Editor’s Note: Paul Bradley, PhD, is the Chief Data Scientist at ZirMed, a company empowering healthcare organizations to optimize revenue and population health with an end-to-end platform of cloud-based financial and clinical performance management solutions.
Predictive analytics – using data, statistical algorithms and machine learning techniques to identify the likelihood of future outcomes based on historical data – has become a staple of many industries.
Brick-and-mortar retailers use it to determine inventory levels based on factors such as weather models and estimated demand. Companies like Amazon and Netflix use it to offer suggestions on additional products you might like. Even sports franchises use it to predict the potential success of players, which was most famously depicted in the book and movie Moneyball.
Now healthcare providers and payers have started using predictive analytics to help them with clinical challenges such as population health management and closing care gaps as they make the transition to value-based care.
Early innings for healthcare
Healthcare has lagged other industries because of the exponentially greater complexity of analyzing the factors that contribute to human health versus predicting which ballplayer will deliver the best run production. But today a combination of vast amounts of data now being available thanks to the move from paper charts to electronic health records (EHRs), along with the explosive growth of computing power and storage, has made it possible to predict the care actions, medications, devices, etc. that will deliver the best outcomes.
There is another area, however, where predictive analytics can make a significant contribution to the success of healthcare providers: helping them to get paid for all that care they’re providing.
Consider self-pay patients – a category that is growing rapidly thanks to developments such as high-deductible health plans (HDHPs). When a patient comes to a provider and presents as self-pay, the level of risk automatically increases. After all, hospitals, laboratories, physician practices and so forth are highly likely to be reimbursed if they submit a clean claim to a health payer.
Not so with individual patients. But with predictive analytics, the provider can run the patient’s information through its predictive models and determine the likelihood that the patient will pay his or her portion. Armed with this information, providers can take steps, such as collecting from patients up-front, to help them eliminate this very common cause of bad debt.
Predictive analytics can also be used to ensure providers are receiving all the reimbursement from payers to which they are entitled by looking for missing charges. For example, a claim may be missing the charge for a pacemaker device, even though it includes the charge for the procedure to insert the device.
Data to develop a stronger lineup
Currently, providers tend to rely on rules-based technology that says if A and B occur then C should happen. Using our previous example, one such instance may be “if a charge for a device insertion is on the claim there should be a charge for the device on the claim.” Rules-based technology may catch the missing pacemaker device. For this to happen, however, the rule would need to be specific enough to relate the pacemaker insertion procedure with the pacemaker device. As it is stated, the rule could be satisfied if the claim included a low-dollar charge for a lead-wire device, while still missing the high-dollar pacemaker device. In this case, the rule was satisfied, but the provider still misses payment for the pacemaker device. Not a great situation for the provider.
Like a baseball manager looking at stats while creating that day’s lineup, predictive analytics looks through historical data to find correlations between the 9,000 to 16,000 individual elements that make up a single healthcare encounter. In the case of the pacemaker, because the charge for the pacemaker insertion procedure ordinarily occurs in conjunction with the charge for the specific pacemaker device, predictive analytics will expect the device charge to be there. If it is not, that instance is considered an outlier and can be escalated to the proper resource for follow-up.
Implementing these types of predictive analytics typically yields a 4:1 to 5:1 ROI. In a healthcare setting that could mean millions or even 10s of millions of dollars added to the bottom line.
Rewriting the rules
The other great advantage predictive analytics brings is its flexibility, especially when combined with machine learning. Unlike retail and manufacturing, healthcare is an extremely dynamic industry; every human differs from every other human, and changes in knowledge and treatment occur constantly. What is true, or the best practice, today may not be so three months for now.
With rules-based technology, you must know or anticipate everything as you build the analytics if you want them to be included. With machine learning, changes over time become factored into the algorithms automatically. The technology “learns” over time so it can always reflect the current state as it occurs. The result is decisions are based on today’s rather than yesterday’s conditions, helping you drive better financial outcomes.
For the win
Many industries have discovered the truth to the saying “the past predicts the future.” Now it’s healthcare finance’s turn.
By creating algorithmic models that use past data to predict outcomes, then comparing those answers with the actual known outcomes, healthcare organizations can use predictive analytics to help them improve financial performance by cutting costs, reducing bad debt, and charging properly and completely for their services. And that’s a home run.
