Computerized system and method for identifying members at high risk of falls and fractures

Abstract

A computerized system and method for automatically estimating the likelihood of having a fall leading to a fracture/dislocation within a specified period is described, and comprises a predictive model for guiding patients to the right course of treatment and encouraging discussions with their doctors for better outcomes. The system and method extracts member's health information from health administrative claims data, including clinical and pharmacy data, and estimates the probability of a fall for that member. Patients with high risk scores are selected for various clinical programs and interventions to manage their health conditions and reduce their likelihood of falling.

What is claimed is: 
     
         1 . A system for identifying one or more members of a health insurance member population at risk for falling within a predetermined time period, the system comprising:
 one or more computing devices comprising a falls prediction model;   one or more processors;   one or more electronic storage devices comprising:
 falls model triggers comprising at least one of: alcohol abuse, Alzheimer's disease, blood vessel injury, cognitive dysfunction, concussion, dementia, dialysis, difficulty walking, epilepsy and convulsion, face, eye, or neck contusion, face, neck, or scalp injury, fracture, hallucinations, hip, knee or joint pain, hypotension, motor problems, muscle weakness, nerve or spinal injury, obesity, one or more previous falls, other injury, Parkinson's disease, and stroke; 
 falls predictors comprising at least one of: injury/poisoning incidence count, a lower limb fracture, head, neck, or spine trauma incidence count, previous falls, an upper limb fracture, a neck or trunk fracture, a dislocation, a narcotic prescription, age, a hospital emergency room visit, obesity, governmental agency health score, a bone disorder, gender, race, an anti-depressant prescription, and an anti-hypertensive prescription; and 
 executable software instructions, which when executed by the one or more processors configure the one or more processors to:
 receive member data and consumer data for each member of the health insurance member population, wherein said member data comprises data selected from the group consisting of: medical claims data and pharmacy claims data, wherein said consumer data comprises data selected from the group consisting of: demographic data, geographic data, and financial data; 
 analyze said received member data for each member of the health insurance member population to identify a subset of members within said health insurance member population having one or more of said falls model triggers present in said member's data for the respective member; 
 process the member data and said consumer data for said subset of members using an algorithm configured to extract features from the member data for said subset of members, wherein said falls predictors are selected to correspond with the extracted features; 
 provide a plurality of training conditions to said one or more computing devices comprising the falls predictive model to develop the falls predictive model, wherein said training conditions comprise at least one of: an unintentional fall, a skull fracture, a neck fracture, a trunk fracture, an upper limb fracture, a lower limb fracture, and a dislocation of bones; 
 provide said member data and said consumer data for said subset of members to the one or more computing devices comprising said falls predictive model; and 
 receive a calculated falls risk score from the one or more computing devices comprising said falls predictive model, wherein said falls risk score represents the likelihood that the respective member of said subset of members will visit an emergency room as a result of experiencing a fall within the predetermined time period, and wherein said calculated falls risk score is determined, at least in part, based on the presence or absence of each of said falls predictors in said member data for the respective member. 
 
   
     
     
         2 . The system of  claim 1  wherein:
 said algorithm utilizes a technique selected from the group consisting of: variable selection, principle component analysis, and clustering 
 
     
     
         3 . The system of  claim 1  wherein:
 said features are extracted using a temporal feature extraction technique. 
 
     
     
         4 . The system of  claim 1  wherein:
 said falls model is developed utilizing one or more modeling techniques selected from the group consisting of: decision tree, logistic regression, artificial neural networks, and ensemble. 
 
     
     
         5 . The system of  claim 1  further comprising:
 additional software instructions stored at the one or more electronic storage devices, which when executed, configure the one or more processors to sort the received calculated falls risk score into one of a plurality of groups according to a severity level indicated by the received calculated falls risk score. 
 
     
     
         6 . The system of  claim 5  further comprising:
 additional software instructions stored at the one or more electronic storage devices, which when executed, configure the one or more processors to assign a clinical program or intervention for each of the members in said subset of members, wherein said assignment is determined based on the group into which said member's calculated falls risk score has been sorted, and wherein said intervention is adapted to reduce the member's calculated falls risk score. 
 
     
     
         7 . The system of  claim 6  further comprising:
 additional software instructions stored at the one or more electronic storage devices, which when executed, configure the one or more processors to enroll said member in said assigned clinical program or intervention. 
 
     
     
         8 . The system of  claim 1  further comprising:
 additional software instructions stored at the one or more electronic storage devices, which when executed, configure the one or more processors to accomplish the temporal feature extraction by:
 gathering member data for each feature for a time period prior to a date in question, wherein said member data comprises a number of events, each of which is associated with a particular time; 
 dividing the time period into a number of equal intervals spanning the time period; 
 sorting the gathered data such that each event is sorted into the interval corresponding with the particular time for the respective event; 
 summing the data falling within each interval; 
 assigning a weighting to each interval in decreasing fashion such that the interval temporally closest to the date in question gets the highest weight and the interval temporally farthest from the date in question gets the lowest weight; 
 multiplying the summed value for each interval by the weighting for the respective interval to determine a weighted sum for each interval; and 
 summing the weighted sums to determine a cumulative sum, wherein the cumulative sum is utilized to determine the calculated falls risk score. 
 
 
     
     
         9 . The system of  claim 1  further comprising:
 additional software instructions stored at the one or more electronic storage devices, which when executed, configure the one or more processors to accomplish the temporal feature extraction by:
 gathering member data for each feature for a time period prior to a date in question, wherein said member data comprises a number of events, each of which is associated with a particular time; 
 dividing the time period into a number of equal intervals spanning the time period; 
 sorting the gathered data such that each event is sorted into the interval corresponding with the particular time for the respective event; 
 fitting a predictive model to determine a temporal feature value for each extracted feature; and 
 weighting each extracted feature with the respective temporal feature value, wherein the weighted values are utilized to determine the calculated falls risk score. 
 
 
     
     
         10 . A method for identifying one or more members of a health insurance member population at risk for falling within a predetermined time period, the method comprising the steps of:
 receiving, at one or more computing devices, member data and consumer data for each member of the health insurance member population, wherein said member data comprises data selected from the group consisting of: medical claims data and pharmacy claims data, wherein said consumer data comprises data selected from the group consisting of: demographic data, geographic data, and financial data;   analyzing, at said one or more computing devices, said received member data for each member of the health insurance member population to identify a subset of members within said health insurance member population having one or more of said falls model triggers present in said member's data for the respective member;   extracting features from the member data and said consumer data for said subset of members using an algorithm;   providing a plurality of training conditions to said one or more computing devices, wherein said training conditions are selected from the group consisting of: an unintentional fall, a skull fracture, a neck fracture, a trunk fracture, an upper limb fracture, a lower limb fracture, and a dislocation of bones to develop a falls predictive model at the one or more computing devices;   process said member data and said consumer data for said subset of members using the falls predictive model; and   receive, from the falls predictive model, a calculated falls risk score, wherein said falls risk score represents the likelihood that the respective member of said subset of members will visit an emergency room as a result of experiencing a fall within the predetermined time period, and wherein said calculated falls risk score is determined at least in part based on the presence or absence of each of said falls predictors in said member data for the respective member.   
     
     
         11 . The method of  claim 10  wherein:
 said falls predictors are selected to correspond with the extracted features. 
 
     
     
         12 . The method of  claim 11  wherein:
 said falls model triggers comprise at least one of: alcohol abuse, Alzheimer's disease, blood vessel injury, cognitive dysfunction, concussion, dementia, dialysis, difficulty walking, epilepsy and convulsion, face, eye, or neck contusion, face, neck, or scalp injury, fracture, hallucinations, hip, knee or joint pain, hypotension, motor problems, muscle weakness, nerve or spinal injury, obesity, one or more previous falls, other injury, Parkinson's disease, and stroke; and 
 said falls predictors comprise at least one of: injury/poisoning incidence count, a lower limb fracture, head, neck, or spine trauma incidence count, previous falls, an upper limb fracture, a neck or trunk fracture, a dislocation, a narcotic prescription, age, a hospital emergency room visit, obesity, governmental agency health score, a bone disorder, gender, race, an anti-depressant prescription, and an anti-hypertensive prescription. 
 
     
     
         13 . The method of  claim 12  wherein:
 said algorithm utilizes a technique selected from the group consisting of: variable selection, principle component analysis, and clustering. 
 
     
     
         14 . The method of  claim 13  wherein:
 said falls model is developed utilizing a modeling technique selected from the group consisting of: decision tree, logistic regression, artificial neural networks, and ensemble. 
 
     
     
         15 . The method of  claim 14  further comprising the steps of:
 sorting the received calculated falls risk score into one of a plurality of groups according to a severity level indicated by the received calculated falls risk score. 
 
     
     
         16 . The method of  claim 15  further comprising the steps of:
 assigning a clinical program or intervention for each of the members in said subset of members, wherein said assignment is determined based on the group into which said member's calculated falls risk score has been sorted, and wherein said intervention is adapted to reduce the member's calculated falls risk score. 
 
     
     
         17 . The method of  claim 16  further comprising the steps of:
 enrolling said member in said assigned clinical program or intervention. 
 
     
     
         18 . The method of  claim 10  wherein:
 said features are extracted utilizing temporal feature extraction, which comprises the sub-steps of:
 gathering member data for each feature for a time period prior to a date in question, wherein said member data comprises a number of events, each of which is associated with a particular time; 
 dividing the time period into a number of equal intervals spanning the time period; 
 sorting the gathered data such that each event is sorted into the interval corresponding with the particular time for the respective event; 
 summing the data falling within each interval; 
 assigning a weighting to each interval in decreasing fashion such that the interval temporally closest to the date in question gets the highest weight and the interval temporally farthest from the date in question gets the lowest weight; 
 multiplying the summed value for each interval by the weighting for the respective interval to determine a weighted sum for each interval; and 
 summing the weighted sums to determine a cumulative sum, wherein the cumulative sum is utilized to determine the calculated falls risk score. 
 
 
     
     
         19 . The method of  claim 10  wherein:
 said features are extracted utilizing temporal feature extraction, which comprises the sub-steps of:
 gathering member data for each feature for a time period prior to a date in question, wherein said member data comprises a number of events, each of which is associated with a particular time; 
 dividing the time period into a number of equal intervals spanning the time period; 
 sorting the gathered data such that each event is sorted into the interval corresponding with the particular time for the respective event; 
 fitting a predictive model to determine a temporal feature value for each extracted feature; and 
 weighting each extracted feature with the respective temporal feature value, wherein the weighted values are utilized to determine the calculated falls risk score. 
 
 
     
     
         20 . A method for identifying one or more members of a health insurance member population at risk for falling within a predetermined time period, the method comprising the steps of:
 receiving, at one or more computing devices, member data and consumer data for each member of the health insurance member population, wherein said member data comprises medical claims data and pharmacy claims data, wherein said consumer data comprises data comprising demographic data, geographic data, and financial data;   analyzing, at said one or more computing devices, said received member data for each member of the health insurance member population to identify a subset of members within said health insurance member population having one or more of said falls model triggers present in said member's data for the respective member, wherein said falls model triggers comprise: alcohol abuse, Alzheimer's disease, blood vessel injury, cognitive dysfunction, concussion, dementia, dialysis, difficulty walking, epilepsy and convulsion, face, eye, or neck contusion, face, neck, or scalp injury, fracture, hallucinations, hip, knee or joint pain, hypotension, motor problems, muscle weakness, nerve or spinal injury, obesity, one or more previous falls, other injury, Parkinson's disease, and stroke;   extracting features from the member data and said consumer data for said subset of members;   providing a plurality of training conditions to said one or more computing devices, wherein said training conditions comprise an unintentional fall, a skull fracture, a neck fracture, a trunk fracture, an upper limb fracture, a lower limb fracture, and a dislocation of bones to develop a falls predictive model at the one or more computing devices; and   process said member data and said consumer data for said subset of members using the falls predictive model to generate a calculated falls risk score for each member in said subset of members;   wherein said falls risk score represents the likelihood that the respective member of said subset of members will visit an emergency room as a result of experiencing a fall within the predetermined time period, and wherein said calculated falls risk score is determined at least in part based on the presence or absence of each of said falls predictors in said member data for the respective member.