AIMHI

OPIOID RESEARCH GRANT

Federal HHS grant working on the opioid epidemic focused on American Indian and African American population

We mapped geographical data using a GIS system

Many different types of data and categories plotted on one map

The challenge: Finding useful patterns in the data

  • The GIS system is useful because analysts can gain information from the map.
  • Microsoft Excel and business intelligence tools help us find trends, but the questions posed are limited by the analyst’s biases.
  • We looked at various AI tools including Google Tensorflow™, Pytorch™,  and Amazon SageMaker™.
  • We chose to work with simMachines™ because the time and cost to develop predictive models was a fraction of other solutions and excellent reference accounts such as American Express and Starbucks.
  • simMachines meets the recent explainable AI requirements put forth by the FDA.

One of the models predicts if a county will have increased or decreased opiod prescription rates

The simsMachines model will explain why it made the prediction

The "why" of Artificial Intelligence
(AI) is critical

  • The federal government and FDA require explainability.
  • The simMachines model provides detailed high-resolution information about the “why” in a single pass. Explainability is built into the algorithm.
  • Predictive models degrade over time and explainability helps quickly find the reasons for the degradation. This lowers the time and cost for maintaining a good model.
  • Sometimes the “why” is more helpful than the prediction!