Dimensions of Data Quality

Clinical research data quality is multidimensional. Different frameworks use different terminology, but several dimensions appear consistently in clinical data management and health data quality literature. Understanding these dimensions helps data managers design checks that go beyond simple missing value counts. Accuracy refers to whether data correctly represent the source observation or real-world event. If a participant's laboratory report shows hemoglobin of 8.5 g/dL but the database records 85 g/dL, the value is inaccurate. Completeness refers to whether expected data are present. If the primary outcome is missing for many participants, the dataset may be incomplete even if other variables are well populated. Validity refers to whether data conform to defined formats, ranges, and allowable values. A visit date entered as free text rather than a date format may be invalid. Consistency refers to whether related values agree logically. A discharge date before admission date is inconsistent. Timeliness refers to whether data are entered, reviewed, corrected, and available within required timeframes. Timely data are particularly important for safety monitoring, adaptive study decisions, and operational dashboards. Uniqueness refers to the absence of unintended duplicates. Duplicate participant records can distort enrollment counts and analyses. Integrity refers to protection from unauthorized, undocumented, or inappropriate change. Integrity depends on audit trails, access control, source traceability, and controlled correction procedures. Another important dimension is interpretability. Data are not useful if future users cannot understand variable definitions, coding, units, collection methods, or missing value meanings. For example, a variable coded as `1`, `2`, and `3` is not interpretable unless the data dictionary explains what those codes mean. Interpretability connects data quality to metadata, FAIR principles, and reproducibility. No single quality dimension is sufficient. A dataset may be complete but invalid if values are outside plausible ranges. It may be valid but inaccurate if values were copied from the wrong source document. It may be accurate and complete but not timely enough for interim safety review. Data managers should therefore design a portfolio of checks that assess multiple dimensions. **Table 5.1: Data Quality Dimensions in Clinical Research**