Abstract Background Continuation of interferon (IFN) β-based therapies is important for maximum treatment effectiveness in patients with multiple sclerosis (MS); however, few real-world data are available on discontinuation from IFN β. The aim of this cohort analysis was to estimate real-world discontinuation rates up to 3 years among MS patients in the United States taking subcutaneous (sc) IFN β-1a three times a week (tiw) and to identify whether the factors associated with discontinuation change over time. Methods Patient data were pooled from the MarketScan© Commercial and Medicare Supplemental healthcare claims databases. Patients with ≥1 multiple sclerosis diagnosis who were sc IFN β-1a tiw naïve, had ≥1 year of continuous eligibility before treatment, and ≥1 prescription were followed from first prescription (index date) until date of discontinuation, switch, or end of observation. Treatment status was analysed at exactly 1, 2 or 3 years after index. Multivariable models were used to identify drivers of discontinuation. Results Data from 5956 patients were included; 2862 patients (48.1%) discontinued therapy. Discontinuation rates were 36.9% (1 year), 49.5% (2 years) and 55.8% (3 years). A greater proportion of discontinuing patients had poor adherence (<80% [94.0%] versus ≥80% [51.7%]) or were taking additional medication at follow-up versus the overall population. Factors independently associated with discontinuation irrespective of time on therapy were increasing number of magnetic resonance imaging scans (1 year adjusted odds ratio 1.45, 95% confidence interval 1.26–1.67; 2 years 1.18, 1.06–1.32; 3 years 1.20, 1.07–1.34) and adherence <80% versus ≥80% (1 year 180.95, 135.84–241.03; 2 years 135.80, 100.10–184.23; 3 years 174.89, 115.27–265.38). Factors associated only with early discontinuation (at 1 year) were ≥3 sets of laboratory investigations versus none (2.54, 1.20–5.38), and anxiolytic use at follow-up (1.40, 1.06–1.82). Factors associated only with later discontinuation (at 2 years and/or at 3 years) were antidepressant use at follow-up (2 years 1.46, 1.10–1.94) and greater number of relapses (2 years 1.60, 1.11–2.30; 3 years 2.31, 1.27–4.22). Conclusions Potential drivers of discontinuation change over time. Improved awareness of the drivers of discontinuation could lead to targeted interventions to improve adherence.