October 2004

Fighting Infections with Data

By Corie Lok

Every year, two million Americans contract infections in hospitals, and about 90,000 of them die as a result. Despite the severity of this quiet epidemic, physicians have no systematic way to quickly and effectively identify spreading hospital infections. Now a small but growing number of U.S. hospitals are adopting data-mining technologies widely used in other industries to alert doctors to problems they might otherwise miss.

One company helping hospitals connect the dots is MedMined of Birmingham, AL, which has sold its data analysis services to more than 80 hospitals since physician and computer scientist Stephen Brossette founded it in 2000. Hospitals transmit encrypted data from patient records and lab tests to MedMined, which then uses its data-mining algorithms to tease out unusual patterns and correlations. At first, only the most computerized and technologically savvy hospitals were interested, but "now we're seeing more of a mainstream push," says Brossette, the company's president and chief technology officer. One incentive: increased public scrutiny of hospital-acquired infections. Illinois, Pennsylvania, and Missouri have passed laws requiring hospitals to publicly reveal their infection rates, and similar bills are pending in Florida and California.

MedMined and competitors such as Cereplex and Theradoc (see "Computerized Germ Catchers," below) also track emergency-room and outpatient-clinic data to look for community outbreaks and bioterrorism events. Automating disease surveillance "is going to be hugely beneficial for patient care," says Jerome Tokars of the Centers for Disease Control and Prevention in Atlanta. "Instead of collecting and counting data, personnel can start doing more prevention of hospital-acquired infections."

One concern about the technology, says Robert Weinstein, chair of infectious disease at the John H. Stroger Jr. Hospital of Cook County in Chicago, is that it may send doctors chasing after too many false positives -- seeming clusters of infections that turn out to be random statistical anomalies. Even so, most infection control specialists agree that they need help from computers in crunching the mountains of patient data that may conceal evidence of an impending outbreak.