Sensors to go are coming of age

By | April 2, 2020

When healthcare experts in Europe started debating whether or not to reimburse digital health applications two years ago, Josef Hecken, a very senior German health politician, said in an interview that health insurance companies should not be obliged to pay for “fun applications”. Hecken is head of the German healthcare system’s governing board for reimbursement decisions (G-BA). He is among those who really has a say on where the money flows in German healthcare. 

Quotes like Hecken’s led the German minister of health, the Christian Democrat Jens Spahn, to create a new reimbursement pathway for digital health applications that (partly) bypasses the traditional decision making chains for reimbursement issues. The law that established this new pathway came into effect at the end of 2019. Whether it will work, remains to be seen. 

All around the world, there is still a sound scepticism towards digital health applications and wearable sensor technology in the realms of “deep healthcare”. Up until now, there is no healthcare system at all that offers straight-forward reimbursement for, say, a digital application that continuously monitors blood pressure or seizures. Most of what is being paid for is driven by marketing considerations of healthcare organisations that want to be seen as being innovative or of health insurance companies that want to attract younger – and less costly – customers.

The Apple Heart Study as a trailblazer

A key issue with wearable sensors, of course, is the lack of clinical evidence. Collecting physiological data is the easy part. Proving that collecting data 24/7 really makes a difference is far more difficult. Gerd Antes, a statistician who was closely associated with the international Cochrane Collaboration for many years, likes to put it this way: More data does not necessarily mean better medicine, and this holds true even if the data itself is correct.”

What Antes and others mean can be illustrated by the most famous clinical trial on wearable sensor technology that has been conducted so far, the Apple Heart Study, results of which were published a year ago. In this impressive research endeavor, 400,000 citizens – who were recruited exclusively online – measured their heart rate with the PPG-sensor of a smartwatch in order to screen for atrial fibrillation. The study showed that a smartwatch based atrial screening was feasible. What it didnt show was that this type of screening makes sense, i.e. that it does more good than harm.

Apple is quick to report anecdotal evidence from customers. According to Sumbul Desai, head of Apples healthcare business, the company receives many letters of grateful smartwatch users who claim that the screening saved their lives. But proving net clinical benefit” is tougher than providing anecdotal evidence. We simply dont know yet whether a smartwatch based atrial fibrillation screening really reduces strokes, or whether overtreatment (with anticoagulants) will overcompensate any early detection benefits in the end. Desai, still a part-time practising physician, knows that: Thats how research happens. The whole essence of research is answering the unknown, so thats what we look forward to doing in partnership with the medical community.”

Read More:  Residential school remains could reveal how 215 children lived — and died, experts say

Heart rhythm specialists cautiously endorse wearables

In the case of atrial fibrillation screening, Apple is now partnering with Johnson & Johnson for a second big clinical trial, the Heartland Study. This trial that will focus on high risk patients aged 65 and older will not be about screening for atrial fibrillation only, but about detecting atrial fibrillation and subsequently treating it. When talking to Healthcare IT News sister publication HIMSS Insights in London, Sumbul Desai said that news on the start of this trial will be announced very shortly. If everything goes as planned, Heartland will be the first large-scale wearable sensor screening trial with a clinical endpoint in the history of medicine.

That cardiologists, in particular, are very interested in wearable sensors related to the heart rhythm was illustrated at the Consumer Electronics Show (CES 2020) in Las Vegas. The Heart Rhythm Society (HRS), the professional body of cardiologists dealing with heart rhythm diseases in the US, used the event to publish a position paper on wearables together with the Consumer Trade Association (CTA). It clearly endorses the recent developments: Wearables can contribute to an early diagnosis and to a better management of diseases”, according to Dr. Nassir Marrouche from Tulane University School of Medicine. It is not an uncritical endorsement though: The HRS states that symptom-related data capturing, specifically, could be very helpful and should be encouraged, but that the jury was still somewhat out for periodic or continuous data capturing unrelated to symptoms.

More and more wearables take clinical trials seriously

Wearable sensors are not only gaining momentum in the medical management of heart rhythm disorders. The Dutch startup NightWatch is focusing on neurology, specifically on epileptic seizure disorders. A major problem here is what neurologists call sudden unexpected death in epilepsy, or SUDEP. It is estimated that one in 5,000 children and one in 1,000 adults with epilepsy die from SUDEP each year. SUDEP usually happens at night, which is why seizure detection mattresses were invented that generate warnings related to seizure-typical movement patterns.

Read More:  Early detection of Alzheimer’s possible through algorithm

The problem with seizure mattresses is that they are error prone. This is why NightWatch has developed a wearable, a bracelet, which not only analyses the patients movement but also features a continuous heart rate measurement. Combining both has proved highly effective in detecting nightly seizures in a recently published clinical trial: The system generated an alarm in 86 per cent of suspicious situations, as evaluated by specialised epilepsy nurses. And there were only 0.03 false alarms per night, much better than mattresses,” says NightWatch executive Jeroen van den Hout.

That NightWatch hasnt simply pushed its solution into the market, but conducted and published clinical trials with relevant endpoints, is paying off. In the Netherlands, the bracelet has become part of the regular epilepsy treatment for many patients, particularly children, and it is being reimbursed regularly. The company is now entering the German and other markets, and it is hoping for similar reimbursement schemes there.

The future of wearables is AI-empowered

When looking at what works and what doesnt in healthcare related wearables, it is becoming increasingly clear that sensor technology in itself wont do the trick. According to Steven LeBoeuf, founder and CEO of Valencell, successful companies combine robust wearable sensors with AI algorithms that either improve the accuracy of measurements, provide for medically relevant interpretations of measurements, or both: When we look at the future of wearables, it is not so much about new modalities. Good things will come from old school modalities that collect datasets and apply machine learning.”

LeBoeuf is considering Valencell a proof in point, of course. The company is not so much a wearable device manufacturer in itself, but rather a technology and algorithm provider for hardware producers. Valencell technology is used by many earbud manufacturers, for example, to measure the heart rate in the context of fitness applications based on an in-ear PPG sensor. The challenge that the company is taking on now is tougher than heart rate: Valencell has announced a technology that measures blood pressure continuously based on in-ear PPG measurements.

PPG-based blood pressure measurements are not new per se, but up until now, they have a reputation of being inaccurate and thus not particularly suitable for medical applications. Valencell is addressing this problem with machine learning. According to LeBoeuf, more than 15,000 datasets have so far been collected in shopping malls in the US, the Philippines and Vietnam. Work is continuing, but the clear goal is to make blood pressure tracking with audio earbuds as reliable as heart rate tracking – not thanks to a new sensor technology but thanks to sophisticated machine learning.

Read More:  Why anxiety keeps coming back

Sleep apnea: From screening to diagnosis

Sleep apnea diagnosis is another area in which machine learning empowered wearables are about to make a major difference. There are myriad tools out there that claim to diagnose” sleep apnea based on wearables that measure blood oxygen content, another PPG-sensor-based measurement. The truth is, though, that sleep apnea cannot be diagnosed based on oxygen measurements. These measurements are suitable for screening, and that is it. Diagnosing sleep apnea requires analysing respiration plus electroencephalography, a combination of measurements that, at the moment, is only available in sleep laboratories.

But the times, they are changing in sleep medicine too. The US-based company Dreem has recently developed an easy-to-use headband that features electric recordings of brain activity. This moves much closer to polysomnography, the gold standard for diagnosing sleep apnea. The device also records heart rhythm, respiration, and movement, and it uses machine learning algorithms to calculate sleep stages out of the various measurements. In a clinical trial that so far has only been published on the bioRxiv preprint server, the headband-based sleep staging showed a very strong correlation with polysomnography-based sleep staging. This shifts the home-based, artificial intelligence empowered sleep laboratory a significant step closer.

Abandon fun applications, enter the medical wearable

Once diagnosed, respiration of sleep apnea patients could even be monitored much more effectively. A group of researchers at Ghent University in Belgium have developed and validated a wearable thorax patch that uses machine learning to accurately depict the breathing pattern of the sleeper based on bioimpedance measurements. Smart clothes could even leave the fitness-realm and, thanks to machine learning, enter serious healthcare. China-based wearable specialist Xenoma, for example, has developed sensor-equipped pyjamas and leisure suits that allow remarkably detailed movement analyses and also generate alerts in case of falls. Simply labelling wearables as fun applications” might still be good for a quick quote, but it doesnt reflect reality anymore. 

This article was first published in the latest edition of HIMSS Insights, Data Meets Privacy. Healthcare IT News and HIMSS Insights are HIMSS Media publications.

News from healthcareitnews.com