Unlocking The Power of Data Science In Healthcare

Posted on January 29, 2014 I Written By

Kyle is CoFounder and CEO of Pristine, a VC backed company based in Austin, TX that builds software for Google Glass for healthcare, life sciences, and industrial environments. Pristine has over 30 healthcare customers. Kyle blogs regularly about business, entrepreneurship, technology, and healthcare at kylesamani.com.

Vinod Khosla, Founder of Sun Microsystems and Khosla Ventures, recently stated that “in the next 10 years, data science and software will do more for medicine than all of the biological sciences together.”

The rise of population health and healthcare analytics companies aligns with Khosla’s claim. There are hordes of companies implementing healthcare analytics and helping providers identify at-risk populations to engage in proactive care. Despite their efforts, most of the analytics companies have been struggling to help providers actually improve outcomes.

Why?

Because data science in and of itself is meaningless. Effective data science can only provide insights. The challenge is in acting on insights provided by data. This is a widely acknowledged problem that every data science / analytics company faces; this problem has been particularly difficult in healthcare where a backwards culture and incentive structure have skewed the system towards complacency and volume rather than proactive care and value.

In healthcare, the actionability and effectiveness of data science hinge on communication between providers and patients, and on patients’ ability to act on those insights. There are a few methods of provider-to-patient communication and actionability:

At the point of care (in person or virtual visit) – providers have been educating patients at the point of care since the dawn of the profession. With advanced data analytics, providers can give more accurate, more customized education during the encounter. But the problem is that patients must act on that information at home when the doctor isn’t looking over the patient’s shoulder. Patients consistently fail to do what providers have asked them to do. The problem here is that the patient education and actionability based on education are intermediated by time and (lack of) context. Patients simply forget or are unwilling to do what their providers ask them to do in order to better care for themselves. Patients aren’t being educated in the right context. Point of care education won’t encourage patients from smoking the next cigarette, taking their meds on time, or skipping cheesecake at the office party.

Patient portals – the federal government has mandated that providers enable patients to engage with providers via patient portals. The basic premise of this mandate is that with access to their own health information, patients will take better care of themselves. Patient portals have some potential to empower patients to learn about their conditions at home and investigate conditions in more depth, but they don’t solve the context problem. Patient portals won’t do anything to help patients order a salad instead of a hamburger.

Messaging and notifications – this is the least explored, least understood, and in my opinion, the most potent communication channel to impact patient behavior. Automated notifications on iOS and Android can be presented contextually provided the device has contextual data to present notifications. Context is king. We live in the age of context. As devices learn more about their owners, devices can present contextual information to help change behavior. If your smartphone (or Google Glass, Jawbone, iWatch, etc) knows that you’re about to smoke a cigarette, it can automatically connect you with your husband/wife so that they can yell at you. If your device knows that you’re out at a steakhouse for dinner with business guests, it can remind you to order grilled chicken instead of a fried steak. The number of opportunities are endless.

To provide a better sense of the power of context, let’s examine Google and Facebook ads. Facebook ads are anything but contextual. When I’m scrolling through my news feed, I don’t care about the latest Hobbit movie, some new workout shake, or Dell’s newest laptop. I logged into Facebook to check out what my friends are up to, not to learn about the Hobbit or a laptop.

But when I Google “flight from Austin to New York January 18th” there’s a huge probability that I’m already committed to spending several hundred dollars to fly to New York, get a hotel, and spend money in NYC. With that search, only relevant advertisers – airlines, taxis, hotels, and local NYC attractions – will bid for my attention; I’m not going to see an ad for The Hobbit when searching for for a trip to NY.

This sense of context is reflected in Facebook and Google’s click through rates (CTR). 1-3% of all Google searches result in the user clicking on an ad. Between .01-.3% of FaceBook ads are clicked on. Google is measurably 10-100x more effective than Facebook. That’s the power of context.

There’s nothing wrong with emailing patients PDFs and interactive digital education tools after an encounter; there’s nothing wrong with patient portals and BlueButton. All of these communication channels fall short in that they don’t take advantage of real-time two way contextual communications. All of these channels are intrinsically one-way and lack context.

Books were the the first few-to-many communication channel. Then newspapers and magazines. Then radios. Then movies and TV. The defining characteristic of the Internet is that it is the first to enable two-way, many-to-many communications. The federal government’s healthcare communication model is fundamentally based on 20th century communication strategies. The power of data science will drive meaningful changes in patient behavior only when communication strategies leverage 21st century communication models.