Research grant seeks to reduce errors in blood collection, transfusions
Giving people red blood cells that are incompatible with their blood type is a primary cause of preventable death in transfusion medicine. Accidentally collecting blood from people infected with HIV or hepatitis affects the overall safety of the nation's blood supply. The data analysis work of KU's Costas Tsatsoulis will look for ways to reduce errors in both the collection and transfusion of blood.
Tsatsoulis' work is part of a $3.2 million National Institutes of Health grant that hopes to establish a way of reporting near-miss errors within transfusion medicine. Near-miss reporting-or gathering data on mistakes that almost happened-already is standard in critical fields like aviation or nuclear power.
Most error reporting looks at what happened, but near-miss reporting looks at why something almost happened. Near misses, says Tsatsoulis, are far more interesting and potentially more useful than errors. More numerous than critical errors, near misses provide more data and also reveal a wider variety of problems.
"The hope is that we can change blood procedures," says Tsatsoulis, whose lab is part of the Information and Telecommunication Technology Center on KU's west campus. "Errors aren't only caused by humans. Errors can come from the environment, such as bad lighting. Or they can come from a certain business culture, such as employees who say, 'We never call our supervisor on the weekend.' The data analysis will identify where the faults lie."
More than 25 organizations-from small and large hospitals to blood collection sites and the Red Cross-are noting blood procedures from the point of donation through transfusions. Together these organizations handle a large percentage of the nation's blood supply.
The error and near-miss data they gather will be forwarded to Tsatsoulis and his students for analysis. Using computers, they will apply different algorithms-or sets of well-defined problem-solving rules-to the data to identify various patterns.
"Humans," he says, "can't identify the patterns. They need automated methods that show causality and a clustering of events around certain parameters. Once our computers identify a pattern, humans then will study it."
For instance, the data might show that more near misses occur at certain times of day or on holidays. Perhaps a blood procedure itself may be the root cause of near misses. A case in point-as a result of near misses, blood bag labels are now color-coded by blood type to reduce the chance a patient will be given the wrong blood.
"Before anyone can stop errors, they have to find them," Tsatsoulis says. "When you find them, you then have to learn why they occurred and then change the procedures."
The $3.2 million grant is coordinated by Columbia University in New York. Tsatsoulis' four-year subcontract is for almost $350,000. Project directors invited Tsatsoulis to join their study after reading about his work with data mining.
For more information, contact ITTC.