Accuracy of hospital discharge data for cancer registration and epidemiological research in Northern Ireland

Abstract

Background: Cancer registries provide a basis for many epidemiological studies in cancer. Electronic data provide for prompt, economical data capture for disease registries; doubts, however, exist about their data quality.

Materials and methods: We examined the accuracy for cancer registration of a subset of 7043 electronically captured hospital discharge data.

Results: Note availability was 82%. Of the notes available for examination demographic data accuracy was high; however, 7.4% of cases coded on discharge as cancer had no malignancy recorded in case notes while 4.1% had in-situ or benign tumors. Almost a third had some inaccuracy in coding tumor site. Prevalent cases accounted for 17.2% of cases examined reflecting a new registry. Electronic data capture reduces time spent examining notes; only 40% of cases notified from PAS required a quick validation check. It enhances data completeness; without electronic discharge data 11.5% of the final database would have been missed. The validation check prevented over-inflation of the cancer registration database by 7.5%. Measures of accuracy in the final database were high.

Conclusion: This study shows that discharge data are a valuable data source for cancer registries but require a targeted note review aimed at cases without corroborating data.