The objectives of this study were to estimate the data quality of haematological malignancy diagnoses in a hospital discharge registry, and to quantify the impact of any misclassification of diagnoses on survival estimates. We included all patients > or = 15 years living in North Jutland County, Denmark with a first-time discharge diagnosis of a haematological malignancy registered in the Hospital Discharge Registry and the Danish Cancer Registry, the reference standard, from 1994 to 1999. We estimated completeness and positive predictive value (PPV) of haematological malignancies and specific subcategories, as a measure of data quality, and compared mortality rates based on data from the two registries by Cox regression analysis. Completeness in the Hospital Discharge Registry for all haematological malignancies was 91.5% (95% confidence interval (CI) 89.6-93.1) and PPV was 84.5% (95% CI 82.2-86.5). Reviews of the pathological files showed misclassified cases in both registries and thus indicated that both completeness and PPV of the Hospital Discharge Registry were underestimates. Mortality rate ratio for all haematological malignancies when registered in the Hospital Discharge Registry compared with being registered in the Danish Cancer Registry was 0.98 (95% CI 0.88-1.09). Discharge data had some misclassifications but these had no major impact on survival estimates.