Flyway Health produces independent, claims-based research on real-world treatment patterns, formulary impacts, and therapy adherence. This page explains the general standards behind that work. Study-specific cohort definitions, sample sizes, and time windows are described within each individual report.
Flyway's analyses draw on open and closed medical and pharmacy claims covering more than 330 million lives, sourced through Health Verity. Where relevant, claims data is supplemented with formulary data from MMIT and provider taxonomy from the National Plan and Provider Enumeration System (NPPES).
The underlying claims dataset spans 330M+ lives across commercial and public payers. This scale lets Flyway observe low-frequency patterns, segment by payer and provider, and follow patients longitudinally across multiple years of treatment.
Flyway uses retrospective, claims-based cohort analysis. Cohorts are defined by diagnosis codes and treatment fills, then tracked across defined time windows to measure treatment episodes, switching, discontinuation, and persistence. Each study applies explicit inclusion and exclusion logic to control for coding artifacts, and reports sensitivity analyses that test the finding against alternative explanations.
All analyses use de-identified claims data. Flyway does not receive or expose patient-level identifiable information, and reports findings only at aggregate and cohort levels.
Claims-based research has known boundaries. Medication obtained through in-house dispensing such as 340B, direct distribution programs such as ADAP, or cash-pay channels is not fully visible in commercial claims. The observational design means associations cannot be interpreted as causation. Secondary diagnosis coding can introduce noise, which Flyway controls for using same-provider and lab-confirmed subsets where applicable. Each report states the limitations specific to its analysis.
Flyway's research is independent. No manufacturer funds these analyses, and findings are not reviewed or approved by sponsors before publication.