dFDA: A Decentralized Framework for Drug Assessment Using Two-Stage Real-World Evidence Validation

Only 1-10% of adverse drug events get reported to the FDA. Billions of people generate continuous health data through wearables (heart rate, sleep, activity, symptoms daily) while starting and stopping treatments. This data could detect treatment effects but remains unharvested. A two-stage system solves the observational data problem. Stage 1 ($0.100 (95% CI: $0.030-$1.00)/patient): aggregate millions of natural experiments, score causal confidence using Bradford Hill criteria via the Predictor Impact Score. Observational data finds signals cheaply but can't prove causation. Stage 2 ($929 (95% CI: $97-$3K)/patient): confirm top signals through pragmatic trials embedded in routine care. These are 44.1x (95% CI: 39.4x-89.1x) cheaper than traditional Phase III because they use existing clinical infrastructure, not specialized trial sites. This cost estimate is based on a meta-analysis of 108 pragmatic trials plus large-scale implementations like RECOVERY and ADAPTABLE. Output: treatment rankings by effectiveness with validated effect sizes. Companies register treatments in minutes at zero cost. Patients generate outcome data worth $41K (95% CI: $20K-$120K)/patient. The Predictor Impact Score operationalizes six Bradford Hill criteria using saturation functions, enabling automated causal signal detection from observational data at population scale. **Important limitation:** The Predictor Impact Score is a prioritization metric, not proof of causation. High PIS signals warrant experimental investigation; low PIS does not rule out true effects. This framework complements traditional RCTs. Stage 2 pragmatic trials are required to establish validated causal claims. Until PIS thresholds are validated against RCT outcomes, treat them as theoretically-motivated heuristics.