OBJECTIVES: The rates of adverse events (AEs) are an important component of chemotherapy cost effectiveness. However, AE rates in clinical practice are difficult to obtain. This study aimed to determine whether Australian administrative data could identify the incidence of selected chemotherapy AEs.   METHODS: The Elements of Care study (EoC) was a prospective cohort of individuals in New South Wales undergoing chemotherapy for breast, colorectal or lung cancer. Primary data, including AEs experienced, was collected through questionnaires and medical record review. Linked administrative data of prescriptions and medical services for each participant were available from the Pharmaceutical Benefits Scheme (PBS) and Medicare Benefits Schedule (MBS). This data was used to identify if an individual was treated for one of the selected AEs (diarrhoea, vomiting, anaemia and neutropenia) in the three days after a chemotherapy dose. These proxy-identified AE rates were compared with the self-reported AE rates using 2x2 contingency tables, with significance of any differences calculated using odds ratios and chi-square statistics. RESULTS: There were 482 individuals in EoC. The proxy identified much lower rates of AEs than were self-reported, capturing 30% of self-reported cases of nausea and vomiting, 1.3% of self-reported diarrhoea, and less than 1% of self-reported anaemia and neutropenia. Additional analyses did not identify a pattern in the grade of AEs or type of treatment received that the proxy was more likely to identify. CONCLUSIONS: Overall there was poor concordance between the two measures of AE rates. This may be due to low treatment rates for AEs, poor capturing of AE treatments by the proxy, or over-reporting of adverse event by participants. Regardless, it would appear that administrative data such as the MBS and PBS are not suitable for estimating the incidence of AEs in clinical practice, and bottom up data collection techniques such as EoC are essential.