A Cambridge, Massachusetts, research lab is addressing some of modern medicine’s most overlooked issues with cutting-edge IoT technology and an open-source approach, weaving aging devices and deeply siloed data into an accessible web of medical information.
The Medical Device Interoperability Program, or MD PnP, in affiliation with Massachusetts General Hospital and Partners Healthcare, is a hub for research into making medical devices dramatically smarter by making it simpler for them to share the data they gather.
With more and more people being monitored by IoT devices in hospitals and monitoring themselves with Fitbits and Apple watches, there’s suddenly a lot more digital data than there was before in the world of healthcare.
One challenge is to gather and analyze that data from disparate devices so it provides medical professionals with more complete information about the condition of their patients. Another is to make that process simpler for the IT staff that has to set up the systems.
The heathcare field ratchets the usual difficulties of integrating all these systems up to a much higher level. There are too many device makers, too many technical hurdles, too many regulatory issues – and the penalty for getting something wrong is that people could die.
Too many alarms
One of the main issues the lab tackles is alarm overload. Walk through any busy hospital ward, and one of the main ambient sounds that greets the ear is a chorus of beeps and boops from monitoring device alarms, like birds tweeting and whooping in a jungle canopy. And like biologists or National Geographic photographers, the healthcare workers who spend a lot of time in hospital wards eventually start perceiving the sounds as background noise, not something to be actually alarmed about. That, obviously, is a problem on the occasions where the alarm isn’t just going off because a patient’s blood oxygen dipped below the threshold for a millisecond or when the clip slipped off of a fingertip.
It’s not their fault, according to MD PnP lead engineer, Dave Arney. Traditionally, medical devices are a little bit “dumb,” whether they’re simply older technology or constructed to fulfill strict requirements such as: if the patient’s blood pressure reading EVER drops below a certain level, then alarm.
“All those alarms are almost certainly working exactly how they were intended to be used,” said Arney.
But Dr. Julian Goldman, the director of the MD PnP program, said, “If you were to rethink that environment today, you might say that maybe it should be conditional – maybe you should bring up a different alarm app for a patient in the ICU after surgery, and a different intelligent alarm app for a neonate,” he said.
Goldman and Arney think that a big part of the solution to alarm fatigue is intelligence and interoperability. Arney used the example of a pulse oximeter – a simple device clipped to a finger tip used to measure blood-oxygen levels – being connected to a blood-pressure cuff. If the oximeter knows that the blood-pressure cuff is taking a measurement at any given moment, it won’t alarm when the patient’s O2 saturation drops, since it knows that there’s a perfectly good reason for that to be happening.
Context-aware alarms, a la aviation, are what’s needed in medicine, according to Goldman. Many airplanes are equipped with systems that do things like sound an audible alarm if the pilot appears to be trying to land without lowering the landing gear.
“How is a plane so smart that it knows you’re trying to land with the landing gear up? Why doesn’t the alarm go off when the landing gear is up at 30,000 feet?” asked Goldman. “The point is when you integrate sensor data, your alarm becomes a lot more useful. It becomes trusted. That’s the kind of thing we need in health care. And to do that, you need to integrate data from many sources, and IoT’s capabilities add to that richness.”
Today, once healthcare workers move away from the bedside, all they have is the alarms, and the inability of simple machines to prioritize their alerts and fit them into a larger context is a problem.
As an example, Goldman and Arney showed off an image of a set of vital signs, with the ECG on the heart-rate monitor displaying a flatline. Despite the fact that the reading is the product of random electrical noise – not an actual cardiac event – the heart rate monitor sees only that the ECG reading has flatlined. It doesn’t know that the rest of the patient’s vitals, like blood pressure and oxygen saturation, are totally normal, so it makes an alarm as though the patient is having a heart attack.
The point is that these are not, on their own, particularly smart devices.
New IoT systems could be a boon for healthcare
Both Arney and Goldman wince a little at the idea that the goal is “interoperability,” per se, and Arney argued that the word is merely a placeholder for something a lot more complex.
The current state of affairs makes for a huge workload for hospital IT departments that are trying to take devices from dozens of different manufacturers and weave them into a useable infrastructure. Frequently, this drags third-party developers into the mix to create one-off software systems that integrate a hospital’s equipment.
The goals become lightening the load on IT and generally making it easier to enhance the capabilities of devices. “How can you allow for hospitals and companies to assemble interoperable components and build new algorithms and apps so that we can democratize the improvements that are needed in healthcare?” he asked.
The alternative proposed by MD PnP – an open platform that can be used to seamlessly connect devices without a lot of integration work for IT staff or the expense of hiring outside contractors – makes more advanced uses of medical technology more freely available to healthcare providers that might not otherwise be able to afford it.
Getting healthcare data into a more readily available format can ease the work that harried health professionals have to perform. For example, an alternative way to view alarm data other than having to look through a bank of monitoring devices to figure out which one is beeping would be more efficient and less disruptive. Or in another example, an infusion pump that would refuse to work if it was loaded with a prescription the patient is allergic to could avoid dire consequences.
OpenICE and testing
OpenICE, or the Open Integrated Clinical Environment, is the MD PnP lab’s signature accomplishment so far. It’s an open-source framework that can run on any Java-capable computer (the team uses Beagleboards for its testing modules) and connect medical devices via those nodes into an information-sharing whole monitored by a Supervisor module running on a laptop or smartphone. It automates node discovery, publishing and subscription among different nodes on the network, and can translate proprietary data formats from different devices into a common language.
The idea is to enable any device that outputs digital information of any type to be easily connected to a broad network of other devices. Simply by adding an OpenICE-enabled module to an existing piece of medical equipment, doctors and developers can make that piece of equipment programmable and smart, allowing it to share data among other devices on a network for access to smarter alarm functionality, for example.
OpenICE is based on a prototype standard developed separately under a National Institutes of Health grant. The platform gets around the byzantine complexities of medical-device connectivity formats by integrating them via the OpenICE framework. An aging dialysis machine might be monitored remotely by using this method, for example.
The MD PnP lab isn’t building finished solutions. Rather, it’s working on a logical infrastructure that it hopes third parties can use to create smarter medical devices more easily. OpenICE is an underpinning for informatics and clinical detection apps that the team thinks will help save lives and that it hopes will fire the imaginations of healthcare innovators.
The lab features gear that can simulate medical conditions for medical IoT devices to react to. One such rig – the team calls it Randall Jones – is an apparatus with a CO2 tank hooked up to a pair of bellows that behave like human lungs, allowing the team to test ventilator equipment and other gear that monitors lung function. The fake lungs could be programmed to behave as if they had a bronchial anomaly, that a developer might use to test an app centered on early detection of lung conditions, for example.
Using this type of simulated patient allows the team to make sure they’re getting realistic-looking data from medical devices for use in testing OpenICE-enabled apps.
It’s all about making medicine a smarter, more scientific and more evidence-based process, according to Arney.
“Evidence-based medicine is this radical notion that patient treatments should be based on scientific evidence, but that’s hard to do in medicine. It’s hard to conduct clinical studies of all these things,” he said. “There’s a saying that half of what you learn in medical school is wrong, but nobody knows which half.”