Population health brought evidence-based medicine to the family doctor's office, vastly improving the overall quality and consistency of care in this country, but in doing so, it forgot about the individuals who make up the population.
To standardize care, population health uses a rules-based approach. Treatment protocols, developed and endorsed by medical societies and others, guide a range of intervention decisions by physicians, population health leaders, and others. If it's an activity or treatment aimed at improving a medical condition or preventing a medical condition from progressing, there are rules defining when and how the intervention should be used. This includes rules around preventive screening and annual exam intervals, pre-defined triggers for care management programs, pharmaceutical pathways, and many more types of interventions.
These rules are based on averages across a population, and that's the problem with population health: Nobody's average.
Our system of approving new medications relies on the same population-level average approach. Without question, pharmaceutical interventions--FDA-approved drugs--are held to the highest standard of all interventions, yet about half of FDA-approved prescription drugs do not work as intended for individuals. For some conditions, it's much higher--about 75 percent for cancer drugs.
Population-level averages are leading us to waste hundreds of billions of dollars a year on interventions that have no value whatsoever to the individuals they are intended to help and on the downstream medical costs resulting from ineffective interventions.
This is not new information for the many dedicated population health leaders and healthcare providers tasked with implementing the flawed system. So why is now the time to change?
A convergence of needs, resources, and tools is creating the perfect opportunity to put the individual first--and do so on a population-level scale. The number and cost of interventions is rising exponentially, making the already painfully high cost of healthcare in the United States, and relatively poor health status, unbearable. Also, an exponential increase in data and investments to deal with it shows no signs of slowing, leaving healthcare leaders drowning in data and searching for a return on investment.
Meanwhile, breakthroughs in machine learning and cloud computing are unlocking the value in data to ease the pain caused by escalating interventions. Machine learning is the pinnacle of analytics. It goes a step beyond predictive modeling, which can forecast what will happen next, to predict what will happen in the future under specific sets of conditions.
After all, what's the point of predicting the future if you can't change it?
Machine learning can reveal cause-and-effect relationships in data, not just correlations, making it possible and practical to predict many future "what if" scenarios, to compare outcomes, and use that knowledge to optimize decisions, actions, and investments. GNS, a pioneer in the precision medicine movement, has been applying its patented machine learning technology for more than 15 years. Today health plans, healthcare providers, pharmaceutical companies, researchers, and others are using machine learning to address many costly and persistent healthcare concerns and to inform essential operational decisions.
To improve medication adherence, for example, population health rules (HEDIS and NCQA delineated differentiators) use 80 percent as the intervention threshold. The effect is this: Members with adherence below 80 percent are typically selected for care management interventions. Members with adherence at 80 percent and above get no intervention.
The 80 percent rules-based threshold is blind to the potential value intervention holds for each individual within the population, like Ethel, a hypothetical member who illustrates the knowledge blind spot. Ethel has a chronic condition and has been prescribed medication to manage it. Her adherence level is 82 percent. The rules-based population health approach would not select Ethel for an intervention to improve her adherence.
When we use a data-driven, machine learning-enabled approach to look at many more data points, it is possible to see and act on the potential value of intervention at a person level. For Ethel, and every other individual within the population, GNS computer models answer questions about risk, efficacy, engagement, and intervention:
· What is this person's risk for an ER visit or hospitalization at their current adherence level?
· How much do we have to improve this person's adherence to prevent the health event?
· Will this person engage to change their behavior?
· Which intervention will be most effective for this member?
The models reveal that Ethel is at high risk of a health event related to her medication adherence level, and a 10 percent improvement in her adherence would prevent a complication predicted to cost $10,000. On this basis, the data-driven approach would select Ethel for an intervention to improve her adherence. The value to Ethel is enormous. The approach also delivers significantly improved results across the population.
Improving medication adherence is just one example of how data-driven approaches are transforming care management and care delivery by enabling person-level decisions on a population-level scale. Machine learning is being applied to prevent pre-term births, extend progression-free survival for patients with metastatic cancer, prevent metabolic syndrome from developing or progressing, and more. Data-driven solutions also eliminate knowledge blind spots affecting business decisions, such as by revealing insights about individuals composing new member populations.
Population-level averages were the best we could do 20 years ago, but it's time to break the rules of population health to put the people, not the population, first.
Colin Hill is chairman and CEO of GNS Healthcare.