Friday, October 22, 2021

Machine Learning, AI, and Platform Technology: Drug Discount Management’s Big Leap Forward

Today’s guest post comes from Scott Johnsen, VP of Product at Kalderos.

Scott discusses the advantages of using artificial intelligence and machine learning to efficiently process reimbursement data and avoid duplicates.

Learn more about Kalderos’ solutions to solve drug discount management challenges at scale during a free webinar on November 17: The Power of the Platform: How a Trusted Network Delivers Outcomes That Are a Win for All.

Machine Learning, AI, and Platform Technology: Drug Discount Management’s Big Leap Forward
By Scott Johnsen, VP of Product, Kalderos

The cost of drug discounts for patent-protected brand-name drugs is rising steeply, reaching $187 billion a year in 2020. Drugmakers know the complexity and cost of noncompliant discounts are rising steeply too—but are resigned to operating without adequate tools to navigate this changing landscape. Until now, approaches used to identify noncompliant drug discounts have largely relied on humanpower.

Humans are tasked with reviewing something similar to a Niagara Falls’-sized flow of drug discount data, sorting through disparate invoices and manually reconciling inconsistent spreadsheets. This approach is capable of identifying some noncompliance, but according to Kalderos’ internal data, likely leaves somewhere around 90% of existing noncompliance undiscovered. This leads to revenue leakage and erodes the bottom line.

Essentially, the problem is one of scale. No matter how skilled or hardworking your team is, the sheer magnitude and fragmented nature of industry data makes their task impossible. There is a limit to how much information a human analyst can process, and scaling team size only extends that ability in a ratio-based way. Further, as we add more humans it creates new problems in the form of coordination and efficiency, creating diminishing returns in the effort to scale.

This isn’t a knock on human ability; in fact, our capacity for creativity and critical thinking will likely never be matched by a machine, leaving countless interactions where humans provide insights to always be uniquely… human. But when it comes to processing data, computers have the edge in one incredibly powerful area: the ability to see all available data and make instant determinations. And machine learning—in which an algorithm uses computational methods for continuous learning—is the game changer that unlocks this ability.

Said simply, what distinguishes advanced technologies such as artificial intelligence and machine learning from other types of computer technology is an ability to rapidly self-evolve, gaining greater ability and precision. Now, instead of scaling your team's ability to identify noncompliance and revenue leakage in a ratio-based way, you can use technology to scale that ability exponentially.

At the core of the Kalderos platform lies our compliance engine, an AI-based approach that applies machine learning to the aforementioned Niagara Falls’-sized flow of drug discount data. In this cloud-based approach to Drug Discount Management, our algorithm uses a series of checks and validations to flag transactions that are high risk for noncompliance. Some of these checks are obvious pass/fail, while others evaluate probabilities.

That’s where AI comes in. Machine learning enables the algorithm to perpetually refine the checks and validations it uses to identify noncompliance. Essentially, it takes the logic and pattern recognition that a human would use to make a determination, and applies it exponentially. With access to a comprehensive and accurate data set, the algorithm can confirm where it was correct and where it was incorrect, developing not just greater detail, but greater perceptiveness.

This process is ongoing. Even after the algorithm reaches a high level of acuity, it still continues to learn. That’s because the claims data itself changes over time, creating new patterns for the algorithm to identify and apply.

A common concern when using machine learning to solve complex problems is how to ensure that the algorithm has access to good data, and can use it the correct way; put another way, how do we know that the algorithm is learning the right lessons from the right sources, and ultimately making the right decisions?

To make machine learning successful, you need a “gold standard” dataset as input. At Kalderos, we’ve solved that challenge with a multisided platform that aggregates data validated by multiple stakeholders. For instance, we offer a tool for covered entities that enables them to respond easily and efficiently to good faith inquiries from multiple manufacturers; using this tool, our covered entity partners review claims that are identified as high risk of being duplicate discounts between 340B and Medicaid, and confirm where duplicates exist.

This scenario is a “win-for-all.” Manufacturers have verification that a duplicate discount exists before disputing with Medicaid. Covered entities are able to identify where and why noncompliance exists, strengthening their programs. The algorithm uses the responses to learn what factors uniquely contribute to duplicates, further refining its skills in pattern recognition. And we’ve created multiple other points across the platform that gather, check and confirm data for continuously increasing accuracy. Ultimately, this benefits everyone who participates in drug discount programs, enabling stakeholders such as covered entities, state Medicaid agencies and drug manufacturers to identify and correct errors more quickly and comprehensively.

To learn more about how Kalderos is developing machine learning and AI solutions in the cloud to solve Drug Discount Management challenges at scale, join us on Nov 17, 2021 for a webinar titled The Power of the Platform: How a Trusted Network Delivers Outcomes That Are a Win for All.

Sponsored guest posts are bylined articles that are screened by Drug Channels to ensure a topical relevance to our exclusive audience. These posts do not necessarily reflect our opinions and should not be considered endorsements.

To find out how you can publish a guest post on Drug Channels, please contact Paula Fein (

No comments:

Post a Comment