A large federal database that healthcare researchers use frequently was found to contain "a clinically important rate of apparent discrepancies," including 11 cases where patients were recorded as being both dead and alive.

In a report posted on the Annals of Emergency Medicine website (PDF), Dr. Steven Green of Loma Linda (Calif.) University Medical Center detailed his examination of 10 years of records from the Centers for Disease Control and Prevention's National Hospital Ambulatory Medical Care Survey. A total of 348,367 emergency department visits were recorded for the period, which covered 2000-09. In 11 instances, Green found, patients were recorded as having died in the emergency room and been admitted into the intensive-care unit.

In addition, clinicians reportedly performed intubation on 875 patients and then discharged 81 (9%) of them and admitted 153 (17%) into a noncritical-care unit—dispositions that are "incompatible with this procedure," Green wrote. After intubation, he said, surviving emergency-department patients are almost universally admitted to the intensive-care unit.

"While it is theoretically possible that a patient could be intubated, extubated and discharged, it would be a highly unusual circumstance," Green said in an e-mailed news release. "It is impossible to believe that more than one-fourth of these patients were placed anywhere other than the ICU or the morgue, and yet that is what 10 years of federal reports indicate. Something doesn't add up."

The report noted that the Annals of Emergency Medicine has published 25 studies between 2000 and 2011 that relied on the NHAMCS database. It was noted that the advantages of using the database include its "enormous sample size." Its disadvantages, according to the report, include researchers not having access to source medical records to resolve questions and discrepancies.

Green wrote that free text entered in records under "cause of injury" was available for only 97 of the incompatible visits, but that it offered some clues as to mistakes made by the individuals entering the records. Injuries described included gunshot wounds to the head or sternum and a person being "dragged by car for 30 feet." Cases such as these "indicated a near-certain probability of critical-care admission or death," Green wrote. Also, Green speculated that in 11 cases, chart reviewers mistakenly characterized a tracheotomy as intubation.