Using AI for eligibility file management

When a company decides to offer a health care administration platform to its employees, they’ll need to provide a certain level of information to give members the best possible experience. As a general rule, the more information a service provider about an employee’s benefits package, the better a member’s experience will be. Essential information includes name, contact information – such as email or phone number, medical plan information. Additional information could be voluntary benefits like 401K, long-term disability, short-term disability, etc. These files–which rarely come in a single file–are what the industry calls eligibility files.

c Although there are some standards in the space, they are limited and not universally used. A single typo can cause a benefits system to fail, requiring hours of detective work to find the problem. Older systems can be outdated and result in compatibility issues, especially when multiple benefit partners are in the mix.

The answer is not creating more standards (though that would help). The better solution, or at least part of it, is to enlist artificial intelligence to make processing these files easier. Using tools such as Excel, which was developed in 1985, isn’t ideal, especially since multiple files are often involved. Rather, technologies like fuzzy matching and approximate string matching can be used to process these files with little or no human interaction and give quicker results.

What is fuzzy matching?

To explain fuzzy matching, it’s best to start by describing how traditional computer logic works. With traditional logic, something can be represented with a value as true (1) or false (0). With fuzzy matching, the number can fall anywhere in between–partially true or partially false. Take tap water as an example: with traditional logic the water is either hot or cold. With fuzzy logic, you would have a scale of temperatures that runs from extremely hot to extremely cold.

Excel’s traditional, precise calculations make sense for financial spreadsheets, but using fuzzy logic processing on eligibility files allows for better matching between non-standard files or fields.

Results from using fuzzy match on eligibility files

As we prepare early for December 2018 enrollments, developing AI fuzzy matching has been a quick and easy project to dramatically cut our new client onboarding time while helping us with the wide variety of data formats and files we receive. We’ve been able to decrease the human capital required to process these files by 90 percent and have decreased our error rate by 5 percent at the same time. We still need human intervention and don’t think we’ll ever have a process that’s 100 percent automated, but that’s not the end business goal.

For companies that want to use fuzzy matching in processing eligibility files, there are a lot of open source projects to help you get started. Sourceforge and GitHub are two excellent resources. Even Microsoft Research has gotten into the game and released Fuzzy Lookup Add-In for Excel, so you Excel pros don’t need to feel left out and can dip your toes into using AI. I expect more of these tools to be released over the next couple of years as AI becomes more prevalent.

Rick Ramos is chief marketing officer for HealthJoy.