OBJECTIVES: Evolution of asthma disease severity over time can be highly dependent on the prescription patterns and drug compliance of patients. The purpose of the modeling is to analyze longitudinal observational data of cohort of asthma patients to describe and quantify the dynamics of adherence, prescriptions, and outcomes and their interaction over time.  METHODS: We explored and analyzed 5 different observational studies following asthma patients in France over up to 2 years. Main patients’ demographics along with prescriptions, ACT and 3-level GINA control scores could be defined every quarter and exacerbations at a given quarter were adjucated based on hospital admissions. Medication possession ratios could be defined quarterly and used as a proxy for adherence. A patient-level dynamic Bayesian inhomogeneous Markov model with quarterly time-step was then developed to jointly describe prescriptions and outcomes over time in relation with adherence proxy using medication possession ratio, adjusting for patients demographics and seasonality. Internal and external validation was performed.  RESULTS: . Such Bayesian model could be fitted to the available data with different parameters informed by one or another data source. Treatment switches were associated with severity at the previous quarter while adherence was significantly improved when patients are switched and when they are less controlled in the previous quarter. Risk of exacerbations was depending on the control score and season at the present quarter and on the risk of exacerbation at the previous quarter. Control was significantly improved by a better adherence and to a lesser extent by a treatment escalation and improved severity at the previous quarter.  CONCLUSIONS: This Bayesian dynamic model allowed quantifying the most important interactions between drug uses and effects on control and exacerbations over time, hence providing a powerful tool for real-world outcomes predictions in asthma patients.