PITTSBURGH -- A new mathematical model of sepsis may help physicians make better clinical decisions, researchers here hope.
PITTSBURGH, June 14 -- A new mathematical model of sepsis may help physicians make better clinical decisions, researchers here hope.
In a proof-of-concept study, the model accurately predicted the outcomes in a cohort of 1,888 sepsis patients in 28 academic and community hospitals in Pennsylvania, Connecticut, Michigan, and Tennessee, according to Mark Roberts, M.D., of the University of Pittsburgh.
The key factor in making the model so accurate was including information on the duration of the illness, something that has been lacking in previous models of sepsis, Dr. Roberts and colleagues reported online in the journal Critical Care.
The study supports the "clinical intuition that the history of the disease in an individual matters," the researchers said.
Previous models used a constant rate of disease progression, they said, whereas this model assumes that progression will vary from patient to patient, based on recent disease history.
The model might be used, Dr. Roberts said, to develop treatment protocols, help design clinical trials, or to evaluate the cost-effectiveness of various therapies.
The model was based on data from the Genetic and Inflammatory Markers of Sepsis (GenIMS) Study, a multicenter cohort study of subjects with community-acquired pneumonia (CAP) at risk for severe sepsis.
Most patients in the cohort were older, with a mean age of 67.7 years, 80.7% were white, and 48% were women, the researchers said. They were also relatively ill at baseline, with an average:
Almost 16% required intensive care, 6.5% died within 30 days, 31.2% developed severe sepsis, and, of those, 26.9% died within 90 days.
The model -- an empirically based, Monte Carlo model -- was required to match computer-generated patients to those in the actual cohort on the basis of three factors: component or total SOFA score, the direction of change in SOFA score, and duration of illness.
When that was done, Dr. Roberts and colleagues found, the model predicted 30-day discharge and death rates that were statistically indistinguishable to those seen in the real patients.
On the other hand, the researchers found, when the model did not include information on the duration of illness the results were significantly different (at P<0.001) from the real cohort.
"This observation reinforces a clinical intuition that a patient with a given level of organ dysfunction on day three is very different from a patient with that same level of organ dysfunction on day 15," the researchers said.