Using Modern Data Science Methods and Advanced Analytics to Improve the Efficiency, Reliability, and Timeliness of Cardiac Surgical Quality Data

Within existing national surgical quality improvement (QI) programs, there are numerous opportunities to improve the efficiency of data flow from the point of collection to the time at which performance-based feedback is provided to stakeholders. Current limitations of the QI data cycle include: (a) reliance on hand

abstraction for data collection; (b) a retrospective and episodic (e.g.: quarterly, bi-annually, etc.) approach to analysis and feedback which creates a time lag from when the hospital’s performance is declining and when it is made aware; (c) small clusters of clinically meaningful poor performance may go of undetected using current

episodic analytic structures. To address the first limitation, modern data science methods (MDSMs) could be used to automate the collection of some, or all, of the variables within surgical QI registries. Full or partial automation of data collection could allow the substantial resources currently committed to manual data

abstraction to be repurposed to support more continuous, proactive engagement in local QI activities. To address the limitations associated with episodic performance evaluation, alternative approaches for analyzing data in more real-time could be applied to provide an early warning of declining performance. The Veterans

Affairs (VA) Surgical Quality Improvement Program (VASQIP) is one of the most successful and longest- standing national clinical registries used for surgical QI and has been the template for a number of similar programs in the private sector. As such, VASQIP represents an excellent model for evaluating alternative

approaches to data collection and analysis that could allow for more efficient data flow through the quality improvement cycle and enhance national surgical QI efforts. The overall goal of this proposal is to evaluate alternative, potentially more efficient strategies that can be readily implemented within the existing

infrastructure of contemporary surgical QI programs and aid in the more efficient flow of data. The specific aims are to: (1) develop and validate MDSMs to use structured and unstructured electronic health record data to automate cardiac VASQIP data collection; (2) compare the risk-adjusted CUSUM (a statistical process

control methodology borrowed from industry) to quarterly observed-to-expected ratios (i.e.: VASQIP’s current approach to assessing performance) for evaluating VA hospital cardiac surgical performance; (3) conduct semi- structured interviews with diverse stakeholder groups to set a national research agenda for expansion and

improvement of surgical QI programs. This mixed-methods proposal will involve observational studies using VASQIP and VA Corporate Data Warehouse data for patients who underwent cardiac surgery at a VA hospital between 2016 and 2020 as well as qualitative interviews with stakeholders who can help to inform future

changes that can improve the data available within VASQIP. This project is important and novel because it will provide real-world, generalizable data that can be used to inform national surgical and non-surgical QI initiatives within VA and the private sector.