information. Such databases, especially if expanded to hospital networks with large numbers of children, could provide a useful resource for certain pediatric device safety studies because they would provide greater specificity about device type, more relevant clinical details, larger numbers of children with characteristics of interest, and a denominator for safety and adverse event analyses.
Compared to hospital-only sources, large automated billing and medical record systems maintained by some HMOs and other entities offer the advantage of including data on both inpatient and outpatient care. Unfortunately, as is often the case for inpatient-only databases, detailed information about devices is frequently not available because use of a device is subsumed in another, broader event or transaction such as billing for a medical or surgical procedure. Moreover, as discussed earlier, current device coding options are often insufficient to identify the brand or model of a device implicated in an event. Without such codes, a strategy based on review of computer-based medical records cannot automatically link event reports to specific devices, although the significance of coding limitations varies among devices.
The difficulties for device surveillance and epidemiology created by coding limitations is captured by the following comment by a researcher on problems encountered in determining the specific model of an ultrasound device used for many obstetrical patients, even though investigators knew when and where the procedure occurred. “We needed to contact each obstetrics office and ask a busy person to find and read to us the make and model information, [and] since some offices had more than one model, we couldn’t assign their patients to specific devices” (personal communication, Richard Platt, M.D., Harvard Medical School and Principal Investigator, HMO Research Network CERT, April 8, 2005). This comment reflects experience in the HMO Research Network, which is one of the Centers for Education and Research in Therapeutics (CERTs), a program created by the FDA Modernization Act of 1997 and overseen by the Agency for Healthcare Research and Quality (AHRQ). The network, which includes 14 large HMOs, is conducting several retrospective cohort studies of drug safety and medication errors. Past studies have investigated the association between certain nonsteroidal anti-inflammatory drugs and serious coronary problems (Ray et al., 2002). Device studies have proved substantially more difficult for the Network (personal communication, Richard Platt, M.D., Harvard Medical School and Principal Investigator, HMO Research Network CERT, December 9, 2004; personal communication, James Donahue, Ph.D., Project Scientist, Epidemiology Research Center, Marshfield Clinic Research Foundation, December 9, 2004). In conversations with experts from the Network and other investigators working with large database systems in the CERTs program and elsewhere, the committee found