#27 - Big Data, Little Healthcare
Walmart gets more Human(a), your health plan wants your data, and more
Healthcare is strange. Just like the observable laws of physics fall out of fashion when you descend to the subatomic level, our innate and learned conceptions of how businesses, markets, and consumers function is distorted by the strange, powerful forces within healthcare. We can see these odd effects at play in this week’s discussion of health insurers not caring if you’re healthy, Lisa Suennen’s Morgue’s Law of health tech, and how some people can get their health records on their phone in 5 minutes while others need to wait for them to arrive in a paper envelope. It’s a strange place, this $4 trillion industry of ours. Let’s explore.
Things That Happened
Walmart hired Sean Slovenski, previously the VP of Innovation at Humana, to lead their health and wellness business. While this indicates closer ties to Humana, with whom they were rumored to be discussing an acquisition earlier this year, other salient details have emerged about Walmart’s healthcare future. Bloomberg reported an internal memo on the hire stated the company would be placing “more focus on our Health & Wellness business in the near term.”
Remember the nice Korean family I wrote about a few weeks ago who were charged over $15,000 for a nap and some baby formula? Well, either in response to my critique here, or more likely the original Vox and Kaiser Health News story, Zuckerberg San Francisco General Hospital has waived the bill. They apologized to the family and also wrote that the case “offered us an opportunity to review our system and consider changes.” We’ll see about those changes.
Things Worth Understanding
“Your zip code is a better predictor of your health than your genetic code.”
This quote, from public health researcher Melody Goodman, has become a touchpoint for discussing the social determinants of health (SDoH). It’s entirely sensible that where you live, and how much you earn, and how much education you have will all correspond to your overall health. Even just housing can be hugely deterministic when it comes to overall health, as has been shown in numerous studies.
New companies have sprouted up around addressing these SDoH. Cityblock Health takes an urban, community-based focus to helping with broad-based preventive healthcare in high-need populations. Circulation, the transportation startup I wrote about a few weeks ago, ensures people have reliable transportation to and from their doctor. There is much work to be done, and much to be gained, in helping those who are predisposed to worse health due to their socioeconomic status.
But startups with social missions aren’t the only ones interested in the SDoH. NPR’s Marshall Allen published a story this week describing how health plans are purchasing massive data sets on the personal behaviors, socioeconomic indicators, and other lifestyle habits of their members. The goal of this activity is to use these non-health determinants to predict a member’s healthcare utilization in the future.
It’s really quite astonishing the level of insight into a person, both factual and predictive, you can gain from publicly available and volunteered data. Lenddo, a lender, uses a person’s social media connections to evaluate their creditworthiness; the richer your interpersonal connections, the less likely you are to default.
But companies can learn a lot about you, and draw conclusions, without you even volunteering anything. A new study has shown creditworthiness can be determined from the basic data websites gather while we browse. From Daniel Miessler’s blog post on the study:
In general, there were a few things that jumped out as predictors.
iOS vs. Android (iOS users were around half as likely to default)
Emails with name in them were better
Desktops defaulted far less than mobile
People with numbers their emails defaulted more
People with old domain emails (Hotmail, Yahoo) defaulted more
People who ordered at night instead of in the afternoon defaulted more
The most interesting thing about this is how easily this stuff can be (and is) gathered from users just during a regular browsing session. Especially if you’re on the website itself that is going to making the decision.
By the time you’ve filled out your online application for a health insurance policy, the health plan could already have a pretty good idea of how much care you’re likely to consume in the coming year. In fact, Facebook had gone as far as asking hospitals for patient data to investigate the relationship between the medical and social data set (before the project was shelved following the Cambridge Analytica debacle.)
Health plans aren’t allowed to deny members coverage based on pre-existing conditions or charge sicker patients more, thus reducing the direct effect this predictive modeling will have on consumers. This is, of course, as the law currently stands.
Even if the regulation does hold up, and if we enact data protection laws that more European in their restrictiveness, that doesn’t solve the heart of the problem. As it usually goes in healthcare, the problem lies in the incentive structure. In this case, the fact that health insurers have, on average, short relationships with their members means they aren’t incentivized to improve overall health but instead to mitigate near term cost. If health plans had a longer time-horizon to realize ROI on their policies, perhaps these reams of data could be put to use identifying their most at-risk members and helping them get care to mitigate major health events down the road. Incentive misalignment creates opportunities for innovation to be used in ways that are ultimately value destructive.
We can predict who will get sick, we have the money and skills and capacity to treat them, we just need to rearrange the pieces so they can work together.
Things To Read
Christina Farr and Todd Haselton played with Apple’s HealthKit API, demonstrating how infuriatingly easy it is to get your health history on your phone. I say “infuriatingly” because none of my providers are participants in the program. Not only is the health data ready to go in under five minutes, but the needlessly complex user experience of the patient portal is gone and your health data is presented using Apple’s pixel-perfect design standards. Check out the full story (and many screenshots)
Even if your health plan knows your shoe size based on your browsing habits, and Apple can put your health records in their pleasingly modern San Francisco typeface in 3 minutes flat, health records persist as a major problem point in healthcare. Gina Kolata writes in the New York Times about a promising database that links genetics to treatments and outcomes for cancer patients, granting unparalleled and life-saving insight into opportunities for precision treatments. The major problem holding up the growth of the database: getting the treatment records. Even after patients have given their consent, “Simply getting the records delivered, in whatever format, has been a nightmare. Records usually arrive as faxes or via snail mail.”
“Healthcare operates on a path closer to Morgue’s Law, which, according to me, says that companies that try to completely replace the human touch with technology are destined for the morgue and definitely not on the fast track for adoption,” Veteran healthcare VC Lisa Suennen writes in her blog during a meditation on technology in healthcare, and on the importance of market timing for tech-based healthcare startups.
Startup Of The Week
K Health, an Isreali startup, just announced the launch of their K app in New York City, alongside a $12.5 million round of financing. -
The Opportunity
People like to look up symptoms on the internet. This is the phenomena that made WebMD into a web property commanding a $2.8 billion purchase price when KKR took them private last year.
And while WebMD’s information is typically medically verified, people tend to look further. Last week we talked about how Amazon’s Alexa, mining web information, makes a terrible doctor. And the week before that we discussed the medical pseudoscience and misinformation that abounds on the internet. Dr. Google, as it were, should probably lose board certification.
The Solution
K Health is a machine learning powered chatbot aiming to unseat Dr. Google, and WebMD, as the first choice self-diagnosis tool. It’s an app (isn’t everything these days?) where you enter basic information about yourself, answer questions about your symptoms, and then it’ll tell you what could be wrong.
More specifically, K tells you the diagnoses received by patients similar to you when they presented with the same symptoms in the past. K doesn’t diagnose. That’s an important regulatory consideration, and will also help them avoid getting physicians in a tizzy like Babylon Health. K uses machine learning models trained on "millions of medical charts” to work its magic.
The Prognosis
From a product perspective, K Health looks strong. It’s a clean, ad-free interface that, provided it actually works, solves the customer’s problem of judging the severity of medical symptons; “am I dying or is this just a head cold?” It’s also free to use, which helps immensely.
The business model is compelling as well. I can’t say I’m privy to the investment deck that spells out revenue strategies, but when you’ve got a consumer technology guy like Allon Bloch as CEO (he founded Wix and online car marketplace Vroom) it’s a safe bet K Health is trying to become the default first step for consumers when navigating a health issue. CVS Minute Clinic, Walgreen’s, Walmart, and urgent care centers in general are all trying to be that same first step. K Health has the advantage of being both free and more convenient. It won’t actually treat you, or even tell you exactly what’s wrong, but it will help you figure out if you even need to bother getting out of bed.
If K Health’s service is compelling to consumers, there are many, many ways to leverage those users. They could go the traditional route and earn advertising revenue, but I think they’ve got a bigger target. I’d imagine they want to be involved in controlling where those patients end up going to the doctor. It could be information on which local clinic has the shortest wait, which doctor has the best reviews, or perhaps even a referral to a physician service of their own construction. The optimal value match for an on-demand self-diagnosis tool is an on-demand physician, so it would make sense for K Health to build their own telehealth service with the app as their patient funnel. The larger point is that, once you’re the first stop for customers on their health (read: purchasing) journey, you control your own revenue destiny. It worked for Amazon.
Things I Listened To
My son, who is just now 16 months old and starting to develop his own opinions, paused his wobbly jaunt down the sidewalk at the sound of bagpipe music lurching through the door of a Scotland-themed store. After considering whether it was music, or just someone doing something heinous to a cat, he made his decision and started dancing.
So this week, my musical recommendation is dedicated to my son. Elliott, there are far better things to dance to than bagpipes. Try The Go! Team’s 2015 album The Scene Between. (spotify)
While I Have You…
The Healthcare Handout is a labor of love. That’s why it’s entirely free to read. All I ask in return is that if you find it interesting or informational, share it with a friend or colleague. Help spread the word and you’ll make this healthcare writer eternally grateful.
Have a great week,
- Isaac Krasny