Data science is a buzzy term not only in the technology sector but in the wider culture, as well. It has seeped into the common vernacular and promises increased insights and knowledge extracted from the vast quantity of data being generated every day. The use of data science in healthcare is growing, but the potentially identifiable nature of health records makes working in this space a challenge. I recently spoke with data scientists from 3 different healthcare companies on how their groups are using data to improve the quality and efficiency of healthcare.
axialHealthcare, Nashville, TN
Lindsey Morris is Director of Data Science and Analytics at axialHealthcare in Nashville, TN. Since joining the company, Morris has watched axialHealthcare grow rapidly from 8 to more than 100 employees. Focusing on pain management and opioid care, axialHealthcare leverages medical, behavioral, and pharmacy claims data to drive improved patient care and financial savings for health insurers through technology-enabled capabilities. In essence, their goal is to understand which treatments are both effective and safe in the treatment of pain. Axial is also focused on determining if treatment approaches beyond potentially addictive opioids are viable for particular patients.
A big question at axialHealthcare is, “what does safe and effective pain management look like?” The answer seeks to ensure that 1) opioids are prescribed judiciously given their propensity for causing dependence and addiction and 2) that other pain reduction therapies are considered when warranted.
Morris emphasized the unique challenges of working with health information, including the critical need for data security and privacy as well as navigating the complex United States health system. The focus for most of U.S. healthcare is on what is reimbursable by either health insurers or Medicare/Medicaid (state payers). Thus, every company trying to improve cost efficiency in healthcare must think about how their recommendations fit into the payers’ current reimbursement framework.
axialHealthcare has organized its Research and Development (R&D) group into data science and statistics/communication branches that communicate closely but have different functions. These teams work with the product team whose job is to think about the value they can extract from data insights and models to benefit customers. The R&D groups also provide support to the company’s clinical outreach team comprised of licensed clinical pharmacists and engagement specialists who work to change provider behavior and improve patient outcomes, which ultimately reduces costs on the healthcare system and protects insurers from spending on ineffective treatments.
Most of the company’s data comes from insurance claims, but some is also gleaned from patient behavioral and electronic health record data. Although the team is always focused on new data-derived models of improved care and cost savings, it is critical for the data science team to align their projects with what the market needs
“Insurers, our main clients, are very focused on short-term costs and so it’s often critical that our company frame the work in a way that indicates both short-term and long-term costs can be improved through data insights. Selling clients on the long-term cost savings can be difficult, especially if the short-term effects are increased costs to insurers,” according to Morris. This point illustrates the challenge of providing solutions that are good for the business of healthcare and also for the health of patients. Framing information in a way where payers can see the long-term savings generated from costly approaches in the short-term is critical to enacting meaningful, effective interventions.
Ultimately, the company hopes to collect its own data for two reasons: 1) access to data can be a challenge and 2) variables that may be of interest to the data science team aren’t always available in the data collected by a third party. Nevertheless, axialHealthcare’s current approach has proven effective for patients and payers.
Artificial Intelligence (AI) in Healthcare
At the 2018 Health:Further Festival held in Nashville, TN, August 27-29th, I talked with two companies working to use artificial intelligence (AI) to improve healthcare.
Change Healthcare AI group, San Francisco, CA
Change Healthcare is the largest independent healthcare information technology company in the U.S., handling approximately 60% of medical claims. Alex Ermolaev is part of a growing AI team there. The goal of Change Healthcare’s AI group is to improve efficiency and add value on top of existing data management and analytics solutions. Change Healthcare uses large, aggregated and de-identified claims, medical history, and treatment plan data from their databases to provide insights on how to increase the effectiveness of healthcare, particularly how to provide treatment that is more efficacious and economical. While Change has mostly claims data, Ermolaev said most of it is very extensive, often up to 400 pages per claim, so that meaning and insights can be extracted from the various doctor notes and other details from healthcare providers. Text from claims can be read and analyzed using natural language processing approaches to identify relevant information in the record.
Ermolaev, formerly at Nvidia, mentioned that most AI models can achieve very high accuracy (>95%) as long as the following 3 factors are available: 1) large amounts of data, 2) bigger/more complex models, 3) more computing power. “The main limit to using AI in healthcare is the lack of large enough data sets,” according to Ermolaev. This shortage of data is not unexpected given the sensitivity of personal health information and the vast privacy protections in place. Thus, companies with access to the data have a great advantage when competing in the AI healthcare space.
As AI and predictive analytics grow in healthcare, Ermolaev believes we are moving from evidence-based healthcare, where treatment decisions are based on what’s been proven effective for the population in general, to intelligence-based care where a particular patient’s medical history informs more personalized treatment.
I was surprised to hear Ermolaev mention that genetic information (frequently promoted in academic circles as the key to precision medicine) is often not required to develop personalized insights. This highlights the fact that currently available medical history, behavioral, and symptom data is often adequate in creating dramatically more effective personalized treatment plans.
While Ermolaev could not provide many details on specific projects at Change Healthcare, he did mention that they plan to announce several AI initiatives at the company’s Inspire Conference in October, 2018. Stay tuned for that announcement.
Droice Labs, Brooklyn, NY
While at Health:Further, I also met with representatives from Droice Labs on Entrepreneur Alley, a showcase area where over 70 startup companies could meet with conference attendees.
Droice Labs brings the power of artificial intelligence to hospitals. The company’s technology provides personalized predictions of how a given treatment (e.g., a drug or a medical device) will perform for a given patient. This software solution is based on a combination of the latest medical research and learning algorithms, which together analyze how a treatment has performed on similar patients in the past by aggregating data from millions of patient records and treatment plans. This allows doctors to consider all of their options in real-time and choose the right treatment.
Droice Labs has approximately 30 employees and works with providers, payers, pharma, and government clients to build applications that augment processes and improve workflows. The goal of this work is to improve the quality of care for patients while also decreasing the burden on physicians with the ultimate outcome of increasing the efficiency of the healthcare system.
The company has been around for just over a year and a half and operates, impressively, only on revenue generated (i.e., it’s completely bootstrapped). The founders of Droice have extensive backgrounds in technology and AI and take a “deep dive” approach to their projects, trying to understand the causation behind their insights and results that they communicate transparently to their clients.
The relatively small company has a very collaborative culture with employees from a variety of backgrounds—tech, healthcare (including clinicians), scientific research—that bring different but complementary perspectives to their work. “The company is structured to be very horizontal, an organizational setup that fosters the sharing of ideas among all individuals in the team,” according to Droice Labs Co-founder Mayur .
Speaking to Saxena and Co-founder Harshit Saxena (no relation), it is clear Droice Labs has a growth-focused, startup-like culture with a hunger from employees to continue to innovate and do more. The company has offices in New York and recently expanded into Shanghai, China, and has a mission that appeals to young workers that want to work for a values-driven company. “One measure of success at Droice Labs is how many people we were able to impact by our work today,” says M. Saxena. “Do the insights we develop increase the wellbeing of an extra 1,000 people? How can we improve things to increase that impact?”
Both of the founders agree with a point mentioned on the main kickoff stage at the Health:Further Festival: healthcare touches everyone. When asked what got them into using their analytic skills in the healthcare space instead of a traditional technology company, M. Saxena emphasized the common human experience of healthcare: “my thought was, if I am going to consume it, I might as well work to improve it.” He went on to say that healthcare is full of great people that work tirelessly to improve human life and that the industry needs technology to enable healthcare workers to do their jobs more easily and effectively.
At companies like Droice Labs, Change Healthcare, and axialHealthcare, the approaches may differ but the goal is the same: to improve healthcare in the 21st Century through data and insights.