HOW ACCURATE IS OUR PREDICTIVE TECHNOLOGY?
Charts with predictions for two drugs that we support.
Downward predicted Tacrolimus results MAD = 9%-11%
Tacrolimus is a widely specified anti-rejection drug used post transplant by kidney and liver patients. It requires constant and careful monitoring and can be difficult to control in early stage use. Taken post transplant it is effectively a drug for life.
Upward predicted Tacrolimus results MAD = 4%-6%
For the statisticians amongst us: Average MAD = 5.8% and Average RMSE = 6.4%. These values were calculated by assuming equal numbers of upwards, downwards and steady state patients. In reality the numbers are likely to break 1:3:6 suggesting Average MAD and RMSE values could be lower.
Typical Warfarin prediction results MAD = 17% - 22%
Warfarin is a widely specified anticoagulant, often used post transplant alongside Tacrolimus. A skittish drug and difficult to predict and control throughout its use cycle, MAD values can be as high as 55% - 65%. Using H-AI's predictive technology improves (INR) predictions by up to 10%.
Predicted results for typical skittish cycle MAD = 23%-27%
Our statistics for Warfarin are: Average MAD = 23% - 27% and Average RMSE = 26% - 39% These values were calculated by averaging the results from the 10,000 data sets we use for testing our predictive technology.Whilst they may seem unconstrained, they represent a quantum improvement on the standardised PDE method of predicting Warfarin INR values.