Wednesday, Aug 12, 2020
The future of big data analytics for health systems
Jean DrouinChief Executive Officer & Founder, Clarify Health
Healthcare is notoriously conservative in its adoption of technology, yet, we have seen the use of telehealth skyrocket because of the pandemic. We interviewed Thomas Graf, MD, FAAFP, Chief Population Health Officer, Ascension Medical Group and Jean Drouin, MD, Chief Executive Officer, Clarify Health, to get their perspectives on whether health systems should make a similar push to invest more in big data infrastructure and healthcare analytics.
Dr. Graf’s institution, Ascension Medical Group, is a 22 state, 100 plus hospital health system, with about 10,000 employed clinicians and a full range of facilities, from critical access hospitals to academic medical centers. Dr. Drouin’s organization, Clarify Health, is a healthcare analytics platform company that serves payers, providers, and life sciences companies. Its business applications deliver insights into opportunities to improve care, optimize networks and referral patterns, and identify unmet needs and risks of patients and populations. Its software products leverage its platform, comprised of longitudinal patient journeys for over 300 million Americans and one of the most powerful analytics tech stacks in the healthcare industry.
HLTH: Why have health systems lagged other industries in their ability to leverage big data analytics?
Dr. Graf: Health systems invest a fraction of what other industries do in technology. We have invested in the basics, like electronic health records, operating systems for pharmacies, labs, and radiology. While I have seen an explosion of information that could be used to better diagnose and treat patients, I also see that we are limited in our ability to transform massive amounts of data streams into usable and actionable information. It took providers ten years to be able to reliably prescribe beta-blockers for patients who had a heart attack in the hospital. Not because doctors did not know how, not because nurses did not understand the drug, not because the pharmacy did not want to dispense it, but because we could not find a way to keep the drug top of mind and do it in a reliable fashion. I think our recent experience with COVID-19 highlights the fact that health systems are absent really good information about our populations when they are outside of our four walls. We need to understand what is happening to patients in real-time.
HLTH: Do health systems have the data they need to leverage machine learning and AI?
Dr. Drouin: Unfortunately, electronic health record data is not the be all and end all. It is only a small picture of a patient based on limited interactions with the health system. There are many other forms of information. There is critical information to be gleaned from medical claims, labs, genomics, prescriptions, and social determinants of health. Bringing all these data together would manifest the fullest picture of a patient and then could be used to improve the care that we deliver. Healthcare has fallen behind other industries in its ability to bring together an integrated, real-time, and actionable perspective on its consumers (i.e., patients) because our data is siloed, delayed, and disaggregated. This is what makes it so difficult to use the full power of machine learning, without which we cannot be precise about predicting a person's health. Consumer companies, like Amazon, and banks, like JP Morgan, have addressed this challenge head on because understanding their consumer is vital to their business. They obtain all the relevant bits of data and information needed and build the richest consumer profile possible.
HLTH: How well do health systems understand their consumers today?
Dr. Graf: I think that depends on how you define the consumer. We understand most of our “loyal” customers (people who use our facilities a lot), moderately well. If the consumer is a patient who is in the hospital today or was in the hospital a month ago, we have their EMR data. If you define the consumer as a patient who touched the health system two years ago, then I would say we have a slightly imperfect understanding of their healthcare needs. However, if you define a health system consumer as anyone who is within its community, broadly and geographically defined, then I would say we have a huge gap in our understanding. We have a very fragmented picture, at best, of what their needs are and how they would best respond to care. Companies like Verizon and Amazon know more about me than my local health system does. Those companies know I do not exercise as much as I should and I drink too many sugary drinks, and they know the impact of that on my health. For health systems, there is an under-appreciation of the power that information can bring.
HLTH: Can data and analytics help health system improve care?
Dr. Graf: Most hospitals struggle to use medical data in a way that is actionable. Take blood pressure data, for example. We accurately measure and monitor blood pressure, but that does not mean that every patient has their blood pressure under control. Medical data is not the answer. We know that patients who have transportation issues and food insecurities struggle with disease management; depression impacts blood pressure and cholesterol management, we get that. But, what about the opportunities for impacting care that we do not see or do not know about? We need to have that data and have it in a coherent fashion.
Dr. Drouin: You can reveal a lot about a patient’s health needs when you link data from different sources, for example social determinants of health data and clinical data. With large enough data sets it is possible to train models to answer even some of the toughest questions. Our healthcare system is currently faced with many questions like, “what will care patterns look like in the new normal?” “Should we expect an unforeseen tsunami of other diseases coming on account of care being delayed?” Machine learning makes it possible to answer these questions. You can put a shock into a model, such as COVID-19, and predict the burden of disease moving forward. With social determinants of health data, the model can include factors such as job loss rates, transportation access declines, and food insecurities and therefore predict more precisely. This is the kind of a thing a bank would do to assess their forward risk and learn about their liabilities on mortgages, for example.
HLTH: How should health systems use data and analytics in the future?
Dr. Graf: I think the future lies in being able to understand the clinical status of our communities in real-time; to understand when they are getting better or worse and to determine which interventions are most needed. This is where we are headed at Ascension. At Ascension, we serve the poor, the vulnerable, and the underserved. We care not only for the folks who come to us but all the folks in our community. We want to create a healthier community by connecting individuals with the part of the network that fits their needs. We want to be proactive about care. If a patient has a procedure in the hospital and their wound gets infected at home, I want to know. I can prescribe antibiotics early to avoid admitting them into the hospital for IV antibiotics. We need to be alerted when there is a significant change in pattern, see what is causing it, and then push that information to the most appropriate individual. If a patient is depressed, their primary care physician should know and intervene. If someone has lost their job, we should connect them to social services or job training. If a patient loses their housing and no longer has a refrigerator to store their insulin, we need to know. Being proactive is much better care for the patient, less expensive, and lowers hospital readmissions.
HLTH: What types of analytics are available to help health systems today?
Dr. Drouin: Our healthcare system has fostered incredibly transactional relationships. However, we are moving to value-based and longitudinal care models, which require payers and providers to have a better understanding of the whole person and the whole community. Ultimately, health systems, physician groups, and even health plans are dealing with day to day patient care and thus need actionable solutions. They need to understand things like patient volume and service line utilization, referral flows and patterns, a patient’s individual health risks and care needs, and a provider’s cost-efficiency and clinical performance (particularly if participating in value-based care models). At Clarify, our mission is to leverage healthcare data and advanced analytics to provide these actionable insights via on-demand software solutions.
You May Also Enjoy
Thursday, Jun 25, 2020
Role of AI and Machine Learning in Chronic Disease Management During COVID-19
Anand K. Iyer, PhD MBA