Can a computerized physician-alert system lower the incidence of deep vein thrombosis or pulmonary embolism among hospitalized patients?
Yes, by about 41%, according to a study involving 2,506 randomly selected patients conducted at Brigham and Women's Hospital, Boston, and published in the New England Journal of Medicine.
"Entering orders actually changed outcomes for the better," said Samuel Goldhaber, M.D., lead author of the study, in a telephone interview. "We did get a 41% reduction in clinically important DVTs with no increased risk of bleeding."
Goldhaber, a cardiologist who has been at Brigham and Women's since 1976, said there has been a notable paucity of orders for DVT prophylaxis at the Harvard Medical School-affiliated hospital despite solid medical evidence that they improve patient outcomes. In addition, the 725-bed hospital has had its computerized physician order-entry system wired with an alert to suggest DVT prophylaxis whenever a patient order for bed rest is entered.
"It wasn't enough," Goldhaber said. His article cites an earlier audit that found prophylaxis was used at Brigham and Women's with only 52% of patients who developed DVT while hospitalized for other reasons.
But Brigham and Woman's isn't alone in that regard.
"It's happing all across the country," Goldhaber said. "I ran a survey of 183 hospitals and we found that a majority of patients who developed DVT had not had prophylaxis."
So, Goldhaber and his colleagues spent two years planning a study, four years carrying it out and a year analyzing the results to see if a more sophisticated alert system could be created to improve the situation.
From September 2000 to January 2004, 2,506 patients from medical and surgical services were identified as being at increased risk for DVT. The patients were assigned to 120 physicians who were not made aware of the trial.
The clinicians were divided into two groups. An intervention group, whose physicians would see 1,255 at-risk patients, would receive new, risk-based alerts. A control group, whose physicians saw 1,251 at-risk patients, would receive no new alerts. (The old global alerts were left to run on the CPOE system for both groups.)
The new alerts were generated by a computer program that calculated a cumulative risk score based on eight common risk factors for DVT in each hospitalized patient's risk profile. Risk factors were weighted -- prior venous thrombolism, for example, was given three times more heft than obesity -- and a target cumulative risk score was determined as a trigger point.
If the target score was reached, the program searched electronic orders for ongoing use of mechanical or pharmacological prophylactic measurers. If it found none, it produced an alert that required the physician's acknowledgement and an offer to place instructions to either withhold or order prophylaxis from the same computer screen.
Physicians also could link to the hospital's DVT guidelines that included various drug regimes. Control-group physicians had access these guidelines, too, but they were not prompted to use them.
Even with the new, targeted alerts, prophylactic measurers were ordered for just 34% of those at-risk patients in the intervention group. Viewed another way, Goldhaber said, despite the alerts, doctors in the intervention group chose to ignore the advice of the computer two-thirds of the time.
In comparison, however, DVT or pulmonary embolism prophylactics were ordered for just 15% of the at-risk patients in the control group.
Goldhaber said the risk-assessment scale his team developed could be used to launch a similar alerts program at a hospital without a sophisticated IT system.
"Those hospitals without CPOE could assign a nurse, check (a new patient's) risk score against the point-scoring system we invented and check the handwritten order to see if those patients were receiving prophylaxis -- and page the doctor if they weren't.
"We certainly proved the principle," he said, though he added, "I think it's easier to do with a computer."