St. Luke's Hospital in Kansas City, Mo., employs a web-based technology* that helps patients understand their risks with an upcoming cath lab procedure. The technology executes the American College of Cardiology's (ACC's) multivariable risk prediction models with patient-specific data so that individualized estimates of outcomes can be generated within the routine flow of clinical care and used to support shared medical decision making.
We currently use ACC models to estimate mortality (so that clinicians are keenly aware of patients' risks), bleeding (so that appropriate bleeding avoidance strategies, such as radial approaches and bivalirudin use, can be selected), and one-year target vessel revascularization rates with bare-metal and drug-eluting stents (DES) (so that patients can compare the benefits of DES with the need to take prolonged dual-antiplatelet therapy).
To ensure that patients receive a validated estimate of their risks, we changed our informed consent process from a generic, hospital-wide template to a more educational, individualized consent that includes patients' risks within the consent form itself (Circ Cardiovasc Qual Outcomes 2008;1:21-28). Nurses in the cath lab holding area generate and print the form, which takes less than five minutes. The doctor then obtains the consent and uses the risks to plan the appropriate procedure—including the vascular approach, anticoagulant strategy and a patient's stent preferences.
Since implementing this program, we were able to document improvements in patient satisfaction with the consent process, an improved sense of engagement in their medical decision making and reduced anxiety, as well as lower rates of DES use in patients at lower risk for restenosis. St. Luke's also has documented a marked reduction in bleeding from 5.5 percent in 2007 (about twice the national average) to 1.8 percent in 2010 (about half the national average). Congruent with these changes in bleeding, we were able to reverse the "risk-treatment paradox" (JAMA 2010;303:2156-64) in which those with the most benefit from bleeding avoidance treatments, such as bivalirudin, are now most likely to get it.
An additional important advantage of prospectively applying the ACC's risk models is that the risks of a major complication, like bleeding, are a powerful justification to admit an outpatient undergoing PCI to the hospital for closer observation. As such, software-generated risks have become our core strategy for triaging outpatients to overnight admission and serves as our foundation for defending recovery audit contractor (RAC) reviews.
Since piloting the program at our institution, it has now been deployed at 10 other U.S. centers, including Kaiser San Francisco, Mayo Clinic in Rochester, Minn., Yale-New Haven Hospital in New Haven, Conn., and Integris in Oklahoma City. After implementation, each center, interestingly, has focused on different aspects of the risk models to improve its quality, safety and cost-effectiveness.
St. Luke's is now embarking on a program to execute Medicare's 30-day readmission models for heart failure, acute MI and PCI, all of which have important economic consequences for our hospital. By identifying those at highest risk, we hope to be able to tailor the intensity of our discharge and follow-up planning to the needs of the individual patient. Collectively, we believe these efforts have enabled us to markedly improve our personalization of healthcare and work toward achieving the Institute of Medicine's goals for better quality of care and reduced errors.
*Disclosure: Dr. Spertus is a developer of ePRISM software, and holds an equity position in Health Outcomes Sciences of Kansas City, Mo., that is commercializing the program.