Importance of Data Quality Improvement and Assessment in Anatomic Pathology

Abstract

Many industries have developed standards for measuring and classifying data quality. In contrast, data quality in anatomic pathology (AP) is not well defined or consistent. Modern data capture tools that result in collection of structured data elements are not employed often enough by laboratory information systems (LISes) in AP. As well, workflows often exist which allow introduction of inconsistencies, thereby compromising data quality. For example, in many LISes, part types and descriptions are captured in structured text fields upon initial case accession. Part descriptions as initially accessioned may be inserted in the final report; however, they are inserted as editable free text, allowing for potential discrepancies between part descriptions in the final diagnosis field and what was originally accessioned. Rare reports of major accessioning errors prompted investigation of frequency and severity of potential part description discrepancies. In a preliminary study, 852 surgical inside cases accessioned into Sunquest-CoPathPlus over a randomly chosen 3-day period were evaluated. Each part description present in the structured data field post sign-out was compared to part descriptions present in the final diagnosis section of the signed-out report. Part descriptions entered at the time of accessioning captured in an HL7 message crossing to the EMR were also available. We found 11% of cases contained part descriptions that had been edited in the final diagnosis field of the report without any change in meaning, while for approximately 6% of cases (55/852), those edits added additional information compared to the original description at accession (most commonly proceduretype addition). Major discrepancies (eg, discrepant laterality) were not detected in this limited dataset. Discrepancies in part type reporting within a pathology report not only make for a confusing report, but also potentially distribute inaccurate information to the patient record. Tools for efficiently assessing data quality and classifying data discrepancies within AP are currently lacking.