A0354
Title: Framing the evidence: A text mining approach to regulatory persuasion in pharmaceutical pricing
Authors: Paul Hofmarcher - University Salzburg (Austria) [presenting]
Abstract: Pharmaceutical companies engage in high-stakes persuasion when submitting benefit assessment dossiers to health authorities. In Germany, this process requires firms to demonstrate the added value of new drugs to negotiate reimbursement prices. While the scientific evidence provided in these submissions is standardized, it is hypothesized that the "framing of information" may influence the reimbursement prices. A novel statistical framework, the structural text-based scaling (STBS) model, is proposed, which combines Poisson factorization topic modeling with author-level framing effects to detect variation in textual emphasis across those dossiers. The Bayesian hierarchical model allows topic-specific deviations in language use to be regressed on covariates such as company identity, orphan drug status, and therapeutic class. Estimation is performed via variational inference. Using a corpus of oncological drug dossiers submitted to regulatory authorities, preliminary results suggest modest but systematic differences in linguistic framing across firms and drug categories. This might have potential implications for regulatory pricing decisions.
