Surveillance of medical device-related hazards and adverse events in hospitalized patients
Although much research has used computerized databases to identify adverse events associated with medication use, parallel studies have yet to focus on computerized detection of adverse events with medical devices.
In this pilot study conducted over a 9 month period at a 520-bed tertiary care facility in the United States, Samore and colleagues used a rule–based flagging system in conjunction with an electronic medical record to detect adverse events (producing patient harm) and hazards (unassociated with patient harm) that occurred with device use. Examples of device-related hazards included device malfunction; misuse, including misdiagnoses; improper treatment and failure to recognize and act on information from monitoring devices. Devices were also categorized according to the body part or organ system in which they were used as well as other characteristics of their use (e.g., implantable, reuseable, disposable devices).
In a sample of 20441 regular and short-stay patients that excluded obstetric patients and newborns, the incidence and positive predictive value of device related problems identified through the use of electronic rules were compared to figures obtained through other ascertainment methods (i.e., voluntary incident reports, telemetry problem checklists, clinical engineering work logs, patient survey results, discharge diagnosis codes indicative of device-related adverse events).
Different ascertainment methods identified minimally overlapping subsets of device related problems. Also, a greater frequency of hazards and events was identified via rule-based computer ascertainment than with voluntary reporting. The positive predictive value of a rule-based computerized flag was quite variable ranging from 0% for flags due to extended surgery, change in urine output and low respiratory rate (<8/min) to 35% and 38% for catheter associated bloodstream and urinary tract infections, respectively. In addition to the catheter-associated infections, only line insertion/removal was associated with actual patient harm. All other flags were associated only with a potential hazard, if anything, and all had positive predictive values of <5%. Artifactual or physiological abnormalities were common but were rarely related to device-related problems. Across all types of flags, 7.8% (552/7059) were indicators of a confirmed device related hazard or adverse event with an incidence of 27.7 device-related adverse events detected by computer-based flags per 1000 admissions.
When identified though voluntary staff reporting, post-discharge patient survey or ICD-9 discharge codes, the observed incidences of device-related adverse events were 1.6, 18 and 64.6 per 1000 admissions, respectively. The overall incidence of device-related adverse events detected by at least 1 of these ascertainment methods was 83.7 per 1000 admissions. Use of a telemetry checklist over a 516 hour period of the study showed 593 false alarms and 4 missed ectopic events, with none resulting in patient harm. Review of clinical engineering logs also provided a means for identifying potential device related problems but were not linked to specific patients or events, precluding calculation of incidence rates.
On the basis of this research, Samore and colleagues concluded that identification of device-related adverse events has important patient safety implications but that additional strategies are needed for optimal detection of such adverse events.
Comment: In identifying possible human factors and device-related threats to patient safety, this research extends the use of rule-based computerized flagging from adverse medication events to device-related adverse events and hazards. In addition to demonstrating that significant numbers of individuals experience harm from device-related events, this study highlights the fact that computerized surveillance alone is likely to miss some device-related events. Furthermore, devices are quite diverse in the adverse events associated with their use and diverse surveillance strategies will most likely be needed to identify risks to patients.
Additional optimization of computerized and other approaches is required as the low positive predictive values with most devices will entail labor-intensive screening of multiple flagged “events” to identify a single device-related adverse event or hazard. Such surveillance strategies will also need to incorporate means for detecting the preventability and severity of identified hazards and adverse events, neither of which were a part of this study.
Finally, the study findings may not be generalizable to other facilities. Positive predictive values depend upon the prevalence of the event in the tested population and the incidence rates for device-related adverse events are calculated per thousand admissions rather than in hours, days or other measure of actual device use. Consequently, variations in device use or device-related adverse events across facilities may influence the reliability and validity of these surveillance techniques.
