AI is transforming Medicare shopping

Between the four Medicare parts (A – D) and various supplemental coverage options, Medicare is a complex but customizable system. Medicare shoppers can pick and choose from a wide variety of plans and coverage options, which while convenient, the sheer number and complexity of these options can make the shopping process more difficult. At the same time, there’s increasing interest in online shopping from Medicare-aged adults 65+. As they become more comfortable shopping online on both desktop computers and (increasingly more) on mobile devices, the insurtech industry is exploring Artificial Intelligence (AI) and how it can help simplify the process for online Medicare shoppers. AI-based Medicare shopping solutions are expected to rely heavily on Machine Learning (ML) to automate certain Medicare shopping functions previously completed by licensed insurance agents, like needs-based analyses, product selection and plan enrollment. Over time, these technologies can learn how to recommend the best coverage options for a Medicare consumer based on their individual needs building them an insurance package they can feel confident about.

Josh Candamo is the Chief Information Officer for Ensurem, overseeing the development of the next generation of insurance technology and product innovation. He recently spoke with ADM about how Medicare and technology are beginning to find a happy place together.

ADM: How many seniors are looking for Medicare information online?

Candamo: Online information is ubiquitous, and older adults are feeling more comfortable by the day looking for information online. In 2014, when 80% of younger Americans owned a smartphone, only 50% of adults over 50 owned one. Things have changed since then. According to an AARP study published in 2020, over 81% of seniors entering Medicare eligibility for the first time own a smartphone, and there are 10,000 Americans turning Medicare age every day.

Older adults are not only researching online; they’re also buying more products online. We call this trend the “Amazonification” of consumer shopping. With this trend, older adults are also feeling increasingly confident using a wider selection of devices, including newer generations of technology. For instance, usage of voice-activated home assistants, like Amazon’s Alexa or Google’s Home device, by older adults increased a staggering 325% from 2017 to 2019. 

Older adults are becoming very comfortable searching online for information that’s important to their day-to-day lives, and Medicare is no different. In fact, with Medicare being such a complex product, it’s a likely target for online searches.

ADM: What technology devices are seniors using the most and how is AI is transforming Medicare shopping?

Candamo: Over recent years, our connectivity to the Internet of Things has increased tremendously. A new generation of electronic devices, from smartwatches to smart vehicle dashboards, now feel like a “necessary” part of our lives. The availability of these devices, as well as the existence of sophisticated vendors, have transformed traditional services, like a taxi ride or grocery shopping, into self-service consumer experiences online. The availability of these electronic devices is only going to increase with time. In fact, we are already seeing a significant amount of traffic on our website coming from smaller electronic devices such as smartphones and tablets. Our website, Ensurem.com is a great proxy for senior behavior because we focus on health insurance products for seniors. About 46% of people visiting our website during 2020 were using mobile devices or tablets.

ADM: What’s the difference in apps designed for seniors versus younger demographics?

Candamo: At Ensurem, we develop innovative ways for seniors to learn about complex products like Medicare Advantage and Medicare Supplement. Medicare itself is notoriously complicated; however, that doesn’t mean our websites and apps must be. I’ve found that for seniors, simplicity is king. When we think of simplicity, we must consider the Graphical User Interface, aka the GUI (what users see), and even more importantly, the simplicity of the general flow of the application (what users can do).

A simple GUI should use standard interface controls that people use on a day-to-day basis. Standard controls are easy to use because people are already used to their standard behavior and purpose. For instance, people know that form fields are used to type information in, and checkboxes are interactive boxes that can be toggled to indicate an affirmative or negative choice. In addition to standard GUI elements, you should always create consistency among these elements.; Consistency can be accomplished through the reoccurring use of similar controls which could include properties like colors, sizing and layouts. Control properties, when used strategically, are remarkably effective ways to create correlations between purpose and functionality. For example, on most websites, blue, underlined text can be used to easily identify hyperlinks.

When considering the flow of the application, it is also important to avoid unnecessary content or features to avoid distracting the user, as well as be clear and concise in the language used in messaging. An example would be to break up a complex task like a long application process into simpler, smaller tasks like individual questions.

ADM: What is Machine Learning, and how it can help Medicare enrollment?

Candamo: During the last few years, you have probably heard the term Machine Learning (ML). Why? Because it pretty much runs the world. Your Google searches, Facebook feed, and movie recommendations on Netflix are all powered by ML.

ML applications, in simple terms, are computer programs that learn from data without human intervention. Initial “training data” is collected, organized and used to develop “machine learning models,” from which a model "learns” and refines its rules over time. Machine learning is a form of Artificial Intelligence, a branch of computer science concerned with building smart machines capable of performing tasks that have traditionally required human intelligence to be performed. 

Did you ever wonder why over the last few years you hear about data privacy so much? It all circles back to ML applications. Companies want to store every piece of information about you and how you interact with the world. During the last two decades, with the world embarking on a massive usage of online applications and the Internet of Things, the “world of data” exploded. Data continues to be recorded, stored and categorized like never before. And with the infrastructure to quickly process data getting cheaper by the day, it makes these conditions the “perfect storm” to fuel the next generation of ML applications.

Medicare enrollment is an extremely arduous process, and requires highly trained, experienced, and licensed Medicare specialists to help consumers find the right Medicare option for their needs. We are in the infancy of new ML-based tools that will help consumers in the future. I predict we will inevitably see a heavy push of investment towards the advancement of ML-driven technology in the Medicare space. Why? The answer is simple. Machine learning thrives with complex problems that have many variables involved. When complex data is involved, patterns and associations of data can easily be overlooked by human observation, but not by ML systems.

ADM: Is Machine Learning easy to adopt in the Medicare space?

Candamo: The short answer is no.

Although there have been significant advances made in ML to make it more accessible over the past few years, ML is still a complex subject that requires a highly specialized technical skillset. In addition to the general ML complexities, Medicare is a highly regulated space and is subject to strict security standards like the Health Insurance Portability and Accountability Act of 1996 (HIPPA). However, the potential reward to our society and the consumer will certainly push the Insurtech (insurance technology) industry to make incremental usage of ML technology over the next few years, starting with consumer educational content, ultimately leading towards 100% automated enrollment tools that include needs analysis and decision making.

ADM: Will cyber agents be replacing live Medicare agents any time soon?

Candamo: Again, the short answer is no.

The live Medicare agent will not (and should not) disappear overnight. Like with many other technological advancements, new Medicare shopping technology will follow a natural evolution. Consider the transportation industry as an example. Founded in 2009, Uber took the ride-share industry by storm, but it doesn’t mean that taxis don’t exist today.  And, while Ubers and taxis coexist in the same space, they’ll soon be joined by completely autonomous, self-driving cars which are becoming more of a reality every year.  When you look at modern car manufacturers as well as the taxi and ride-share industry, you see how the landscape has changed immensely over the last decade, but legacy systems still have their place

Insurtech is still in the infancy stages of leveraging data for ML applications in the Medicare space. Nevertheless, the amount of data that is available to use is vast and increasing rapidly both in terms of volume and quality. As mentioned earlier, this is a trifecta of ideal conditions for ML to thrive. First, during the next couple of years, you will see an increasing trend in improving the availability and quality of consumer educational tools. In parallel, live insurance agents will also see improved ML-driven tools introduced for use behind-the-scenes that will help them be more effective in the over-the-phone enrollment process. Finally, you will see new enrollment tools in the marketplace going from semi-automated to fully-automated options that include every step of the shopping process from needs analysis to product selection and enrollment. 

ADM: When using Machine Learning, what can go wrong?

Candamo: Researchers and software developers should always use data responsibly, but even more so when personal or medical information is involved.

When ML engines don’t work as expected, the problem is likely with the training data or training method being used. A common problem is unintentional bias in the training data. Most commonly, data is defined as biased when it is not representative of the underlying population or problem being studied. If you use biased data, the models trained using biased data, will in turn reflect those biases. For instance, in 2020, Twitter started to investigate its image-cropping function that users complained favored white faces over black. If most training data involved white subjects, then, you could have a potentially biased application towards non-white subjects.

Additionally, while Medicare is an area tightly regulated by the Government, organizations will have to increasingly rely on internal compliance resources to interpret laws and regulations designed for more traditional services and tools to stay in the vanguard of technology innovations.

Technological advancement is nothing to be afraid of. In fact, in general it’s a big part of qualitative improvements to modern life as we know it. My advice for healthcare and insurance companies is to keep pushing forward because machine learning will pave the way for many technological advances that will help our seniors and society in general. But, as companies embrace Machine Learning for the next generation of consumer products, they ought to use data responsibly and stick to a rigorous, methodical, and scientifically sound product creation process.

About Josh Candamo

Josh Candamo is the Chief Information Officer for Ensurem, overseeing the development of the next generation of insurance technology and product innovation. He believes that your most valuable intellectual property has nothing to do with patents or technology, but everything to do with your people. He is passionate about team building and creating the right corporate culture to develop amazing software products. Josh holds a PhD degree in Computer Science from the University of South Florida, specializing in pattern recognition and image processing. He has published six first-authored, peer-reviewed publications in several prestigious International journals and conferences in the area of pattern recognition and Unmanned Aerial Vehicle navigation.