Dashboards to Improve Extractability of Cardiovascular Indicators in a Learning Health Care System: Mixed Methods Study

Abstract Background Cardiovascular risk management (CVRM) guidelines have been developed to evaluate and manage all patients at higher cardiovascular risk, being either symptomatic or still asymptomatic. Although these guidelines have long existed, adherence varies. A learning health care system (LHS) could address adherence by continuously analyzing routine care data to inform and improve health care practice. Dashboards may be used to inform clinicians on the care provided and potentially improve structured registration of CVRM indicators in electronic health records (EHRs). Objective We evaluated whether the implementation of dashboards in our LHS led to changes in the structured registration of cardiovascular indicators in patients at increased risk of cardiovascular disease (CVD). Methods In our mixed methods study, patients who visited the University Medical Center Utrecht between January 2022 and November 2023, the period during which the dashboard was implemented, were included. We assessed the extractability of the CVRM indicators (ie, BMI, blood pressure, smoking status, medical CVD history, lipid levels, glycated hemoglobin, hemoglobin, and the estimated glomerular filtration rate), stratified by department. We compared the extractability of the indicators with the extractability before the Utrecht Cardiovascular Cohort-Cardiovascular Risk Management (UCC-CVRM) LHS was initialized and with the period during which the UCC-CVRM was protocolized, but without the use of dashboards. To explain our quantitative findings and to gain a deeper understanding of how the dashboards were viewed and perceived, we conducted semistructured interviews with clinicians and analyzed these thematically. Results The extractability of CVRM indicators among 8941 first hospital visits remained low and stable during the period in which the dashboards were used. Overall, hemoglobin (5344/8941, 59.8%) and estimated glomerular filtration rate (5682/8941, 63.5%) were most often extractable, and patients’ CVD history (1946/8941, 21.4%) and smoking status (2543/8941, 28.4%) were the least extractable. Compared to the protocolized UCC-CVRM, indicators were up to 45% less extractable, meaning that CVRM indicators were less often registered in structured fields of the EHR. Interviews with clinicians (N=5) revealed that the low extractability could be attributed to unclear responsibility for CVRM, lack of harmonized agreements for registration in EHRs, perceived challenges related to the EHR system (eg, some structured fields were not easily accessible), time constraints, and habits (eg, maintaining habitual ways of working that are perceived to best suit their workflow). Conclusions Dashboards did not improve the registration of CVRM indicators in structured fields of the EHR. This was explained by perceived organizational, technical, and operational issues, such as unclear responsibility for CVRM care, suboptimal technical knowledge and limitations of the EHR system, and time constraints. Our findings provide guidance on what aspects to consider for the extractability of CVRM indicators to be improved, which will be beneficial for both clinical practice and scientific research using real-world data.