Predicting Next Year`s Allowed Amount

 Predicting Next Year’s Allowed Amount
Neuron Use Case
TM
ColdLight is a company focused on making the complex process of data science simple for
companies through the use of the Neuron platform. This document is an overview of a selfinsurance cost/risk use case in a major US-based manufacturer.
Use Case Overview
Due to rapidly rising insurance and healthcare costs, some larger companies are selecting to
administer and underwrite insurance coverage for their employees. The idea of ‘self-insurance’ is
based on the theory that by retaining control, calculating risks and paying resulting claims and
losses, the overall cost of self-insuring is substantially less expensive than buying a traditional group
or company policy from an insurance company.
This use case describes how Neuron enabled a prominent manufacturer to accurately predict which
employees represented the greatest healthcare cost risk over the next 12-months. The success of
their cost savings strategy of self-insuring employee healthcare plans depended on the accurate
forecasting of individual risk and costs.
What Was the Problem?
Historically, advanced analytical techniques, such as regression, have been used to predict future
healthcare expenditures. However, this approach has been challenging due to the wide distribution
of healthcare expenditures, reporting averages for groups of employees and individual scoring
accuracy. As such, accurate results were difficult to achieve with the methods currently in use.
Further, these methods made it difficult to accurately identify risk at an individual level and it was
completely incapable of specifying any action steps to contain costs and improve health outcomes.
Employing teams of data scientists to build and maintain predictive models is costly and difficult to
maintain from a technical and human perspective. Due to human bias, rapidly changing data and
the difficulty in building, validating and implementing manual predictive models, it is nearly
impossible for most companies to scale their infrastructure to support the demand for data driven
decision making capabilities.
Why Neuron Was Selected
Neuron was selected because of its inherent ability to rapidly process and mine historical data
(claims, clinical and pharmacy) resulting in prediction models containing a high level of accuracy –
all automatically developed through the use of artificial intelligence within the platform. The
platform enables continuous, automated, learning experience for the manufacturer by processing
new data as it arrives, predicting, recommending actions and repeating the process through time in
an effort to evolve the accuracy of the models and optimize the performance of the process.
www.ColdLight.com Screenshot: NeuronDiscover - Top features predicting next year cost.
Results
Initially, Neuron stratified employees based on their predicted costs and grouped them into
categories from low to high in terms of performance. Within each of these categories, Neuron
provided a list of unique characteristics that describe the individuals within each category.
Further, Neuron automatically created, tested and validated a predictive model based on all of the
claims data available in the database. Each model is continuously and automatically refined over
time by Neuron’s artificial intelligence engine using the unique characteristics and behaviors
reflected in the data. The accuracy of the predictions is tested using standard industry measures
and the score is provided to support ongoing optimization efforts. The tuning of models and the
ability improve accuracy through time is achieved by adding new data sets and unique features to
enable the model to more precisely predict the outcome it is focused on.
In addition, Neuron produced scored lists of individual employees in rank order from the highest
predicted costs to the lowest along with specific recommendations that can improve the outcome
for each person and cohorts of members. The recommendations are generated to enable
personalized intervention strategies to lower costs, improve the quality of care and improve
business outcomes.
As a result of using Neuron, this manufacturer has a deeper and more accurate understanding of
the drivers of cost associated with their employee/members and more confidence to effectively
finance the health plan.
www.ColdLight.com