Harnessing the power of machine learning in Market Access & Pricing 

We harnessed our internal expertise to train and test the accuracy of machine learning models based on regression algorithm in forecasting the price of an orphan product in our research “Can machine learning accurately predict payer behaviour?” presented at #ISPOREU2023. 

case studies launch sequence

Can machine learning accurately predict payer behaviour?

We leveraged the data collected across England, France, and Germany as part of our Horizon methodology. All products assessed by each HTA and reimbursed since 2016 were scored against a value framework with key attributes reflecting payer drivers for orphan products. Annual cost of treatment was also calculated for all products. We trained models using various algorithm prediction, Spearman correlation analysis, linear and polynomial regression models and tested model accuracy with pegcetacoplan, a recently assessed orphan product.

Our results validated the value framework in each market, where each value attribute contributed to the overall payer decision as there was no single price driver. Statistical analysis of different regression models suggested that the relationship between value and price differs across markets. We observed that price prediction accuracy of pegcetacoplan compared to actual price varied by the type of regression model applied, with linear regression in Germany providing the greatest accuracy of within 9%.   

At GPI, we are constantly researching, developing and testing to support and improve our solutions. This research supports our ongoing development of GPI Horizon , our unique platform solution that provides rapid value assessment and price prediction throughout an asset’s lifecycle. It also links to our broader expertise of using our data in innovative ways to support our clients with market access and pricing strategy. 

If you like to know more about our cutting-edge research, see more here or connect with us at

Sign up for our Newsletter

Share this Blog: