Variability in date of prostate cancer diagnosis: a comparison of cancer registry, pathology report, and electronic health data sources

PURPOSE

The date of cancer diagnosis is a critical data element for clinical care and research. Because this date can be abstracted from various data sources, its comparability from source to source is unclear. This study compared the date of diagnosis from multiple sources within the same population of prostate cancer patients.

METHODS

We linked cancer registry, pathology report, and electronic health data sources from the Kaiser Permanente Southern California health data systems for a cohort of 22,666 members diagnosed with prostate cancer between 2000 and 2010. The magnitude and direction of the differences in date of diagnosis were assessed for each date pairwise comparison. We reviewed 454 medical records to determine reasons for date discrepancies.

RESULTS

Among the date pairwise comparisons, differences in date of diagnosis spanned from 9.6 years earlier to 10 years later than each other. However, the overall median difference ranged from 1 to 16 days, thus suggesting that the vast majority of the date differences were small. Chart review results identified major categories of date discrepancies.

CONCLUSIONS

These data demonstrate variability in date of diagnosis across these data sources. This variability may have implications for epidemiologic estimates or patient identification in research studies using different data sources.