A research team, composed of engineers from the University of Maryland (UMD) and physicians at Massachusetts General Hospital/Harvard Medical School has developed a predictive model that can warn of sudden blood pressure drops in ICU patients before they
A research team, composed of engineers from the University of Maryland (UMD) and physicians at Massachusetts General Hospital/Harvard Medical School, has found that patients in intensive care often experience dangerous drops in blood pressure that are not quickly corrected by clinical staff; and the team has developed a predictive model that can warn of such “hypotensive’ incidents before these occur.
It long has been known that patients in the intensive care unit (ICU) experience incidents of low blood pressure that can be harmful. During these occurrences blood supply to a patient’s brain and other organs is reduced. ICUs have protocols for regulating patient blood pressure which include alarms that alert ICU staff of the need to administer drugs known as vasopressors that raise the patient’s blood pressure.
“[Our study] raises the possibility that many of the [low blood pressure] episodes were preventable via more vigilant clinical interventions, including vasopressor dose increases immediately upon the onset of the episodes, or dose increases before the onset (to prevent the episode of hypotension altogether),” wrote corresponding author Jin-Oh Hahn, Ph.D., an assistant professor of mechanical engineering at UMD, and his colleagues from UMD and from Massachusetts General Hospital/Harvard Medical School.
Recently published in the journal Scientific Reports, of the Nature Publishing Group, their findings are the latest of a number of studies indicating that even with monitoring systems and treatment protocols in place, ICU patients experience more and longer episodes of low blood pressure (hypotension) than expected or desirable.
“We were expecting that our study would show the problem is that clinical staff are not adequately following the treatment guidelines [for responding to patients’ hypotensive episodes]. However, we found that it is really a problem of staff not being able to give enough attention to the patients,” Hahn said.
Their innovation was to focus on hypotension during ongoing vasopressor infusion. When analyzed, most of those hypotensive episodes appeared to be preventable. The episodes were also predictable, using a team-developed statistical model. When the model was run using historical patient blood pressure data, it gave advanced warning (12 minutes average) for 99.6 percent of occurring hypotensive events [blood pressures below 60 mean arterial pressure (MAP)]. In real time (prospective) testing, their model predicted 100 percent of 26 episodes with a median advance warning of 22 minutes before the episodes occurred.
“It is a challenge to provide consistent and optimal care for critically ill, unstable patients for hours or days at a time,” said Andrew Reisner, M.D. of the Massachusetts General Department of Emergency Medicine and associate professor of Emergency Medicine at Harvard Medical School. “Predictive models and other computerized intelligence can enable a new generation of healthcare with greater precision and consistency than ever before possible.”
The collaborative team is now working to develop clinical trials for testing the clinical efficacy of their predictive model.