#26 - Is your AI vertical? Try adjusting your screen

Verily’s compelling business model, Ian Read answers the phone, and more

This is the third week in a row I’ll be taking a longer look at some aspect of AI in healthcare. Is it just a hype cycle? I really don’t think so. AI, and more specifically machine learning, computer vision, natural language understanding, and other modalities are already making our lives different, and sometimes actually even easier. And while many in healthcare view every bit of technology with skepticism, this one is here to stay. AI is the new electricity, and you must do your darnedest to understand what it is, how it works, and how it will change your life. Onward!

Things That Happened

Cerner reached for their wallet. The EHR giant invested $266 million into the parent company of Lumeris, who provide expertise and technology to help healthcare organizations move to value-based care. Lumeris and Cerner will work together to create and scale technology for the Medicare Advantage market. 

The investment is a sign that Cerner sees value-based care, and specifically Medicare Advantage, as a big growth opportunity, and also an acknowledgment at the the workflows for providers in these arenas are sufficiently different from those anticipated in off-the-shelf EHR systems. This exact gap in the market is what forced companies like Iora Health to develop their own in-house EHR. 

Things are looking brighter for Obamacare exchanges. Bright Health, the narrow-networks insurance startup founded by former UnitedHealthcare CEO Bob Sheehy announced this week they’ll be expanding their individual marketplace plan offerings into five new cities for 2019. Oscar Health, the more famous narrow-networks health insurance startup, is doubling the size of their individual market footprint in 2019. And Centene, the least famous narrow-networks health insurance company, is also expanding in individual markets by way of a $3.75 billion acquisition of NY based Fidelis Care. 

It seems these companies, which were purpose-built to operate in the individual markets, remain bullish on growth despite the Trump administration’s kitchen-sink level obstructionism around the ACA. For more on Centene, take a look at this superlative piece describing their path to success in the marketplaces, from the team at Bloomberg. 

(a deep-learning generated representation of a pharma executive on the phone)

Ian Read answered the phone. The Pfizer CEO, after his company raised prices on about 40 drugs and found themselves in the crosshairs of the president’s tweeting fingers, spoke first with HHS Secretary Alex Azar, but requested a call directly with the president. After the resulting call, Trump tweeted that Pfizer would be “rolling back price hikes” and Pfizer confirmed, with the caveat that the price hikes were only being deferred until 2019, or until the president’s drug blueprint was implemented, whichever came first. Note: My money is on 2019. 

Ed Silverman writes in STAT, “Trump can threaten all he wants… But his ranting simply reveals an inability to accept that delivering on his promise will be much harder than he realized.” Check out Silverman’s piece for a deeper explanation of why lowering drug prices isn’t as simple as it should be. And if the White House’s strategy is to bully drugmakers on an individual basis, Trump might want to invest in one of those ergonomic headsets for his phone…

Jonathan Rockoff@jonathanrockoff

So far this year, pharma companies have taken 3,653 price increases on 1,045 different drug products, according to Raymond James & Associates. Among them: a spray Ambien whose price was just increased by as much as 843% https://t.co/b1Pr0itldP

July 14, 2018

Things Worth Understanding

This week, ResMed announced a collaboration with Verily (the Alphabet subsidiary) to develop “sleep apnea treatments and connected health products.” Last week I discussed Microsoft’s solidifying venture into the healthcare AI space, and the week prior it was IBM Watson Health. So I’ll take this opportunity to explore perhaps the most advanced player in the healthcare AI universe: Alphabet. 

When it comes to commercial AI, Alphabet is arguably the leader. Alphabet has, throughout their many subsidiaries, placed an emphasis on developing AI technologies and integrating them directly into new and ongoing businesses. This means they have the most experience with commercialized AI, and probably the best engineering talent. 

Verily, where Alphabet is focusing most of their healthcare ventures, is using partnerships to make vertical AI plays on multiple fronts. Bradford Cross, founder of the AI focused VC firm Data Collective, argues that vertical AI startups are the most likely to succeed because they deliver the value of AI directly to the customer. He describes vertical AI startups as meeting these four criteria: 

  1. Full stack products: they don’t just create the AI models, but also the interfaces that deliver the value to the customer, and that enable them to collect ongoing data. 

  2. Subject matter expertise: they have connections within the target industry and understand its idiosyncrasies

  3. Proprietary data: they collect available data, and enrich it, and then improve their data set by collecting new data geared specifically to their needs.

  4. AI delivers core value: the AI being developed doesn’t optimize an existing business, but serves as the core value driver of the business 

Using ResMed as an example: Verily identified a partner in a target disease state, created a partnership to co-develop products, and will now likely embark on data collection in ResMed sleep apnea devices in order to create machine learning models that will improve detection and treatment protocols. In terms of the vertical AI model, Verily brings the AI expertise (and likely some funding), while ResMed brings subject matter expertise, access to proprietary data, and the ability to bake the AI into a full stack product where AI can create greater value for the customer. 

Verily has done the same vertical approach in diabetes through partnerships with Dexcom and Sanofi. Verily was involved in developing Dexcom’s newest continuous glucose monitoring device, from which they’ll no doubt gather data on diabetics. Through their partnership with Sanofi they’ve created Onduo, a company that develops applications to help diabetics manage their disease. Through the combination of the two, Verily is providing the AI layer for a full-stack diabetes tool and gaining proprietary data as they go. 

This means that in each new space into which it embarks, Verily will have extremely defensible AI positions. Through partnerships with existing market leaders they gain access to what is likely the best proprietary data, the strongest subject matter expertise, and make themselves indispensable to their partners by delivering greater value to the end customer with AI. This creates what Cross calls the “data flywheel,” where Verily can use proprietary data to enhance user experience, which in turn gives it more and better data. It’s a virtuous cycle that makes it increasingly difficult for competitors to measure up. 

Compared with Microsoft and IBM, Verily is not just the furthest along, but has the most compelling business model. Rather than consulting projects or niche services, Verily is essentially placing itself at the core of new businesses, ensuring it reaps maximum financial rewards from any gains, and can also take learnings and innovations and use them throughout their businesses. 

For a much deeper dive into all of Verily’s, and Alphabet’s, healthcare projects, check out this research report from CB Insights. 

Things To Read

“Although she is wonderfully helpful with basic questions,” Katherine Ellen Foley and Youyou Zhou write about Amazon’s Alexa in Quartz, “…her knowledge of the complex medical world is limited, and Alexa generally only serves up one answer. You can’t flick between a dozen browser tabs to determine which strain of tropical parasite that is (probably not) besieging you.” This first story in a series from Quartz on AI in healthcare and medicine discusses the effect that voice assistants are having on the self-diagnosis problem in healthcare. Most notably, by attempting to provide concise answers, Alexa and the skills developed by third-parties can provide definitive answers without fully understanding the question, and worse of all, premised on unverified and incorrect information. 

On the worst staffed days at an average facility, the new data show, on-duty personnel cared for nearly twice as many residents as they did when the staffing roster was fullest,” Jordan Rau writes in the New York Times. While skilled nursing facilities aren’t one of the major areas of focus in the struggle to improve healthcare delivery systems, their 1.4 million residents consume a significant share of healthcare resources. And when these facilities go understaffed, it makes it more likely these residents will end up back in the doctor’s office or the hospital. 

Startup Of The Week

Higi, a Chicago-based startup, brands itself as a population health enablement company. They also just raised $21 million in venture funding from reasonable healthcare investors 7wire Ventures and Flare Capital Partners. To uncover what population health enablement actually means, we’ll need to start from the beginning of Higi’s story.

Higi was born back in 2013 as a social health platform where users would compete with each other around their higi Score, which was calculated based on their biometric information, social media interactions, and their fitness tracking. Think of it as a credit score, but judging your fitness and lifestyle choices. Sounds like a hoot. Perhaps the most important thing was that higi was very cool with the youth - they launched at south by southwest and even had Lupe Fiasco as their creative director. 

But, as Higi and so many others discovered, even gamification couldn’t get people interested in regularly monitoring their health. While some direct-to-consumer healthcare startups simply die off, Higi managed to pivot, as many other health startups do, to catering to enterprise clients. By modifying their social health platform, and using their 11,000 health kiosks spread throughout the country, Higi is now uniquely positioned to make population health a little bit easier by allowing users to update their blood-pressure, pulse, weight, and BMI through self-service kiosks. If you’re looking to reduce friction, a free no-wait self-service kiosk in a high-traffic location is nearly as good as it gets. 

This round of capital will help Higi continue to hone in on this new population health angle, which seems to align both with their strengths of broad-based biometric data gathering, and with the wider healthcare push towards managing health outside of the traditional healthcare system. 

Things I Listened To 

This week’s post was written while listening to Sketches of Brunswick East, the third of five albums released in 2017 by the Australian psychedelic rock band King Gizzard & the Lizard Wizard. While they usually have a heavier vibe, KG&TLW are subdued and jazzy here. (spotify)