Leading the Call for Reform of Medical Device Safety Surveillance

Abstract

An ideal surveillance system for medical device safety would comprehensively collect data on adverse events across the life span of a device. Preferably, the system would be integrated into electronic health records to allow seamless identification, tracking, and real-time reporting of device-associated adverse events. Moreover, it would parse adverse events to detect substantial safety signals and underperforming devices. Such a system would also allow implementation of corrective actions in a swift and commensurate manner.In this issue of JAMA Internal Medicine, 2 reports 1,2 illustrate the importance of an improved surveillance system, particularly for life-sustaining medical devices. In one report, Davidsson et al 1 present the nationwide experience of Iceland with the Riata defibrillator lead (St Jude Medical), which was recalled in 2011 because of the risk of serious injury or death. The study adds to prior experience about adverse events with the Riata lead. 3 Davidsson et al 1 used a proactive protocol to detect events, which ranged from externalization of the conductor to death, and compared these outcomes in patients with a control group of patients with other defibrillator leads. The report effectively illustrates the extent to which Riata leads underperformed; the lead failure rate was 37% (19 of 52), including 1 instance that resulted in death, compared with the 8% failure rate (4 of 50) for other defibrillator leads (hazard ratio, 4.67; 95% CI, 1.18-18.50). It is noteworthy that Iceland has exemplary record keeping; all adverse event data were captured and are available in a national registry.In another study, Sengupta et al 2 characterize the high burden of serious adverse events (including death) with another cardiac device, the InSync III model 8042 heart failure pacemaker (Medtronic). The report demonstrates considerable deficits in both the swiftness (19 months from first notice of pacemaker failure to recall) and the appropriateness of the response (the US Food and Drug Administration [FDA] classifying the recall as class II-a low probability of serious adverse events). This long unexplained delay before the recall and the inappropriate recall classification raise concerns about patient harms that could have been prevented by speedier and stronger regulatory actions.Although the characteristics of an ideal surveillance system for medical device safety are easy to state, these 2 reports demonstrate the practical challenges. The reports add to a 2015 study by Tseng et al 4 that used systematic interrogation and autopsy of sudden cardiac deaths in patients with cardiac implantable electronic devices and found that passive surveillance efforts may inaccurately estimate device malfunction. Unfortunately, cardiac devices represent the tip of

FDA Device Oversight From 1906 to the Present

“…Development of a UDI system was prompted by the FDA's Unique Device Identification System Rule of 2013, 54 which required manufacturers to assign UDIs to devices. This requirement has facilitated development of a standard for electronic health record (EHR) documentation of device implants in patients, 55 adverse event reporting, and recall notification. UDI availability and use also supports improved aggregation of postmarket data from EHRs, payer claims, and clinical registries, enabling more robust postmarket surveillance.…”

Controlling Safety of Artificial Intelligence-Based Systems in Healthcare

“…Such a system should correctly identify safety signals, practically collect longitudinal data, effectively report adverse events, and strictly define thresholds for device recall [61,103]. An ideal level of post-market surveillance in AI-based medical devices includes three parts: extensively collecting data across the lifespan of devices, integrating results into electronic health records, and full tracking and reporting adverse events [58]. Developing and implementing a clear definition and distinction between information systems (data for human-human interaction) and post-market surveillance systems (data for AI-human interaction) is crucial.…”

A Safety Controlling System Framework for Implementing Artificial Intelligence in Healthcare

“…integrating results into electronic health records, and full tracking and reporting adverse events [67]. Developing and implementing a clear definition and distinction between information systems (data for human-human interaction) and post-market surveillance systems (data for AI-human interaction) is crucial.…”

“…This interaction includes five attributes, such as "Is there any information system developed between a health institution and an AI developer during the lifetime of AI-based medical devices? "[67][68][69]. The attribute of human-AI interaction is associated with designing safety guidelines for interaction between humans and AI.…”