Background The longitudinal pattern of medication use among family caregivers of people with dementia is not well understood, despite the potential for medication over- or under-use. Objectives To investigate caregiver medication use over a five-year follow-up using data obtained via self-report and from a national prescription register, and compare agreement between medication data obtained from the two sources. Methods Medication data for 222 family caregivers of people with Alzheimer’s disease were obtained via self-report and from the Finnish Social Insurance Institution. Generalised estimating equations, Kappa statistics and related samples Wilcoxon signed rank test were used to analyse medication use over time. Results The mean number of medications used by caregivers increased from 3.4 to 4.1 (self-reported current regular medications) and 2.4 to 2.8 (reimbursed prescription medications during the past 90 days) over five years (p < 0.001). Significantly, more medications were identified via self-report (mean 3.6, SD = 3.3) than the national prescription register (mean 2.6, SD = 2.4, Z= –12.300, p < 0.001). Agreement between the two data sources was good for cardiovascular medications and anti-hypertensives (Kappa = 0.883–0.967, p < 0.001) and medications for acid-related disorders (Kappa = 0.508–0.092, p < 0.001). Agreement was moderate for analgesics (Kappa = 0.281–0.477, p < 0.001) and psychotropics (Kappa = 0.281–0.562, p < 0.002). The proportion of caregivers using five or more medications increased from 27.5% to 44.6% (self-report), and 16.7% to 27.7% (register) (p < 0.001). Conclusion Caregivers use an increasing number of medications in the first five years of caring for persons with dementia, and self-report using a higher number of medications than data from the national prescription register suggest.