EXOME

all the information, none of the junk | biotech • healthcare • life sciences

So Far, Little Proof That Digital Health Is Healthy. Does It Matter?

Xconomy National — 

Digital health is popular among investors, there’s no doubt. But is it good for us?

Products such as fitness wristbands and online wellness coaching are generating buzz, and as a techno-skeptic I often roll my eyes at the trendy accessories and general assumption that life is easier online. Meanwhile, the funding keeps rolling in. Through three quarters of 2014 the nascent field has attracted $3 billion, already double the 2013 total, according to Rock Health, the San Francisco digital-health incubator, seed investor, and data provider. (Rock Health is no neutral analyst, but one close observer of the sector says the incubator’s funding totals are actually on the conservative side.)

So I decided to look for examples of high-tech products, from wireless devices to smartphone apps to online programs, that are already proving—in the eyes of the medical community, not just the marketing community—they make people healthier. That, of course, requires evidence.

The short answer is, there isn’t much. Steven Steinhubl is the director of digital medicine at the Scripps Institute of Translational Medicine in San Diego, and he’s on the lookout for technology and products that his group can put through clinical trials. “I believe mobile health technology is the future of medicine and can play a critical role improving health outcomes, boosting patient convenience and health provider satisfaction,” says Steinhubl. “But we have to demonstrate it. The vast majority of supporting data is in the form of pilot studies.”

When asked to define “pilot studies,” he says: “Dirty first looks.”

Steven Steinhubl (Photo courtesy of Scripps.)

Steven Steinhubl (Photo courtesy of Scripps.)

When I asked Rock Health managing director Malay Gandhi about the importance of rigorous clinical studies for up-and-coming digital health companies, he replied, in part, “I would never generalize to the entire category. I think that’s dismissive. While I think it is critical for everyone in healthcare to prove what they’re doing actually works, getting into a strength-of-evidence argument at this point seems slightly misguided. I’m not in support of delaying access to massive improvements over the status quo while we wait for the perfect study to satisfy academic criteria.”

It’s at heart a Silicon Valley attitude: users first. If it looks like a “massive improvement over the status quo,” get it out there.

But many in the field know they have to show real benefits; some have, and for others there hasn’t been time to gather significant samples of data. With the funding explosion this year, more fledging companies doing serious studies should be publishing peer-reviewed data in the next few years.

In the meantime, can those first looks be enough in some cases to convince doctors, nurses, hospitals, patients and insurers to trust, pay for, and put those products into practice? The field has plenty of applications that shouldn’t require the same body of evidence as, say, a device implanted in a patient or a new pharmaceutical product.

Joseph Kvedar, a dermatologist and director of Partners Healthcare’s Center for Connected Health in Boston, says the field needs to find “a happy medium.” He’s worried about the tech startup mentality—“stay lean, pivot when it doesn’t work, and assume when people adopt it it’ll be right.”

“But I don’t necessarily believe a five-year trial is always the right answer, either,” he says.

One good example of that happy medium comes from Steinhubl’s group at Scripps, which tested a band-aid-sized wireless patch called Zio developed by San Francisco-based iRhythm Technologies. Adhered to people’s chests, the patch detected about 50 percent more irregular heartbeats, or arrhythmias, than the more cumbersome standard detector, called The Holter, which patients don’t like wearing for more than a day or two. The National Institutes of Health paid for the 146-person study, and in a January blog post, NIH director Francis Collins held it up as an early example of a significant clinical outcome in mobile health. “Pretty amazing stuff,” he wrote. (You can see pictures of the two devices on the blog, too.)

Based on those results, Aetna added the Zio patch, which is water-resistant and can stay on a patient’s chest for two weeks, to its reimbursement policy. That’s because detection of arrhythmias isn’t easy, so the longer a person with suspected heart problems can wear a monitor, the better. And detection of arrhythmia is a good way to prevent serious complications, including stroke.

I talked to another entrant in the heart monitor business. San Francisco-based AliveCor has an app that turns a smartphone into a handheld electrocardiogram, meant for people with arrhythmia history or at risk for heart disease to use on their own or at a doctor’s office. If feeling symptoms like shortness of breath, a person launches the app and touches fingertips to the electrodes on the customized phone case to find out if he or she is undergoing atrial fibrillation, the type of arrhythmia most linked to stroke.

The FDA has cleared its use without a prescription to automatically interpret an irregular heartbeat as A-fib. The readout says, in part, “Possible AF detected. This result is not a diagnosis.”

But there’s no proof it makes patients healthier. To gain FDA clearance, “we didn’t have to show outcome-based data,” says AliveCor interim CEO Euan Thompson, the company’s third chief executive in two years. “We’re not claiming that the patient will therefore do better, we’re just detecting atrial fibrillation.”

Still, detecting A-fib is theoretically a very good thing. The National Stroke Association says it makes a person five times as likely to have a stroke, and 75 percent of A-fib related strokes are preventable. Thompson says AliveCor’s algorithm analyzes 100,000 heartbeats a month. It identifies A-fib correctly 100 percent of the time, he says, but still has a slight false-positive rate. (“We err on the side of caution.”)

That’s not yet good enough for Aetna, however. In the same policy bulletin that gives Zio a green light for reimbursement, the insurer writes this: “Aetna considers the AliveCor Heart Monitor (iPhoneECG) experimental and investigational because its clinical value has not been established.”

Thompson says AliveCor is in a different product category: episodic monitoring, not continuous monitoring. The product is in clinical studies, and the company would love insurance companies to tie the app’s use to the cost savings of preventing stroke, but “we are not supporting these clinical studies for the purpose of building a reimbursement case,” he says.

One effect of the new wave of heart monitors could be that more patients will be confirmed to have A-fib and wind up being treated with a drug regimen or a surgical correction called ablation. But much of the new health-related technology is about keeping patients away from surgeries, drugs, and other costly interventions. That’s an excellent goal.

On the other hand, keeping people away from expensive products is usually a terrible way to attract venture capitalists. But in the Obamacare era, fueled by widespread consensus that healthcare spending needs to drop, capitalists hope prevention has its rewards, as well.

One pilot study to watch is happening in Wyoming, where the state’s Medicaid program this year began using a smartphone app for pregnant women called Due Date Plus from Wildflower Health of San Francisco. It’s centered on a calendar, with reminders about prenatal checkups and milestones, and a checklist of symptoms that prompt women to call a health hotline. By the end of the year, Wyoming should be able to compare the first Due Date Plus group to years of baseline data: Are Due Date Plus mothers having fewer preterm babies, low-weight babies, and C-sections, for example?

Wyoming is not a populous state, of course, so the pilot data won’t be statistically significant. Neonatal intensive care unit admission [NICU] rate is normally 9 to 10 percent, says Wildflower CEO Leah Sparks, and the goal is to reduce that rate. “It will likely take at least a couple years to get sufficient data that we feel like the outcomes analysis has some level of statistical quality,” she says.

But she also puts weight on Wildflower’s ability to make Due Date Plus—and any other family-health app it develops—a brand-booster. “Everyone assumes that health plans only care about reducing medical costs, and we’re certainly tracking that for maternity,” she says. “But in a world where health plans are more competitive, with exchanges and new markets, they’ll be focused on other metrics like ‘how well am I retaining members,’ how well am I converting them to new products,’ ‘how well am I getting new enrollments,’ ‘what’s my brand level.’ That’s an easier metric on a year to year basis than certain types of medical outcomes.”

Sparks says two Silicon Valley companies who fund employee healthcare themselves have also put Due Date Plus into practice, with a third going live in November. “For those companies, one pregnancy that goes to the NICU and costs a couple hundred thousand dollars can really blow out your benefits for the year,” she says. “They also see maternity about recruitment, retention, and talent war. They have a lot of reasons to knock it out of the park with their prospective parents.”

Other app makers want their products to be so cheap and easy to implement, potential users won’t wait around for iron-clad proof. That’s happening in many U.S. municipalities with PulsePoint, the brainchild of Richard Price, a former fire chief in the San Francisco Bay Area town of San Ramon. Downloaded onto a smartphone, PulsePoint alerts users if they’re near somebody who’s made a 911 call for cardiac arrest. It also maps the nearest automated external defibrillator. The idea is that bystanders can start CPR and even apply the defibrillator before the first responders arrive. Years of public health data show bystander intervention saves lives. But there’s no proof PulsePoint works better than luck.

Nine hundred locations, with San Diego County one of the largest, don’t care. They’ve already integrated the PulsePoint software into their 911 call centers for $10,000, plus an annual license fee that goes from $5,000 to $25,000 based on population.

Steven Brooks, a Canadian researcher at Queen’s University in Toronto, wants to do a randomized clinical trial, but his first attempt to set one up, in collaboration with Toronto’s emergency medical services, fell apart because of Canadian privacy concerns, Brooks says.

In Seattle, the fire department’s associate medical director, Michael Sayre, says he wants King County, which includes Seattle, to adopt PulsePoint, even though the county already has what could be the nation’s highest survival rate for cardiac arrests in public, about 60 percent.

“I think it could make a difference and further boost the bystander CPR rate,” says Sayre. “That would be an experiment, I could be wrong, but it’s not a lot of money in the big picture. It’s more about getting this project on list of things that need to be done. That’s the major barrier.”

PulsePoint is the product of a nonprofit foundation, with the engineering work donated by a big Bay Area software company. But for-profit companies are using a similar strategy: Build upon bodies of evidence that the medical community already trusts, and don’t scare customers away with high prices.

With $28 million in venture funding, Omada Health in San Francisco has designed an online program, inspired by social networks and augmented by coaches only a phone call away, that aims to keep people in danger of developing Type 2 diabetes from crossing the threshold into disease. It’s based on the U.S. government-funded Diabetes Prevention Program, a set of guidelines tested in real-world settings for 15 years and considered the gold standard in behavioral intervention for Type 2 diabetes.

A series of new design ideas for Omada's online Prevent program, posted in Omada's conference room in San Francisco. (Photo by Alex Lash.)

New design ideas in progress for Omada’s online Prevent program, posted in Omada’s conference room in San Francisco. (Photo by Alex Lash.)

According to a metastudy, those attempts at real world practice have shown average weight loss of 2.4 percent after 12 months. Omada’s Prevent program, in a study the company did with two outside investigators, achieved an average 4.8 percent weight loss after 12 months.

It also reduced A1C levels—a proxy of average blood sugar level in the previous three months—nearly 0.4 percent, which Omada medical director Cameron Sepah says took the average participant from prediabetic to normal glycemic range.

Those data are important to Omada. CEO Sean Duffy says, “We’ve gotten payer traction with that trial plus our commercial data,” which remains unpublished. Omada only charges its customers—Blue Cross/Blue Shield of Louisiana, HealthNet, and Kaiser are a few examples—if Prevent users hit the promised outcomes. More data could be coming soon from a current U.S. Veterans Administration study.

Omada now wants to use the same DPP guidelines and apply it to other diseases like hypertension and high cholesterol. It’s encouraged in no small part by an influential U.S. health task force’s recommendation that the intervention techniques would be beneficial to people with other metabolic or cardiovascular risks, says Duffy. “That guidance is influential to the payer community, especially after the Affordable Care Act.”

In a sense, Omada is doing the same as biopharma startups that spin out NIH-funded academic research: build a for-profit venture upon years of taxpayer-funded work. But it’s not spending years toiling in labs and running tests on mice, monkeys, and humans. Omada is building online apps that nudge people to eat better and exercise more, basically—or that get them to nudge each other, through the ever-more acceptable environment of a social network.

Another company involved in diabetes care is WellDoc of Baltimore, MD, although it has started with disease management, not prevention. Like Omada, it has passed the two levels of evaluation that Kvedar says firms need to sell into the healthcare sector: clinical and economic verification. It’s no surprise both have started with Type 2 diabetes. “If someone’s A1C level is above 7, and I lower it a point,” says Kvedar, someone—probably an insurance company—will save on average $4,000 a year. “The return on investment is relatively short-term for diabetes.”

The big challenge, he says, comes with behavior intervention that might not pay off for years. Smoking is bad, and stopping is good. But when? Those ex-smokers might “pay off” by not having major health problems 15 years down the line, when they’re probably in a different health plan. “The question is,” says Kvedar, “who cares enough about that to write a check right now? That’s why so many companies say they’d rather make a cute thing to put on your wrist, make some money, sell it, and move on to the next thing.”

By posting a comment, you agree to our terms and conditions.

2 responses to “So Far, Little Proof That Digital Health Is Healthy. Does It Matter?”

  1. Rob McCray says:

    Alex – this is a very good review of the issues. Within the 15 years of my experience in the field the conversation has moved from health and tech companies not understanding why they should work together to our current situation with tens of thousands of products and apps making mostly unvalidated health claims. We need to apply the discipline of science and clinical validation to the field which otherwise could end up like nutritional supplements – unregulated, generally lacking both validation and quality systems, and financially small $30 Billion compared to about $3 Trillion in health care spending). The problem is that traditional approaches are expensive and slow. They also are ill-equipped to validate tech-enabled services to a population of subjects whose environmental and behavioral inputs are constantly changing. Drs. Steinhubl and Kvedar are key thought leaders who are conscious of these matters. The challenge which my organization is undertaking is to move to a validation/regulatory model that is more dynamic and utilizes the technologies of connected health to assess both risk factors and effectiveness associated with medical products in real world conditions and within the context of the services that support them. In this dynamic approach, both the manner of use of these products and their software components can be modified to improve the safety/outcomes profile. The validation process thus becomes one that is continuous. We have done much better at regulating automobiles than medical products. With the tools at hand, it is possible to move (cautiously) toward a an Open Outcomes Based Regulatory model for medical products.

  2. Josh Trent says:

    The article has a negligible stance towards fitness wearables. It is important to remember that in order for behavioral intervention and/or fitness wearable technology to make dramatic impact leading to a tipping point in adoption, companies must be patient and give the market what it is asking for. As stated by Joseph Kvedar: ”…stay lean, pivot when it doesn’t work, and assume when people adopt it it’ll be right.” Very well put!