Validation for Accuracy of Cancer Diagnosis in Electronic Medical Records Using a Text Mining Method

To validate the accuracy of data in electronic medical record, we compared cancer diagnosis and key words in pathologic reports of cancer patients in a tertiary hospital, using text mining method. We investigated in fourteen kinds of cancers that had highest incidence rates in Korea. Approximately two-third (71.0%) of total patients had right match in cancer diagnosis with pathologic report. The ratio of concurrence was the highest (86.3%) in thyroid cancer patients, however, the ratio was the lowest (49.9%) in liver cancer patients. To prevent the errors in data input, a systematic alarm and feedback to clinicians should be required.