A global health crisis serves to remind us of something that’s easy to lose sight of in today’s technology-driven world: While artificial intelligence (AI) can enhance business operations by automating key processes, there remains great potential to deliver intelligent insights that drive improved decision making for businesses — particularly when facing new or unforeseen circumstances.
Over the years, predictive models have helped to drive operational efficiencies in many industries. In retail and manufacturing, for example, just-in-time inventory management helps streamline supply chains. This has improved productivity and overall performance. At the same time, it’s reduced costs.
These methodologies were improved over several decades and enhanced through the more recent growth in the use of AI to drive further process automation and support more predictive models.
The ideas are simple. You leverage resource utilization data to automate ordering, streamline procedures and make predictions. The goal is to inform business planning and operations. Today, many sectors of our economy leverage AI in this way, and society reaps great benefits.
But what happens when supply chains break down and demand spikes, creating situations predictive models could not have anticipated? The COVID-19 pandemic is showing us firsthand.
A pandemic changes everything
Just weeks into the COVID-19 pandemic, a survey found that three out of four businesses were experiencing supply chain problems. Americans flocked to supermarkets, emptying the shelves of necessities as they prepared to self-quarantine at home. And as workers in meatpacking plants became ill, many plants shuttered. Supplies of beef, pork and poultry tumbled as overall food prices increased.
Historical data provides no way to predict these challenges. The importance of human judgment and the need to adjust readily to the unexpected became evident in real time. It’s a critical lesson for health care, an industry that has its own challenges with COVID-19 but is among those sectors too important to falter.
Simply put, we need to do a better job planning for the future. Building on our current use of data and AI can further empower decision-making and prepare us to deal with challenges that lie ahead, including delivering higher quality care at lower costs, and improving patient experiences.
A lack of data requires another approach
AI depends on data — the right data at the right time. But when the COVID-19 crisis began, and attention turned to projected diagnoses, hospitalizations, capacity planning and mortality, we simply didn’t have the information needed. Early on, because the virus was new, it was often mistaken for and miscoded as influenza. We also lacked up-to-date, national-level data on the number of hospital beds, intensive care units and equipment, such as ventilators and PPE.
Luckily, existing method and tools were somewhat useful to estimate impact and spread of COVID. Epidemiologists, clinicians and actuaries used SEIR (susceptible, exposed, infected, resistant) models — the standard method of projecting spread and trajectory of a new illness — to make predictions, first using what is known of from similar infectious diseases and then incorporating incoming data.
But the rapidly evolving nature of COVID-19 has highlighted the importance of real-time data and the need to build a health care surveillance framework. AI-driven models that could enhance human judgment would be especially advantageous.
Turning to more real-time data
Real-time data is not the norm in health care modeling. Traditionally, we require complete information to make high-stakes decisions like setting health insurance premiums or measuring provider performance, and we get that from claims data.
Often, in the health care industry, it takes three to six months for “completion” of claims data (claims are complete when all services related to an episode of care have been finalized for reimbursement). The ability to access data such as diagnoses or documentation of symptoms in real-time — at least in some cases — will allow us to become more forward-looking. In situations like the COVID-19 crisis, de-identified, real-time data, coupled with stronger surveillance capabilities, would help public health officials develop more accurate forecasts, enhance AI-guided decision-making and act more quickly.
Becoming forward-thinking, using a surveillance framework, will require a new mindset and new skill set developed over time. There will be plenty of questions to answer along the way. What information should we look for? How do we connect data coming from different electronic medical record systems? Do we share data? What are the implications for patient privacy — something that’s always important?
The process may not be easy. But as we’ve seen, being better prepared is crucial.
Building on the momentum
During this crisis, we’ve discovered more about the importance of data interoperability, real-time data and surveillance. And here at Optum, we’ve seen firsthand that when people with different expertise and backgrounds come together, we can accomplish great things, and quickly.
COVID-19 is a difficult situation, but we have learned a lot by working together to navigate the crisis. We can build on what we’re learning and accomplishing.
In the coming weeks, this series will delve into areas of AI development where we’re learning more from our COVID-19 experience, including data modeling and interoperability, epidemiologic tracking and surveillance, and collaboration across health care stakeholders.
Additional stories around the industry response to COVID-19, and our efforts to confront current challenges can be found on Optum® Community Circle. You can also find more perspectives on enabling innovation to accelerate transformation in health care on our data, analytics and technology blog.
About the Author
Steve Griffiths, PhD, MS
Senior Vice President, Chief Operating Officer
Optum Enterprise Analytics
Steve Griffiths brings more than two decades of experience in health analytics management to his role as leader of the Optum Enterprise Analytics organization. Steve holds a PhD in health services research, policy and administration from the University of Minnesota, and a master’s degree in biostatistics from the University of Washington.