concluded that any of the surveillance strategies it investigated detected only a minority of the adverse events that were identified by the strategies collectively and that each strategy had significant limitations. An update of the first published report on the study is currently being prepared (personal communication, Thomas P. Gross, M.D., Director, Division of Postmarket Surveillance, CDRH, February 4, 2005).

A particular focus of investigators was a “computer-flag” strategy that was embedded in the hospital’s computerized patient record system. It involved nurse review of flagged records using a protocol devised in consultation with FDA. The flags were based on “detection rules” for seven categories of events (e.g., complications and hazards related to various types of catheters), and one goal was to identify hazards before they caused harm. This approach yielded more information on adverse events than the hospital’s voluntary adverse event reporting system, but it missed some important events. Disappointingly, the positive predictive value of the flags was low, that is, the great majority of flagged events did not, upon investigation, involve a device-related problem or hazard. Thus, its utility for detecting potential problems before they could cause harm was limited.

Nonetheless, the computer-flag strategy—in combination with other tools—helped investigators better understand the clinical environment. “It appeared that the typical health care worker response to a device problem was to fix it or retrieve a new device that worked and then move on, an appropriate solution at the individual patient level but not an effective systems approach” (Samore et al., 2004, p. 333). This observation was reiterated in committee discussions with clinicians.

To consider patient perspectives, investigators conducted a postdischarge patient survey. It found that people focused on simple, common devices that caused discomfort rather than what are normally defined as serious adverse events. Events reported in the survey had no overlap with other reports.

Another strategy—retrospective review of medical records for ICD-9-CM codes5 that indicate a likely device problem (e.g., code 996.01, mechanical complication due to heart valve prosthesis)—was useful in identifying problems that occurred prior to hospitalization. Still, with ICD-9-CM codes, device problems can only be described by broad category rather than by individual device. Thus, this strategy is not well suited to identify a


The International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) is used in assigning codes to diagnoses and procedures associated with hospital utilization. It is based on the World Health Organization’s International Classification of Diseases, Ninth Revision (ICD-9). In the United States, the National Center for Health Statistics and the Centers for Medicare and Medicaid Services have responsibility for maintaining the system.

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement