Use of an automated predictive model to pinpoint hospitalized patients at the highest risk for clinical deterioration reduced mortality and intensive care unit admissions, and shortened hospital stays, investigators report in a study published online today in the New England Journal of Medicine.
However, the study's lead author was quick to point out that the positive findings were not solely due to scrapping manual risk score calculations in favor of an automated method of identifying high-risk patients, but also relied heavily on effective implementation.
"If you pay attention to doing the workflows and implementing correctly, then you can save lives. If you think you're going to save lives just because you have some fancy schmancy equation, dream on. It really requires the implementation," said Gabriel J. Escobar, MD, regional director for hospital operations research, whose work measures the performance of Kaiser Permanente’s network of hospitals.
For the northern California hospitals that took part in the research study, that meant:
Tapping professional implementation teams that ensured staff were trained in using the model
Identifying internal champions at each hospital — typically a nurse lead and physician lead — to keep staff on target
Standardizing rapid-response team makeup
Devoting 4.2 full-time equivalent per hospital, per year, for nurses dedicated to the rapid-response team 24/7
"There are a number of papers that show that these Early Warning Scores can have good predictive performance but less evidence that they impact clinical outcomes. Speaking as a non-clinician, my sense is that even if a score accurately predicts who is at risk, it is not always clear what should be done to mitigate that risk," Benjamin A. Goldstein, PhD, an associate professor in the Department of Biostatistics & Bioinformatics at Duke Clinical Research Institute, told Medscape Medical News. "This is the big next question: Now that we know who is at risk, what should be done about it?," continued Goldstein, who was not involved in the research.
Automatic, Early Red Flags
Kaiser Permanente first piloted its Advance Alert Monitor (AAM) program at two of its northern California hospitals and, after encouraging results, Escobar and colleagues introduced it via a staggered rollout at another 19 hospitals between August 1, 2016, and February 28, 2019.
To assess the effect of the program, the team analyzed outcomes among 15,487 hospitalizations that occurred after implementation and involved patients 18 years or older admitted to a general medical-surgical ward or stepdown ward. The team compared those outcomes with outcomes among 28,462 hospitalizations involving patients admitted to any of the study hospitals in the 12 months prior to AAM introduction at that facility.
The automated system scanned patient data, including vital signs, laboratory test results, neurologic status, severity of illness and longitudinal indexes of coexisting conditions, care directives, and health services indicators, such as length of stay, to generate hourly scores and automatically triggered alerts for patients at the highest risk for clinical deterioration. An AAM score of 5 was high enough to trigger an alert and equaled an estimated 8% or greater risk for clinical deterioration within the next 12 hours; on average, one new alert was triggered per day per 35 patients.
Thirty-day mortality was 16% lower in patients after implementation compared with the comparator group (adjusted relative risk, 0.84; 95% confidence interval, 0.78 - 0.90; P < .001).
Although, COVID-19 was not the focus of the current study, preliminary data suggests that patients infected with COVID-19 are twice as likely to trigger an alert as patients without COVID infections, Escobar told Medscape Medical News.
Alerts Monitored Remotely
To sidestep alert fatigue, trained nurses monitored the AAM alerts remotely. If an alert reached the threshold, those nurses reviewed the patient's chart and flagged the on-site rapid-response nurse who then contacted the patient's physician. Because AAM alerts flagged emerging risk for deterioration, care teams had a 12-hour window in which to act, which Escobar says was essential.
"I am a physician and when you are working in the hospital, you are so busy. One of the big selling points about this program was we could tell people 'You know when you get an alert, it's not like a Code Blue. You can finish that sandwich. Or if you're writing a progress note, you can finish that note. You have time.' We deliberately built the model so that we would have that time," he told Medscape Medical News.
Using Technology to Protect Patients
But for clinical teams accustomed to sprinting in response to a Code Blue, it was initially bewildering to enter patient rooms in which nothing appeared to be happening. To navigate that new terrain, the team scripted nuanced language that physicians and nurses could use when they approached the patient.
"Our marketing people told us that patients don't like the idea that a computer is watching them, but they love the idea that technology is being used to protect them. So, our scripts were somewhat vague, and they basically said we've been monitoring your vital signs and your labs. And because of this we'd like to assess the situation," Escobar told Medscape Medical News.
Escobar expects that an early warning system will become standard of care at all Kaiser Permanente hospitals within the next 12 months, though some regions may use Epic's model.
Although the system appears to be working in the Kaiser system, Goldstein, the Duke associate professor, told Medscape Medical News that hospitals outside of the Kaiser network face a number of logistical hurdles to adopting the AAM system, including technical, legal, clinical, personnel, and the statistical challenges of how well the score would perform in a new environment.
The authors and Goldstein have disclosed no relevant financial relationships. The study was funded by the Gordon and Betty Moore Foundation, the Sidney Garfield Memorial Fund, the Agency for Healthcare Research and Quality, the Permanente Medical Group, Kaiser Foundation Hospitals, and the National Institutes of Health.
N Engl J Med. Published online November 11, 2020. Abstract
Diedtra Henderson is a freelance journalist based in Washington, DC. She has written for the Boston Globe, The Associated Press, the Denver Post, the Seattle Times, and the Miami Herald.
For more news, follow Medscape on Facebook, Twitter, Instagram, and YouTube.
Medscape Medical News © 2020
Cite this: Inpatient Mortality Dropped With Model ID'ing High-Risk Patients - Medscape - Nov 11, 2020.
Comments