How machine learning improves health and lowers costs

In the mid-1990s, world chess champion Garry Kasparov played a pair of 6-game matches against an IBM computer named Deep Blue. The events drew worldwide attention, with Kasparov and Deep Blue each claiming victory for a match.  

Back then, artificial intelligence was still a scary proposition. The idea that computers might outsmart us made a lot of people uneasy, and by pitting human against machine, the chess matches embodied this fear.  

During the COVID-19 pandemic, interest in online chess soared, and on a daily basis, millions of people now consult with chess chatbots to learn and improve their skills. In the chess world and beyond, computers are now more of an ally than an adversary.  

Today, we’re a lot more comfortable with the idea of harnessing a computer’s seemingly endless capabilities to run myriad scenarios to anticipate countless potential outcomes. This change in perspective is revolutionizing our approach to preventive health.  

In the health care field, the digital revolution has been thoroughly transformative. Providers use computers to store and sort millions of electronic records, analyze test results, even control the delivery of medicines to patients. Computers are an indispensable part of health care.  

On the insurance side of health care, we’re using computer power to identify and even anticipate health care needs that would otherwise require thousands of hours of hands-on work by thousands of people. Blue Cross and Blue Shield of North Carolina (Blue Cross NC) is using machine learning tools to improve the health of our members and lower costs for everyone.  

Research by the Commonwealth Fund shows North Carolina’s preventable hospitalizations among adults aged 18-64 are considerably higher than the national average. And across the country, preventable hospitalizations of adults have accounted for nearly 13% of all nonobstetric hospital stays and about 9% of hospital costs, according to research published in the National Library of Medicine. In addition to the cost of these preventable hospital stays, employers also feel the impact of a preventable loss of productivity among their workers.  

Blue Cross NC is using machine learning to reduce hospitalizations. Our strategy involves using predictive analytics to determine when and how personalized care management interventions can help members stay out of the hospital. Through our care management program, nurses can reach out to members who have been flagged as patients who can benefit from personal outreach and guidance from our registered nurses.  

This proactive approach aims to identify members who may be at risk for rehospitalization by analyzing patterns in claims and clinical data so care managers can reach out to offer guidance. Our vast store of data can help anticipate significant health events before symptoms are evident, particularly in patients with chronic conditions or comorbidities, those on multiple medications, or who show other signs of elevated risk of a major health event.  

In addition to clinical and claims data, Blue Cross NC also analyzes demographic and census information to identify social vulnerabilities that impact health. In a state that offers a diverse mix of populations and geographies, nonmedical drivers of health play a major role in our overall well-being. Predictive modeling can pinpoint people and communities that may be at greater risk of dealing with impacts of factors like food security, accessible health care, safe and adequate housing, and a host of other factors.  

Blue Cross NC has a predictive analytics platform that features 3 models, which were awarded distinguished recognition by the International Society of Service Innovation Professionals. Each model is focused on identifying different risk factors for health events and rehospitalizations. As we continue to explore the potential of these tools, we can adapt the platform to anticipate needs that haven’t even been identified as of today. This flexibility means we can keep evolving to meet the changing health needs of North Carolina.  

The irony of machine learning is that the use of artificial intelligence is actually helping us make health care more personalized. Computers aren’t replacing the human touch, they’re making it more impactful and targeted. Health care is and must always remain personal, even as technologies advance.  

There’s nothing artificial about the intelligence we employ in our digital strategy. Computers merely scour the data and organize it – it’s trained professionals who make determinations on the best ways to support our members and employer customers. The goal is better health care that is affordable, equitable, and easier to navigate. Digital tools and machine learning are crucial tools in our pursuit of that goal.