Background: Clinical registries are the backbone of operational, quality improvement and research initiatives across cardiac centers, nationally.  Many of the data elements contained in these registry databases come directly from the electronic health record and other diagnostic and imaging applications.  Traditionally, the burden for data entry into these registry systems fell to either the clinician responsible for the care of the patient or a team of data entry specialists who were responsible for a multitude of data elements.  Because of the nature of this data flow, the potential exists for data entry error as well as delays and inconsistencies across those entering data.  To minimize all three of these factors, data automation could be developed that would greatly reduce the amount of time spent entering, validating, and fixing data issues while ensuring that the definitions behind the variables remained consistent and accurate.  While our center participates in many registries, the focus for this initial effort was on the IMPACT (Diagnostic and Interventional Cardiac Catherization) and STS (Society of Thoracic Surgeons, specifically Anesthesia and Perfusion components) registries.  

Method: Beginning by interviewing the subject matter experts behind each registry system component, we were able to set definitions and sources behind each data variable that would be in the automation pipeline.  Once the extraction program for the data was set, it would be moved to a staging area which would be used to denote if that record was to be loaded or reloaded based on new or updated information from the previous day.  Upon load the clinical data specialists responsible for the accuracy of the registries would now be able to conduct regular data validation exercises as well as fill in any data not available to be automated (non-discrete).  

Results: The IMPACT and STS Registries were the first 2 in which this automation routine was implemented.  IMPACT included 389 fields, STS Anesthesia encompassed 182 fields and STS Perfusion contained 56 fields which became automated for a total of 627 data elements across a daily average of 15 cases.  On average, the time spent by each individual responsible for entering this data was about 15 minutes.  The time effort involved in developing the programming components of this process totaled roughly 130 hours across all 3 registries. The average time savings for the entire automation program results in 975 person hours (~ 0.5 FTE) regained each year with a slightly lower number of hours in the first year of inception offset by the initial work effort involved in standing up the solution.  Data submission error rates sharply declined from 13.6% pre-automation to 2.2% for IMPACT.  For Anesthesia, the missing data rate decreased from 13.6% of the cases to 1% after automation.  For Perfusion, a drop was seen from 43% of cases missing elements to 3.1%.

Conclusions: The Cardiac Registry Automation program has been a resounding success, providing for the reclamation of thousands of Cardiac person-hours along with improved data quality and the near-real-time entry of data.