For more than a decade, the University of California’s five health centers have pooled data from their cardiac surgery procedures to share best practices and markedly improve patient care.
Advanced analytics on 200 data elements from each patient have provided significant insights into local and systemwide performance, allowing the University of California Cardiac Surgery Consortium (UCCSC) to build standardized and sustainable quality improvements. Importantly, the metrics are vetted, audited and aligned to those of the Society of Thoracic Surgeons (STS), a national database.
“We can benchmark ourselves against national data, and that allows us to be sure that we are among the best quality among elite institutions performing cardiac surgery, not only in California and in our local markets, but nationally,” said , chief of the division of cardiac surgery and a distinguished professor at the .
The consortium’s health centers perform about 4,000 cardiac surgery operations annually. Clinical outcomes in isolated coronary artery bypass grafting (CABG), the most common procedure, showed improved early extubation, reduced blood utilization and reduced readmissions between 2015 and 2023.
Overall, they resulted in 132 bed days saved and a financial margin improvement of about $15 million.
Consortium data
Dr. Shemin founded UCCSC in 2012, bringing together University of California health centers in San Diego, Irvine, Los Angeles, Davis and San Francisco. The cardiac surgery chiefs and database nurse coordinators confer biweekly and formally meet annually. In addition to increasing the volume of procedures and reducing their variability, the consortium aims to optimize costs and explore contracting opportunities, such as joint purchasing.
Each health center uploads its clinical data every quarter. Biome Analytics, a cardiovascular technology firm, then performs multivariable regression analyses on single site and systemwide outcomes.
The health centers’ 4,880 CABG procedures between 2015 and 2023 demonstrated several improvement markers:
- Any blood product utilization decreased by 8.26%
- Rate of early extubation (less than six hours) increased by 22%
- Average initial ventilator hours decreased by 1.5 hours
- Median length of stay in the ICU decreased by 0.38 days
- 30-day readmissions decreased by 3.96%
- New onset atrial fibrillation, at 25.4%, was below STS benchmark
Associated cost savings were also analyzed. For example, early extubation resulted in a margin improvement of about $6.7 million and decreased ventilation in about $3.6 million. The 132 bed days saved equaled a cost savings of about $486,000.
“It's amazing that through collaboration and data analysis we have been able to improve patients’ lives as well as optimize the health systems’ economic efficiency,” said Nancy Satou, RN, director of informatics & database management in the division of cardiac surgery at UCLA Health.
In a further bid to reduce complications and readmissions, UCLA Health uses advanced technologies to monitor patients at home. These include wearables and tablets with Bluetooth to assess vital signs such as blood pressure, heart rhythm and blood oxygen levels. Adverse trends are addressed quickly with a physician consult to help prevent later emergency room visits or hospital readmissions.
“UCLA Health is the only one in the consortium that does this,” said Dr. Shemin, who is also co-director of the cardiovascular center at the David Geffen School of Medicine at UCLA. “We've been able to reduce our readmission rates from the 20% range down to single digits. Patients’ satisfaction levels go up and they feel cared for.
“We're just at the beginning of the revolution of wearables and monitoring technology, giving people who are willing to engage the feedback they need.”
AI and machine learning models
In addition to assessing clinical outcomes, the consortium has begun using machine learning models on preoperative factors to predict resource utilization and clinical outcomes.
A primarily examined length of stay (LOS), with 30-day mortality, acute kidney injury and reoperation, among others, as secondary endpoints. In addition to its predictive capabilities, the machine learning models also identified several risk factors associated with increased resource use.
Future AI applications will require updating the database more frequently than its current quarterly schedule. The consortium’s next challenge is to update monthly, and eventually, in real time.
Further down the line, the consortium aims to measure outcomes beyond its current 30 days. It’s a difficult and expensive proposition that would require linking its robust database with other data sets, such as Medicare, to provide yearly, five-year and 10-year follow-ups.
“Our goal is to minimize complications that drive longer hospital stays as well as having highly efficacious operations,” said Dr. Shemin. “And we're looking at the long-term efficacy of what we do.”
