Workflow Differences Affect Data Accuracy in Oncologic EHRs: A First Step Toward Detangling the Diagnosis Data Babel.

Area of Science:

• Health Informatics
• Clinical Data Management
• Oncology

Background:

• Accurate structured diagnosis (DX) data is essential for reusing clinical information.
• Existing structured DX data often lacks accuracy, hindering its utility.
• Previous studies suggest workflow variations in cancer diagnosis entry, but their impact on data quality remains unclear.

Purpose of the Study:

• To investigate the relationship between workflow-describing variables and the quality of diagnosis data.
• To determine if specific clinical workflows and personnel influence the accuracy of structured diagnosis entries.

Main Methods:

• Extracted diagnosis data from electronic health records (EHRs) for brain neoplasm patients.
• Utilized logistic regression to predict the odds of accurate, inaccurate, and suboptimal DX entry based on workflow variables.
• Selected variables based on their correlation with outcome variables.

Main Results:

• Workflow and personnel variables significantly predicted diagnosis data quality.
• Diagnosis entries outside oncology departments were more accurate (2.89x odds).
• Outpatient locations had lower inaccuracy (98% fewer odds) but higher suboptimal entries (458x odds).
• Physician assistants entering diagnoses had lower suboptimal entries (85% fewer odds) compared to physicians.

Conclusions:

• Clinical workflow and personnel directly affect the quality of EHR data.
• The need for structured diagnosis data recording varies by clinical workflow and information requirements.
• Clinicians and researchers must account for this heterogeneity when analyzing oncologic EHR data.