BACKGROUND: Myelodysplastic syndromes (MDS) appear to be underreported to cancer registries, with important implications for cancer and transfusion support service planning and delivery. Two population-based databases were linked to estimate MDS incidence more accurately. METHODS: Data from the statewide Victorian Cancer Registry (VCR) and Victorian Admitted Episode Dataset (VAED, capturing all inpatient admissions), in Australia, were linked. Incidence rates were calculated based on VCR reported cases and using additional MDS cases identified in VAED. Differences between reported and nonreported cases were assessed. A multivariate capture-recapture method was used to estimate missed cases. RESULTS: Between 2003 and 2010, 2692 cases were reported to VCR and an additional 1562 cases were identified in VAED. Annual incidence rate for those aged 65 years and older based on VCR was 44 per 100,000 (95% confidence interval [CI] = 43-45 per 100,000) and 68 per 100,000 (95% CI = 67-70 per 100,000) using both data sets. Cases not reported to VCR were more likely to have had previous malignancies recorded in VAED (23% versus 19%, P = .003) and to require red cell transfusion (59% versus 54%, P = .003). Using the multivariate model, an estimated 1292 cases were missed by both data sources: the re-estimate was 5546 (95% CI = 5438-5655) MDS cases, with an annual incidence in those aged 65 or older of 103 per 100,000 (95% CI = 100-106). CONCLUSIONS: This study reports a higher incidence of MDS using 2 data sources from a large and well-defined population than reported using cancer registry notifications alone.