Background: Atopic dermatitis (AD) normally onsets in childhood and mostly resolves before adolescences. Disease persistence is known to be difficult to study properly, and current predictors are insufficient to identify more than a small fraction of patients at risk. Objective: To study personal AD medicine history as a retrospective marker of persistent AD. Methods: The study population included all Danish first hospital contacts with a diagnosis of AD (ICD-10, L20) between 1995 and 2012. National register data following the diagnosis were used to define persistent AD activity until 2017 according to personal AD medicine history before diagnosis. Activity was defined as filled prescriptions for topical corticosteroids (TCS) or calcineurin inhibitors (TCI), dermatologist contacts or hospital re-contacts for AD. Risk ratios (RR) for persistent activity (defined as activity >4 of the most recent 5 years) were estimated according to AD medicine history (prescriptions filled prior to diagnosis) adjusted for age at onset, parental AD and basic covariates. Results: A total of 13 628 patients were diagnosed at ages 0-16 years and had up to 21 years of follow-up. 10 years after diagnosis, 67% showed activity (9.5% persistent). Among prior TCS users (69%), the RR_10y of persistent activity increased 1- to 6-fold with increasing strength of strongest TCS/TCI ever, and with number of TCS courses. Prior use of antibiotics (RR_10y 1.32, 95% CI 1.09-1.59) and antihistamines (RR_10y 1.65, 95% CI 1.42-1.91) increased the RR_10y in a dose-dependent manner. In >90% of patients, prior medication use occurred <4 years before diagnosis. Conclusions and clinical relevance: The strength and type of AD medication used in the previous 4 years may predict 10-year persistence of AD. Since children may be misjudged as having milder disease when seen between flares of skin lesions, this information may improve physicians' ability to determine the correct prognosis independently of current AD severity.