HealthScope: Scaleable COst Prediction Engine

HealthSCOPE ACO Use Case This is a demo of the HealthSCOPE ACO view. It is an extension of the ACO Use Case Demo with Cost Analysis detailed below, with the added functionality of allowing a user to review additional information about individual beneficiaries within a population.


HealthSCOPE Individual Use Case This is a demo of the HealthSCOPE individual view. It allows a single beneficiary to input data on their past healthcare usage to predict how much they might expect to pay for healthcare in the upcoming year.


ACO Use Case Demo with Cost Analysis This version of the ACO demo is an extension of the stable ACO Use Case Demo described below. It has the added functionality of returning a series of demographic-to-cost analysis charts. These are included to demonstrate the types of visualizations that are possible to help managers discover correlations between beneficiary features and costs.


ACO Use Case Demo 5/19/2014 - This demo is targeted at the ACO (Affordable Care Organisation) scenario. For this use case a file describing 5000 beneficiaries is uploaded for data input and an aggregated group cost prediction is returned. In addition to prediction a few demographic features are extracted from the input file. A linear regression model has been trained against a subset of the attributes present in the SID 'Core' file. This version of the ACO Use Case Demo is stable and not currently under development.


Singlecost Demo using Linear Regression 5/5/2014 - This version of the Singlecost Demo utilizes a Linear Regression model. The Linear Regression model returns a discrete value in place of cost ranges. This model is trained against a different dataset than the Random Forest model and the user interface uses a different set of input fields.


Singlecost Demo using API backend 5/5/2014 - Same functionality as the Singlecost Demo but with a more scalable PHP based API. Although the modeling remains the same this version of the user interface includes a more convenient auto complete feature for the primary diagnosis code field.


Singlecost Demo 4/10/2014 - An early yet stable version of the single beneficiary cost prediction demo. A Random Forest model is trained to predict which cost range a sample beneficiary will fall into. Although this demo has been deployed against a Hadoop based model, this version uses native Java code for an aproximate three second prediction.



Research Group Association: