OBJECTIVES: Understanding payer perception and price forecasting in the pharmaceutical industry is complex, involving multiple different factors such as: health technology assessments (HTAs) , health economics, and market volatility. Value-based pricing (VBP) aims to align product price with payer perception by evaluating the multiple factors of a product’s target product profile (TPP) considered by payers and relating these to a product’s price point.

Previous work by Global Pricing Innovations (GPI) has demonstrated the high accuracy of price forecasting with a value-based approach. This research utilised a multi-criteria decision analysis (MCDA) framework for orphan indications to assess the overall value of pharmaceutical assets. The framework was built to reflect payer decision making as established in HTA documents and was validated using payer interviews.

METHODS: This study investigates the potential of various regression and machine learning models to assess feature importance and evaluate model performance in determining the value and price of an asset. Various linear, polynomial and machine learning models (ridge, lasso, gradient boosting and XG boost) were tested using orphan datasets containing products reimbursed since 2016 across EU4, UK and the US.

RESULTS: Our analysis revealed that size of patient population was a key driver of value across all payer archetypes, followed by a combination of clinical benefit, trial design, burden, product characteristics, and country-specific considerations (such as HST appraisal in the UK or overall judgement of innovativity in Italy). Additionally, machine learning models achieved lower relative mean absolute error (MAE) across all analysed models and markets, indicating improved performance often seen with more complex models.

CONCLUSIONS: Our results demonstrate that large data analysis and machine learning models enhance understanding of payer behaviour by highlighting differences in value perception across payer archetypes and has the potential to further improve value and pricing forecast methodologies.