The utility of general purpose versus specialty clinical databases for research: warfarin dose estimation from extracted clinical variables

J Biomed Inform. 2010 Oct;43(5):747-51. doi: 10.1016/j.jbi.2010.03.014. Epub 2010 Apr 2.


There is debate about the utility of clinical data warehouses for research. Using a clinical warfarin dosing algorithm derived from research-quality data, we evaluated the data quality of both a general-purpose database and a coagulation-specific database. We evaluated the functional utility of these repositories by using data extracted from them to predict warfarin dose. We reasoned that high-quality clinical data would predict doses nearly as accurately as research data, while poor-quality clinical data would predict doses less accurately. We evaluated the Mean Absolute Error (MAE) in predicted weekly dose as a metric of data quality. The MAE was comparable between the clinical gold standard (10.1mg/wk) and the specialty database (10.4 mg/wk), but the MAE for the clinical warehouse was 40% greater (14.1mg/wk). Our results indicate that the research utility of clinical data collected in focused clinical settings is greater than that of data collected during general-purpose clinical care.

MeSH terms

  • Algorithms
  • Anticoagulants / administration & dosage*
  • Biomedical Research
  • Biopharmaceutics / methods*
  • Cohort Studies
  • Databases, Factual*
  • Drug Dosage Calculations*
  • Humans
  • Medical Informatics / methods*
  • Warfarin / administration & dosage*


  • Anticoagulants
  • Warfarin