Even the most robust technology itself cannot improve healthcare outcomes, as Vanderbilt University Medical center discovered on its way to lowering ventilator-associated pneumonia rates – saving money
During the last decade, Vanderbilt University Medical Center built into its operations the attention to clinical quality that now figures prominently in health reform, using a level of digital documentation and decision support that anyone seeking meaningful use of health IT would love to have.
The problem: All that effort was falling flat where it really counted.
“We had all of the technology anybody’s talking about,” said William Stead, the Nashville-based medical center’s chief strategy and information officer. “We could show that the percentage of the time we did what we wanted to do with decision support went from, say, 10-30 percent up to 40-60 percent – major improvement. But even so, our performance on summative quality measures such as observed to expected mortality were average or actually below average.”
It wasn’t until informatics pros and clinicians figured out how to calculate and graphically present a spectrum of clinical status measures for immediate action that quality measures became worthwhile and patient-care results dramatically improved.
The lessons learned are relevant for hospitals nationwide as they incorporate into clinical settings the quality metrics typical of the federal value-based purchasing and meaningful-use programs now underway.
Most notably, Vanderbilt found that these standard quality metrics by themselves won’t give front-line clinicians the focused information that’s useful in improving clinical results and increasing healthcare value. Measuring and reporting constitute only the first half of the quality-improvement equation. What’s further needed is a sound way to “force” a set of expeditious responses to feedback that, if done well, leads to improved metrics while lowering clinical costs and preventing inpatient deaths.
The method is in place to prevent certain types of infections in high-risk areas, and Vanderbilt is expanding that to target other patient threats in the hospital environment.
Homing in on a serious risk
Few hospital situations are more life-threatening than a critically ill patient, put on a ventilator in an intensive-care unit, who develops pneumonia. Mortality rates discussed in medical literature vary from 20 percent to more than 50 percent.
With six adult ICUs for burn care, trauma, neurological, cardiovascular, post-surgical and general critical care – combining for about 160 beds – Vanderbilt annually had many hundreds of patients who were spending, collectively, tens of thousands of hours on ventilators in settings that were nearly always full.
Ventilator-associated pneumonia (VAP) rates in the mid-2000s at Vanderbilt were running in some cases two to three times above the national benchmark for inpatient populations, said Devin Carr, administrative director of the surgery and trauma patient care center. In 2004, for example, the rate of VAP per 1,000 ventilator days was 29.3.
In the wake of a national campaign by the Institute for Healthcare Improvement to reduce VAP through consistent ventilator management, Vanderbilt organized a multidisciplinary team of physicians, nurses, pharmacists, respiratory therapists and other clinicians in 2007 to develop a “bundle” of care elements proven to reduce the incidence of pneumonia. Then came the challenge of keeping track of how clinicians were juggling the seven elements of necessary care.
Clinical IT provided the electronic means to document what nurses and others did for patients and assemble quality metrics for such actions as elevating a patient’s head to prevent aspirating back into the lungs, appropriate oral care to prevent a buildup of bacteria, and frequently evaluating whether the patient was ready to breathe independently and get off the ventilator.
But for all the advances in IT at Vanderbilt, the VAP improvement team identified some significant shortcomings, Carr said. For example, various ICU medical teams documented differently in separate areas of the IT system according to how they worked, which complicated the task of drawing out and standardizing the metrics in ways that could be brought together and posted in a timely manner.
There also was no alerting function in the documentation system itself.
Due to the broad range of people and clinical roles that come into daily contact with a patient, Stead said, alerts affecting all those roles would be overwhelming in an already-intense environment.
The colors of compliance
The computer-assisted remedy involves taking status information on all care elements from the several different reporting systems, coalescing it in one place and presenting it on each ICU patient in a general way at the bedside for clinicians to see at a glance how ongoing care compared with established targets.
A ventilator management “dashboard” grid, operating as a screen-saver, loads performance metrics on each element – for example, oral care every four hours at a minimum – and displays whether it’s happening as expected, said Thomas Talbot, chief hospital epidemiologist. A box displays green, yellow or red depending on whether an element is being taken care of, is approaching noncompliance or is out of compliance. So after three hours, an oral-care box turning yellow can give caregivers an hour’s notice to attend to that task before the risk of pneumonia rises, he said.
That “forcing function” was the breakthrough in turning metrics reporting into action, closing the loop and allowing the ICUs to gain tight control of the VAP prevention process, Stead said. Instead of just supplying information to the ICU, the dashboard highlights priorities for care-givers – posting the status of each patient’s care in almost real-time and showing medical professionals what needs to be done while they have time to correct. In other words: If it’s yellow, turn it green before it turns red.
Once the process took hold, performance on each of the elements in the ventilator management bundle quickly rose, and the various tasks are being done 80 percent to 90 percent of the time on any given day. But the desired outcomes of all this effort, in terms of mortality and clinical cost, weren’t budging.
Further analysis of the quality metrics, Talbot said, tell a more troubling story: Only about 22 percent of patients are consistently receiving all the recommended care at the same time. An ICU could meet one metric during one hour, a different metric another hour and “still be green most of the day – but if they’re not all green at the same time, then the patient is not getting all the right processes,” Talbot said.
That demonstrates the false sense of achievement that individual quality metrics can engender, said Lee Parmley, director of the critical care division. “If we had 10 patients, we could miss one element on each of them and still be 90 percent compliant. What that would mean is that we had no patients that we had everything right on. What really works is you have to have all bundled elements (performed) on a patient in order for it to be effective.”
Getting it all right at once
Vanderbilt informatics professionals, led by Jack Starmer, went up another level with the reporting and devised a single measure of compliance with every element of a patient’s care plan. A daily report to clinical managers, called the “Z100 score,” tracks in five-minute increments whether all the metrics are in compliance on a given patient. With that added emphasis and the data to power it, pneumonia levels began falling as the Z100 score went from 23 percent in August 2007 to 65 percent in August 2008.
“Once that (score) crossed 60 percent, we had about a 50 percent drop in ventilator-associated pneumonia across our entire operation in six months,” Stead said. The 20-percent rate prior to the program fell to 11.3 percent in fiscal 2009 (ended June 30), when 140 cases were recorded during 12,400 ventilator days. The Z100 score hit 88 percent.
In human and financial terms, Vanderbilt calculated that 108 VAP cases were prevented in fiscal 2009, saving 16 lives and $4.3 million in estimated costs of battling the pneumonia outbreaks, “just based on the cost of each episode and the frequency with which people died if they get it,” Stead said. The calculation included 1,055 fewer hospital days and 431 fewer ICU days.
The lesson Stead shared: “A whole bunch of point improvements – getting one part of what we do for the patient right, or getting it better – actually isn’t enough to get the end-to-end improvement in quality we need to reduce complications, reduce morbidity and mortality, reduce transition problems enough to bring down the cost.”
Vanderbilt has adapted that lesson to combat infections involving catheters. During the first half of 2011, additional metrics were posted on a dashboard that tracks how long a patient has a catheter, prompting the physician to evaluate whether it’s still needed, Talbot said. There’s also a metric on daily catheter maintenance, such as making sure dressings have been replaced. Other care processes are going through the same type of metric application, such as for fall risk and pressure ulcers, he added.
Tapping meaningful use
Meaningful use criteria, including clinical documentation and decision support, are “a necessary precursor” to offering such a clinically precise and vigilant management tool, Stead said. “The bundle is represented as order sets within the computer-based order entry system. It helps people remember what to order. What the dashboard screen-saver does is actually see whether we’re taking the individual steps, and that is derived from documentation in the nurse charting system, it’s derived from information in the respiratory therapy system, it’s derived from information in the order system.”
But once hospitals incorporate meaningful use technology, they may have much more work to do in terms of the way information can be used in the system, he said. Most hospital IT systems are built around transaction processing, while Vanderbilt’s uses informatics approaches around systems engineering, which facilitates cognitive support and rapid-cycle change.
The dashboard is “a statistical calculation of what’s going on,” he said. “It’s not somebody entering something into a box that we then display. We look at several things you’re doing and, from that, calculate the probability that you’re doing what you need to do or not. … It’s a more robust use of IT, and so as we get people to do meaningful use, we’re then going to have to move them on to these other techniques also.”
Such approaches have the benefit of changing the overall management environment.
“Every year our ventilator results have simply gotten better even though they’re no longer a focus of attention as we focus on other things,” Stead said.
From 18 pneumonia cases per 1,000 ventilator days in calendar 2007, the rate fell to 14 in 2008 and 11 in 2009. But thereafter, it continued falling to 7.5 in 2010 and, in the first three months of 2011, a mere three per 1,000 days.
“What we’re really talking about is how to get people to look at care as a system and use controls and feedback to help you adapt and do the right thing,” Stead said. “If you do that, then you fix the system in a way that will fix many quality indicators.”
