Background: The increasing use of routinely collected data in electronic health record (EHR) systems for business analytics, quality improvement and research requires an extraction process fit for purpose. Little is known about the quality of EHR data extracts. We examined the accuracy of three data extraction tools (DETs) with two EHR systems in Australia.
Methods: The hardware, software environment and extraction instructions were kept the same for the extraction of relevant demographic and clinical data for all active patients with diabetes. The counts of identified patients and their demographic and clinical information were compared by EHR and DET.
Results: The DETs identified different numbers of diabetics and measures of quality of care under the same conditions.
Discussion: Current DETs are not reliable and potentially unsafe. Proprietary EHRs and DETs must support transparency and independent testing with standardised queries. Quality control within an appropriate policy and legislative environment is essential.