Using the “proportion of patients covered” and the Kaplan-Meier survival analysis to describe treatment persistence

Research output: Contribution to journalJournal articleResearchpeer-review


View graph of relations

Purpose: Standard Kaplan-Meier (KM) survival analysis is often used to study treatment persistence estimating the proportion of patients who have not yet experienced a treatment break by a given day after treatment initiation. This method only allows patients to be studied until their first treatment break. The “proportion of patients covered” (PPC) method is another approach to study treatment persistence. It measures the proportion of live patients currently covered by treatment. We aimed to describe the PPC method, show how the KM survival analysis and the PPC method can describe treatment persistence, and discuss the interpretation/application of the methods. Methods: We identified new users of statins, selective serotonin reuptake inhibitors, hormone replacement therapy, and ibuprofen. We used KM estimates and the PPC to describe persistence in the 3 years post treatment initiation, using a grace period of 90 days to define a treatment break. Results: Three years after statin initiation, approximately 40% of patients were still in continuous treatment (KM survival) and 60% of patients still alive were in current treatment (PPC). Corresponding numbers were 12% and 25% for selective serotonin reuptake inhibitors and 9% and 29% for hormone replacement therapy. At 1 year, numbers were 5% and 10% for ibuprofen. The PPC showed markedly less variability than the KM survival analysis with different choices of grace periods. Conclusions: The KM survival analysis and the PPC method can be used to study different aspects of treatment persistence. Together, they provide a more complete picture of treatment persistence and drug use patterns.

JournalPharmacoepidemiology and Drug Safety
Issue number8
Pages (from-to)867-871
StatePublished - 2018

    Research areas

  • drug utilization, Kaplan-Meier survival curves, medication persistence, pharmacoepidemiology, proportion of patients covered, survival analysis