In this study we develop a prediction model to detect observations with high probability of data accuracy errors with impact on quality of epidemiological measures produced by ROR-Sul. Logistic regression was used to obtain the model and its reproducibility was evaluated by means of sensibility and specificity. The results identified predicting factors of registry errors; however, the model’s reproducibility indicate that, if there is no way to reach better sensitivities, this method of predicting cancer errors is neither efficient nor profitable.

keywords: cancer registry; quality control; accuracy; data validity; data reliability.