Discrepancies raise questions about database information quality

A new study has found a "clinically important" level of discrepancy --including 11 patients listed as both dead and alive--in a major national database, raising questions about its data quality as a source for research.

Ten years' worth of data in the National Hospital Ambulatory Medical Care Survey (NHAMCS) looked at the disposition of patients who underwent intubation. From the total of 348,367 ED visits, 875 underwent intubation, but what happened next for 27 percent did not follow expected practice.

That would be either death or admission to a critical care unit, either at the same facility or by transfer to another, according to the study published at the Annals of Emergency Medicine. Nine percent were listed as discharged and 17 percent as admitted to a non–critical care unit.


  • In 37 (4 percent) of these visits, there was conflicting information about disposition.
  • In 11 cases, patients listed as admitted to a critical care unit, but also transferred to another hospital.
  • In another 11 cases, patients were noted to have died in the ED, but also to have  been admitted to a critical care unit.
  • In nine visits, patients were described as being admitted both to a critical care unit and to an observation unit.
  • In three cases, patients were noted to have been admitted to a non–critical care unit and also transferred to another hospital.
  • In two cases, patients were listed as not admitted, but then also admitted to a critical care unit. And in one case, a patient was noted to have been taken to the operating room and also transferred to another hospital.

"This study raises serious questions about the data quality of NHAMCS that may change the interpretation of published articles and future inquiries that use this database," wrote lead author Steven Green, MD, of Loma Linda University Medical Center and Children's Hospital in California.

Abstraction in the database is performed both by hospital staff and by field representatives, who may have no formal healthcare training, notes MedPage Today. That article mentions a second review that found data quality issues with NHAMCS.

However, because the original medical records are not available for review, Green's team could not determine whether the discrepancies resulted from errors in chart abstraction, errors in keying in data, or a combination of the two.

Columbia University researchers recently warned that inaccuracies in EHRs could skew research and recommend that researchers take a "radical shift in approach" and study EHRs themselves, not just the data.

Anna Lembke, M.D., an assistant professor of psychiatry and behavioral sciences at Stanford's School of Medicine, recently wrote that "if data-mining is based on inaccurate information, then the whole operation is a house of cards."

To learn more:
- read the study
- here's the MedPage Today article