Machine learning risk factor identification and mitigation system

Abstract

A system performing machine learning to predict and identify claims that have a high likelihood of migrating across a predetermined risk threshold and to generate intervention strategies to mitigate the likelihood of migration. The processing system includes a computer server, database engine, computer programming instructions, network connectivity, associated claims, payment, medical, pharmacy and other relevant data, a plurality of statistical and machine learning algorithms and a method for electronically displaying and attaching the results to a business process. The system will use all available data to analyze the medical treatment pattern of a claimant and based on automated findings make recommendations as to appropriate interventions to positively impact claims costs.

What is claimed is: 
     
         1 . A computer system, including at least one computer processor, comprising:
 a data intake component configured to:
 electronically retrieve data stored in a database, said data representing at least one workers' compensation claim having an initial risk score and a treatment pattern associated therewith, and 
 responsive to retrieving the data, store the retrieved data in a processed data component, wherein the data intake component is coupled to the processed data component; 
   a scoring engine component, comprising:
 a model data layer computer processing module, 
 a variable rendering computer processing module, 
 a model scoring layer computer processing module, and 
 a model retraining computer processing module; 
   the model data layer computer processing module, when executed by the at least one computer processor, configured to transform the retrieved data to enable automated scoring;   the variable rendering computer processing module, when executed by the at least one computer processor, configured to:
 render at least one variable for use by the computer system, wherein said at least one variable comprises both data relevant to an identification and mitigation model, and an importance score in part based upon predetermined risk predictions to which the variables are to be applied; 
   a predictive model computer processing module, when executed by at least one computer processor, configured to:
 responsive to rendering the at least one variable, access a plurality of predictive models stored on a memory storage device of the computer system, and 
 responsive to both storing the data in the processed data component and accessing the plurality of predictive models, execute the plurality of predictive models as a function of the at least one variable rendered by the variable rendering computer processing module to yield a migratory risk score for the at least one workers' compensation claim, wherein the migratory risk score represents a likelihood of the at least one workers' compensation claim changing during the treatment pattern, wherein said executing comprises independently executing each of the plurality of models in parallel; 
   the model scoring layer computer processing module, when executed by the at least one computer processor, configured to score the plurality of models, wherein each model is referenced as a function call whereby the function is passed to the model scoring layer computer processing module and the function returns a model scoring record for each model;   the model retraining computer processing module, when executed by the at least one computer processor, configured to retrain each model of the plurality of models with new or updated data based upon the model scoring record such that each model is dynamically adapted in response to the data intake component electronically retrieving new data; and   a report generation computer processing module, when executed by the at least one computer processor, configured to:
 responsive to yielding the migratory risk score, generate a report identifying the at least one workers' compensation claim as a predicted migratory risk claim based on the migratory risk score, and 
 responsive to generating the report, generate suggestions for altering the treatment pattern to mitigate ongoing claim risk based on the variables, 
   wherein the computer system is further configured to provide, by a communications network to a remotely located portable device responsive to generating the report and suggestions, a display including the generated report and suggestions.   
     
     
         2 . The system as recited in  claim 1 , wherein the data stored in the database relates to one or more of the following types of data or tables: ICD9 code data, workers compensation claim data, claim payment data, medical/prescription billing data, U.S. census data, Social Security disability data, state regulatory issues data, medical coding data, pharmacy database data, chronic condition data, comorbidity data, ICD9 cross reference tables, NCCI cross reference tables, target variable manipulation tables, claim exclusion tables, an evidence based medical treatment crosswalk for comparison to medical bill data, and adjuster notes. 
     
     
         3 . The system as recited in  claim 1 , wherein the predictive model comprises one or more of the following types of predictive models: a model that identifies claims likely to exceed a self-insured retention or deductible, and/or a model that identifies claims likely to exceed a predetermined total cost. 
     
     
         4 . The system as recited in  claim 3 , further comprising an identification and mitigation model database containing the data transformed into model scoring records, said model scoring records representing all relevant variables for the predictions. 
     
     
         5 . The system as recited in  claim 1 , wherein the data intake component, the model data layer computer processing module, the variable rendering computer processing module, the model scoring layer computer processing module, the model retraining computer processing module, the predictive model computer processing module, and the report generation computer processing module each comprise separate computer processors, and wherein said computer processors are each coupled to the processed data component, the memory storage device, and to each other. 
     
     
         6 . The system as recited in  claim 1 , wherein the report generation computer processing module, when executed by the at least one computer processor, is further configured to:
 responsive to yielding the migratory risk score, generate a trend indication, wherein the trend indication represents a prediction trend of the migratory risk score relative to a previously yielded migratory risk score; and   wherein the computer system is further configured to provide, by the communications network to the remotely located portable device responsive to generating the trend indication, an alert display including the trend indication when the prediction trend is increasing or decreasing.   
     
     
         7 . The system as recited in  claim 1 , further comprising additional explanatory predictive models to grade the risk of the risk factor category of the variables. 
     
     
         8 . The system as recited in  claim 7 , further comprising specific suggested interventions stored in the intervention database according to risk factor category. 
     
     
         9 . A computer-implemented method of centralizing the identification and notification generation of elevated risk workers' compensation claims, the method comprising:
 retrieving and validating, by a processor executing a data intake component, input data stored in a database, said input data determined to be relevant to mitigating ongoing risk of a claim, said claim having an initial risk score and a treatment pattern associated therewith;   loading, by the processor executing the data intake component responsive to said retrieving, the validated input data in a processed data component, said processor coupled to the processed data component;   rendering, by the processor executing a scoring engine, variables for use by a plurality of predictive models, said variables comprising the validated input data and each having an importance score based at least in part on a predictive model to which the variables are to be applied, wherein the scoring engine is coupled to the processed data component;   accessing, by the processor responsive to said rendering, the predictive models, said predictive models stored on a memory storage device coupled to the processor;   executing, by the processor executing the scoring engine responsive to both said storing and said accessing, the plurality of predictive models as a function of the importance scores of the rendered variables to yield a migratory risk score for the claim, said migratory risk score representative of a likelihood of the claim changing during the treatment pattern, said executing comprising independently executing each of the plurality of models in parallel, wherein the scoring engine is responsive to both the loading and the rendering for accessing and executing the plurality of predictive models;   retrieving from an intervention database, by the processor executing a report engine responsive to said executing, one or more specific suggested interventions to the treatment pattern from the intervention database according to a risk factor category associated with the claim;   generating, by the processor executing the report engine responsive to retrieving the suggested interventions, a report identifying the claim as a predicted elevated risk claim when the migratory risk score exceeds a predetermined threshold, wherein the report includes the retrieved suggestions for altering the treatment pattern to mitigate ongoing claim risk for the elevated risk claim; and   providing, by the processor executing the report engine to a remotely located computing device via a communications network responsive to said generating, a notification display including the generated report.   
     
     
         10 . The computer-implemented method as recited in  claim 9 , wherein the data stored in the database relates to one or more of the following types of data or tables: ICD9 code data, workers compensation claim data, claim payment data, medical/prescription billing data, U.S. census data, Social Security disability data, state regulatory issues data, medical coding data, pharmacy database data, chronic condition data, comorbidity data, ICD9 cross reference tables, NCCI cross reference tables, target variable manipulation tables, claim exclusion tables, an evidence-based medical treatment crosswalk for comparison to medical bill data, and adjuster notes. 
     
     
         11 . The computer-implemented method as recited in  claim 10 , wherein the predictive model comprises one or more of the following types of predictive models: a model that identifies claims likely to exceed a self-insured retention or deductible, and a model that identifies claims likely to exceed a predetermined total cost. 
     
     
         12 . The computer-implemented method as recited in  claim 9 ,
 wherein the processor retrieving and storing the validated input data is a first processor,   wherein the processor rendering the variables is a second processor,   wherein the processor accessing and executing the predictive model is a third processor,   wherein the processor retrieving the suggested interventions, generating the report, and providing the notification display is a fourth processor, and   wherein the first, second, third, and fourth processors are coupled to each other.   
     
     
         13 . The computer-implemented method as recited in  claim 9 , further comprising:
 generating, by the processor responsive to yielding the migratory risk score, a trend indication, the trend indication representing a prediction trend of the migratory risk score relative to a previously yielded migratory risk score; and   providing, by the processor to the remotely located computing device via the communications network responsive to generating the trend indication, an alert display including the trend indication when the prediction trend is increasing or decreasing.   
     
     
         14 . The computer-executable method as recited in  claim 9 , further comprising transforming, by the processor executing the scoring engine, the validated input data into model scoring records, said model scoring records representing all relevant variables for the predictive model of the plurality of predictive models to which the variables are to be applied. 
     
     
         15 . A non-transitory computer-readable storage medium storing processor-executable instructions, the instructions comprising:
 a data intake component that, when executed by at least one processor, configure the at least one processor to:
 electronically retrieve data stored in a database, said data representing one or more workers' compensation claims each having an initial risk score and a treatment pattern associated therewith, and 
 responsive to retrieving the data, store the retrieved data in a processed data component, wherein the data intake component is coupled to the processed data component; 
   a scoring engine component that, when executed by the at least one processor, configure the at least one processor to:
 transform the retrieved data to enable automated scoring, 
 render at least one variable, wherein the at least one variable comprises both data relevant to an identification and mitigation model and an importance score in part based upon predetermined risk predictions to which the at least one variable is to be applied, 
 score a plurality of predictive models, wherein each of the predictive models is referenced as a function call whereby the function returns a model scoring record therefor, and 
 retrain each model of the plurality of predictive models with new or updated data based upon the model scoring record such that each of the predictive models is dynamically adapted in response to the data intake component electronically retrieving new data; 
   a predictive model computer processing module that, when executed by the at least one computer processor, configure the at least one processor to:
 responsive to rendering the at least one variable, access the plurality of predictive models, and 
 responsive to both storing the data in the processed data component and accessing the plurality of predictive models, execute the plurality of predictive models as a function of the at least one variable rendered by the scoring engine component to yield migratory risk scores for the workers' compensation claims, wherein the migratory risk scores represent likelihoods of the workers' compensation claims changing during the treatment pattern, and wherein the executing comprises independently executing each of the plurality of predictive models in parallel; 
   a report generation computer processing module that, when executed by the at least one processor, configure the at least one processor to:
 responsive to yielding the migratory risk score, generate a report identifying the workers' compensation claims as predicted migratory risk claims based on the migratory risk scores, and 
 responsive to generating the report, generate suggestions for altering the treatment pattern to mitigate ongoing claim risk based on the variables, 
   wherein a display including the generated report and suggestions is provided by a communications network to a remotely located portable device responsive to generating the report and suggestions.   
     
     
         16 . The non-transitory computer-readable storage medium as recited in  claim 15 , wherein the plurality of predictive models comprise one or more of the following types of predictive models: a model that identifies claims likely to exceed a self-insured retention or deductible, and/or a model that identifies claims likely to exceed a predetermined total cost. 
     
     
         17 . The non-transitory computer-readable storage medium as recited in  claim 15 , wherein the data intake component, the scoring engine component, the predictive model computer processing module, and the report generation computer processing module each comprise instructions executed by separate computer processors, and wherein said separate computer processors are each coupled to the computer-readable storage medium and to each other. 
     
     
         18 . The non-transitory computer-readable storage medium as recited in  claim 15 , wherein the report comprises a claim severity ranking for at least one of the workers' compensation claims as a function of the migratory risk scores and wherein the report generation computer processing module, when executed by the at least one computer processor, further configures the at least one processor to provide, by the communications network to the remotely located portable device, an alert display including at least one specific suggested intervention associated with the claim severity ranking. 
     
     
         19 . The non-transitory computer-readable storage medium as recited in  claim 15 , further comprising additional explanatory predictive models that, when executed by the at least one processor, grade the risk of the risk factor category of the variables. 
     
     
         20 . The non-transitory computer-readable storage medium as recited in  claim 19 , further comprising specific suggested interventions stored in the intervention database according to risk factor category.