Some of the patients had suffered heart attacks or were in septic shock; others had cancer, pneumonia, or a whole host of other conditions. From 2001 to 2012, more than 46,000 of them passed through the intensive care units at Beth Israel Deaconess Medical Center, a Boston hospital affiliated with Harvard. Their vital signs were monitored and recorded, as were their lab test results, their doctors’ notes, and reams of other data.
Today, these patients’ de-identified medical records have become a crucial cog with outsized influence in a burgeoning field of artificial intelligence research that’s trying to answer some of the toughest questions in medicine and health care.
MIMIC, as the freely accessible database is called, has been used in more than 500 research papers and conference presentations, according to an estimate from the lab that built it. Most recently, a study published this week used it to develop an algorithm predicting whether patients would be readmitted to the ICU within a month.