OBJECTIVES: Health technology assessment (HTA) is used to estimate the economic and health benefits of new interventions to inform reimbursement decisions. Health economic models are developed based on data available in the literature, trial data and health economic guidelines. A core input in these analyses is the price of medicines.

Previous research has shown that medicine prices in the Netherlands are not static and that this has a substantial impact on health economic outcomes. The development of a prediction model for medicine price dynamics is key to the future implementation of price dynamics in HTA. In this research, we will present the methodology underlying the determination and of medicine price dynamics in the Netherlands.

METHODS: The model was built based on historic medicine price and volume data supplemented with publicly available data from the European Medicine Agency (EMA) and the GIPdatabank. An assemble of machine learning models was built to forecast price developments of medicine based on six primary regressors. The primary regressors were Anatomical Therapeutic Chemical (ATC) code, time of introduction, patent status and generic availability, orphan drug designation, revenue and introduction price.

Accuracy was assessed based on their performance in six measures of accuracy, primarily the root mean squared error and the root squared. This results in the models' most accurate distribution to determine the medicine price dynamics.

RESULTS: With the use of the model, we are able to accurately predict the medicine price dynamics of medicine in the Netherlands. The outcomes certainty is dependent on the availability of data on products with comparable product characteristics.

CONCLUSIONS: With the available medicine market data in the Netherlands, we are able to accurately determine the expected price dynamics based on its primary regressors.