Eliminating Low-Value Care in Virginia: Six Large Health Systems

Low-value care is defined as patient care that provides no net benefit to patients in specific clinical scenarios, and can cause patient harm. Prior research has documented high-rates of low-value care in Virginia; this work has helped to inspire a Virginia government-sponsored quality improvement initiative to reduce low-value care. Funded by an Arnold Ventures grant, six large health systems in Virginia volunteered to partner with the Virginia Center for Health Innovation (VCHI) to reduce use of nine low-value health services (three preoperative testing measures, two cardiac screening measures, one diagnostic eye imaging measure, one low-back pain opioid measure, one low-back pain imaging measure and one peripherally inserted central catheter [PICC] measure). These health systems include nearly 7000 clinicians practicing across more than 1000 sites.

VCHI is implementing a nonrandomized physician peer-comparison feedback quality improvement intervention to reduce use of nine low-value services. Modeling will be used to identify and use propensity score matching to match six intervention health systems to six comparable control health systems. VCHI will provide education, quality improvement training and financial resources to each site, and VCHI will use the Milliman MedInsight Health Waste Calculator to create the peer comparison reports using the Virginia All Payer Claims Database (APCD). VCHI will use additional measures from The Agency for Healthcare Research and Quality (AHRQ). Additionally, VCHI will use AHRQ data to attribute physicians and health care facilities to health systems.

The primary purpose of the initiative is to improve quality of care for Virginia residents and this initiative is not being done for research purposes. Nevertheless, University of California, Los Angeles (UCLA) plans to rigorously study and publish the impact of this intervention across the state of Virginia, which is why the UCLA team pre-registered the initiative. The UCLA team will use the Virginia APCD to evaluate the impact of the intervention. Please note: the APCD has a 1-year time-lag of data collection and is a dynamic database, meaning that its population of enrollees changes from year to year. This intervention was initially designed as a randomized step-wedge intervention; the intervention was delayed by the COVID-19 pandemic and began in September 2020 for all intervention groups. The intervention period was extended through December 2022. As a result, the initial design was modified.