake metric of the selected identifiers using the machine learning model. The method includes determining a second intake metric of the selected identifier and the parameter and transforming the user interface according to the first and second intake metrics. The first intake metric represents an amount of resources expected to be received during a second epoch subsequent to the predetermined epoch.

Systems and methods for user interface adaptation for per-user metrics

A content analysis system includes a processor executing instructions from memory. The instructions include, in response to receiving a request signal from a user device, obtaining feedback items, each having a source indicator; identifying unique source indicators; and, for each source indicator, aggregating corresponding ones of the feedback items. A set of filtered feedback items is generated according to either first or second access levels associated with a user of the user device. A subset of filtered feedback items is selected according to a date range specified by the request signal, a set of automated rules is applied, and natural language processing is performed based on frequency of presence of salient terms to identify themes. A control signal is transmitted to a user interface of the user device instructing display of a representation that indicates a change in the frequency of the identified themes over the specified date range.

Medical processing systems and methods

Systems and methods in this document describe a mental health management system. The mental health management system accesses patient data associated with a patient from a database, determines that the patient is associated with a trigger event, generates a prediction, using a predictive modeling system trained to analyze the patient data, the prediction corresponding to a probability that the patient's current medication data will be modified, stores the prediction in association with the patient data, and transmits the prediction to a computing device.

Predictive modeling for mental health management

A method includes receiving a first set of identifiers selected based on commonality among descriptive data corresponding to the identifiers of the first set. Each identifier corresponds to a user who has been supplied a physical object. The method includes identifying event data for the first set within a specified epoch. The method includes training a machine learning model for the first set using the identified event data. The machine learning model is trained using parallel processing of records from a storage structure storing the event data. The parallel processing includes assigning analysis of the event data of each of a subset of the first set to respective processor threads for parallel execution on processing hardware. The trained machine learning model is configured to receive a selected identifier and generate an output representing an amount of resources expected to be used by the selected identifier for a subsequent epoch.

A method includes generating an intervention model by determining principal components for features of a training set, associating each feature of the training set with one of the principal components, selecting features of the training set most closely correlated with the principal components, performing a regression analysis on the selected features to determine a subset of the selected features that are most closely correlated with a model target, training a machine learning model with the subset, verifying the trained machine learning model with a verification set, and saving the verified trained machine learning model as the intervention model. The method includes determining an intervention expectation indicating a likelihood that the user will take action in response to an intervention being execute, determining a likelihood of a gap in care for the user, selecting and scheduling an intervention for execution based on the care gap likelihood and the intervention expectation.

Automated intervention system based on channel-agnostic intervention model

A method includes generating an intervention model by determining principal components for features of a training set, associating each feature of the training set with a principal component, selecting features of the training set most highly correlated with principal components, training a machine learning model with at least some of the selected features, and saving the verified trained machine learning model as the intervention model. The method includes determining multiple channel-specific intervention expectations. Each channel-specific intervention expectation indicates a likelihood that the user will take action in response to an intervention being executed using the engagement channel corresponding to the channel-specific intervention expectation. The method includes selecting an intervention and scheduling the selected intervention for execution.

Channel-specific engagement machine learning architecture

A method for generating predictions for year-over-year change in drug spending and per member per month spending and for providing the predictions via a web portal includes receiving data collected from a health plan. The data includes per member per month costs of the health plan and demographic information of members of the health plan. The method includes selecting therapeutic classes based on the per member per month costs and demographic information, segmenting the data by the therapeutic classes, and detecting patterns by analyzing the segmented data. The method includes generating models for the therapeutic classes based on the patterns, and generating predictions for year-over-year change in drug spending and per member per month spending for the therapeutic classes by utilizing the models. The method includes providing the predictions via a web portal in at least one of a displayable graphical format and a downloadable data structure.

Artificial intelligence system for modeling drug trends

A computer system for transforming a user interface according to data store mining includes a data store configured to store a parameter related to a user and index event data of a set of events. A data processing circuit is configured to identify a first set of identifiers and train a machine learning model based on event data by the data store. An interface circuit is configured to receive an indication of a selected identifier of the plurality of identifiers, determine a first intake metric of the selected identifier using the machine learning model, and a second intake metric of the selected identifier and the parameter using the machine learning model. The interface circuit is configured to transform the user interface according to the first intake metric and the second intake metric.

A method includes generating an intervention model for a population of users based on contact data, demographic data, and engagement data indicating successfulness of prior interventions for each of the population of users. The method includes, obtaining first data related to a first user, including engagement data indicating successfulness of prior interventions with the first user. The method includes supplying the obtained data as input to the intervention model to determine an intervention expectation, which indicates a likelihood that the first user will take action in response to an intervention. The method includes determining a likelihood of a gap in care. The method includes, in response to the care gap likelihood exceeding a minimum threshold, selecting and scheduling execution of a first intervention. The first intervention is one of a real-time communication with the first user by a specialist and an automated transmission of a message to the first user.

A computer-implemented method includes generating an intervention model for a population of users based on engagement data indicating successfulness of prior interventions for each of the population of users. Each prior intervention corresponds to one of multiple engagement channels, and the intervention model includes multiple channel-specific models. The method includes supplying data related to a first user as input to the intervention model to determine multiple channel-specific intervention expectations. Each channel-specific intervention expectation indicates a likelihood that the first user will take action in response to an intervention being executed using the corresponding engagement channel. The method includes determining a likelihood of a gap in care for the first user, and in response to the gap in care likelihood exceeding a minimum threshold, selecting a first intervention according to the channel-specific intervention expectation that has a highest determined value, and scheduling the selected first intervention for execution.

A computer system for dynamic adaptation of a user interface according to data store mining includes a data store configured to index event data of a plurality of events. A data analyst device is configured to render the user interface to a data analyst and transmit a message that identifies a selected identifier of the plurality of identifiers. A data processing circuit is configured to train a machine learning model based on event data stored by the data store for a first set of identifiers from within a predetermined epoch. An interface circuit determines an interface metric for the selected identifier based on the determined output of the selected identifier and transmits the interface metric to the data analyst device. The data analyst device is configured to, in response to the interface metric from the interface circuit, selectively perform a modification or removal of a second user interface element.

A computer system includes an input configured to receive a first image of medication located in a receptacle, memory, and a processor configured to execute instructions including creating a second image based on the first image, dividing pixels of the second image into first and second subsets, and scanning the second image along a first axis to count, for each point along the first axis, a number of pixels in the first subset along a line perpendicular to the first axis that intersects the first axis at the point. The instructions also include estimating positions of first and second edges of the receptacle along the first axis based on the counts of the pixels, defining an opening of the receptacle based on the estimated positions of the first and second edges, and outputting a processed image that indicates areas of the image that are outside of the defined opening.

Growing community of inventors

St. Louis, MO, United States of America

Christopher G Lehmuth