Though he lived over 100 years ago, Austro- Czech author Franz Kafka’s work is still used as a reference point when depicting the failures of bureaucracy. One of his most impactful works, The Trial, follows the protagonist Josef K, who is arrested by authorities without ever knowing what he’s done wrong, how the justice system works or how to fight the outcome. By the end of the novel, with the state’s reasoning and processes still impenetrable to him, he’s executed.
Critics of current medical device regulations might also say that today’s bureaucracy is failing patients and obscuring decision-making mechanisms in a similar manner. Especially when it comes to device recalls. Across the US, the UK and the EU, recalls are mostly down to the manufacturer. The process is usually voluntary: manufacturers are meant to receive and investigate complaints about their devices and inform the regulators if a defect that could result in a recall is found. Though regulators can issue a recall order if they think a company is being too slow to act.
Regulators will then oversee the recall itself, though it’s the manufacturer’s responsibility to implement it. For example, in the US, the Food and Drug Administration (FDA) determines whether the recall has been rolled out effectively and when it can be terminated. Often, regulators will help spread information about the recall, too, though in practice, the process is opaque and inconsistent – in large part due to failures in collecting, sharing and making decisions based on data in a timely manner.
In June 2011, a report by the United States Government Accountability Office (GAO) noted that the FDA does not routinely analyse recall data, and that their process for overseeing recalls was unclear. For instance, they didn’t have any established criteria for assessing whether a company had removed the faulty devices sufficiently. Further, issues with medical devices are under-reported while data collecting systems are inefficient and lack transparency, reports a 2022 paper in the British Medical Journal (BMJ).
Perhaps the most high-profile case of late was the 2021–2023 recall of Philips Respironics’ CPAP and CBAP machines, where patients were found to be breathing in harmful foam. Despite thousands of reports being filed detailing problems with the device, both the company and regulator have come under fire for being slow to act – all while patients who were unaware of the issues continued to use the machines.
With recalls on the rise – from 2012 to 2022, there was a 125% increase in medical device recalls in the US, with the number of recalls within the most serious classification hitting a 15-year high – what might change to improve the process?
Better information flow
“Recalls can be massively complex,” says Stephen Gilbert, professor of medical device regulatory science at EKFZ Digital Health, part of the Faculty of Medicine of the TU Dresden. As exemplified by the Philips Respironics case, there are serious issues with how manufacturers report recalls, and how that information is accessed.
Here, Gilbert believes that a better system would involve sharing information on device performance and quality between medical practitioners, patients, manufacturers and regulators. Indeed, the BMJ paper mentioned earlier describes concerns about ineffective data collection systems and the poor quality of information collected.
But how to improve the information flow? Gilbert is optimistic that digital avenues hold part of the answer, especially if a device can directly upload information to a publicly available database. If everyone from regulators to journalists and other stakeholders could access this, it prevents complete reliance on the manufacturer to report an issue with a device. There’s no excuse here, he explains, as if we have access to AI, we should be able to readily report live information from medical devices.
In the UK, the Medicines and Healthcare Products Regulatory Agency (MHRA) might point to the Yellow Card app, whereby anyone can report issues to a regulator. But Gilbert believes this just doesn’t work as intended. “It’s definitely not on the patient interface, doctors are immensely busy,” he says. “[Reporting] this way doesn’t happen that often.”
And it’s not just reporting into databases that Gilbert believes holds the answer, but making the databases themselves more accessible. For instance, the US has a database on device issues, but in Europe such data can be harder to access. “It’s about information in the public domain so the authorities know their homework is being marked in public,” he says.
In the case of the Philips Respironics issue, Gilbert believes a recall resolution might have been reached much quicker if it were more widely seen that “these patients are telling their doctors they’re breathing in foam”. The rise of wearables and digital interfaces, where a device is connected to an app or central database, could potentially transform how incidents are reported.
More efficient systems
Gilbert believes that the emergence of AI and digitised systems could also improve outcomes by driving data-sharing efficiencies – in part because the information from devices can be more readily transmitted to larger systems. Trends in performance and usage could also be identified quickly using AI, for instance. “Wearables, types of hospitals at home, internal medical devices with an app…this could transform medicine,” he says.
Here, the MHRA is experimenting with this technology. They’re leading an AI airlock experiment which allows device manufacturers to explore how they can collect evidence to support the safety of products. While this is primarily concerned with readying products for approval, and focuses on AI devices, the approach could potentially help inform recall decisions, too – for instance, in helping manufacturers understand what they’re required to deliver to ensure device safety. “It shows we’re trying to get better in this area,” says Gilbert.
Gilbert is confident these technologies could modernise data-sharing processes – currently, the process centres around preparing paperwork and filing it. But with digital systems, manufacturers and regulators could move towards a process where performance and safety feedback can more easily be given on devices and systems, much the way pop-ups work online. This would be a much simpler way of logging reports. It could also help to standardise reporting systems and make them more transparent. With clear processes for collecting, storing, and sharing information with the right people, firms would have little excuse for inaction.
Transparency matters here, adds Gilbert. “If this is out in the open, companies have to take it seriously… [they] have to make sure their laundry isn’t dirty on the line.” The subtext being: if the right data is made accessible to all relevant parties, it wouldn’t be possible to obscure information about faulty devices – or to drag one’s feet in implementing a recall.
While AI could potentially be used to automate processes and sort through large datasets in device recalls – for instance, an AI chatbot could collect feedback on devices while another model might automatically flag performance deviations – we’d need to figure out how we’d regulate these systems, too. Especially when integrated into a medical device. “We can’t regulate an AI decision support system the same way as a hip implant,” notes Gilbert. It’s perhaps something on the horizon in years to come.
The tools to act
As recall processes evolve, Gilbert warns they mustn’t double down on an already documentationheavy framework. For instance, manufacturers must report incidents with their devices to regulators, document risk assessments, submit their strategy for performing the recall, deliver notification letters and more. “Manufacturers are rightly saying these documents [they have to provide] don’t do much for patient safety,” he explains, noting that even for the most comprehensive audits of devices, which could spot the need for recall or not, it’s all “the manufacturer’s filtered opinion”.
Clearer reporting standards could help make this process less opaque for both manufacturers and regulators. And having streamlined reporting channels that make it easy to log data could cut down the admin burden and get things moving quicker. It gives all parties what they need to take action, faster – and if that data is more widely available, firms could potentially be held accountable should they fail to act. Plus, it gives patients and practitioners peace of mind that their concerns have been filed and received by the right people.
In its 2011 report, the GAO recommended the FDA should regularly assess how a recall is progressing and develop criteria for gauging its effectiveness. A solid flow of information from the devices, to the manufacturer and onto the regulator would certainly be a big help in tackling this problem.
