Next steps for data analytics in disability and benefits plans

Data is everywhere in the fourth industrial revolution, with no industry left untouched by the momentum it’s creating, according to Nikolas Badminton, futurist and chief executive officer of Exponential Minds.

For benefits plans and disability management programs, data analytics can offer insight into how decisions are made, but the landscape is evolving as it becomes clearer how to master the billions of collected data sets. So how are plans sponsors harnessing this data and what’s in store for the future?

Reducing costs, claims

For at least five years, Niagara Casinos has been using data analytics to determine what elements of its benefits plan are adding costs and what future initiatives it can run to lower those costs, says Jennifer Stassen, the organization’s benefits specialist.

Read: 2019 Group Benefits Providers Report: How digital health is affecting the benefits industry

While it pulls data from its internal systems, it also receives third-party information, notes Lindsay Daw, the organization’s disability services manager. “They’ll cover things like our short-term disability claims, long-term disability claims, WSIB incidents, drug costs and top drug claims numbers. And from there, we’ll build our annual strategy. Our goal is to reduce costs and reduce claims.”

From an insurer’s perspective, Medavie Blue Cross also uses data analytics to help identify trends for cost drivers as well as specific benefits that are growing in usage, says Rebecca Smith, the insurance company’s director of group life and disability services.

“This informs product development initiatives and identifies areas where we may want to dig a bit deeper. We gain additional insights in terms of what is the trend of utilization, and it allows us to enhance fraud detection capabilities.”

On the disability side, one of the biggest benefits of data analytics is interventional triage, says Phil LeFevre, managing director for the workers’ compensation and disability unit at U.S.-based MCG Health. “This means identifying those claims early on that are high risk and require more active management, as opposed to a lot of claims that are low touch that you can essentially auto-adjudicate.”

Read: Just a quarter of plan sponsors review claims data regularly: Sanofi

Keeping an eye on a claim by flagging it through an analytical system and then assigning it to a case manager is important, says LeFevre, “because you can prevent that claim from going off track to begin with. And this way, we can all do more with less by making sure those low-touch claims are set to auto-adjudicate without losing the human touch where it’s critical.”

For the many employers looking only at the bottom line, it can be motivating to see the cost impact, says Martin Paquette, vice-president of TeksMed Services Inc. in British Columbia. He notes the company looks at associated costs and forecasts for a claim or set of claims that could affect premiums, then translates that into the potential impact on the bottom line.

“How can [employers] make decisions based on that and how can we present that information and say, ‘Look, it’s beneficial to be safe, to get people back to work, to offer modified duties, to accommodate employees, based on all the data we can provide them’?”

Six areas where analytics can make a difference in insurance claims data

1. Fraud: One in 10 insurance claims is fraudulent. Predictive analysis uses rules, modeling, text mining, database searches and exception reporting to identify fraud sooner at each stage of the claims cycle.

2. Subrogation: Text analytics searches through unstructured data to find phrases that typically indicate a subrogation case. By pinpointing subrogation opportunities earlier on, loss recovery is maximized and loss expenses are reduced.

3. Settlement: Analyzing claims and claim histories can optimize the limits for instant payouts, and analytics can shorten the length of claims cycles.

4. Loss reserve: Accurate loss reserving and claims forecasting is essential, especially in long-tail claims like liability and workers’ compensation. Analytics can more accurately calculate loss reserve by comparing a loss with similar claims.

5. Activity: Data mining techniques cluster and group loss characteristics to score, prioritize and assign claims to the most appropriate adjuster based on experience and loss type. In some cases, claims can even be automatically adjudicated and settled.

6. Litigation: A lot of a company’s loss adjustment expense ratio goes to defending disputed claims. Insurers can use analytics to calculate a litigation propensity score to determine which claims are more likely to result in litigation.

Source: SAS Institute Inc.

One piece of the puzzle

In general, one of the big opportunities in employee health is around integrating programs. Here, the effective integration of data plays a critical role, says Mike Kennedy, principal at Arc Health Management Solutions of Canada Inc.

“Where this is ultimately going is towards the personalization of programs. We can’t forget this part: that at the root of all the data are individuals, people who are either participating in these programs or incurring these claims or using the benefits available to them.”

Read: Employers, insurers have role in managing benefits plan sustainability

To personalize benefits programs, it’s important to understand members’ needs and interests, he adds. “Without integration between data points, it remains very fractured. And an opportunity to create that user experience that employees want — and the outcomes, ultimately, the plan sponsors require — has been missed.”

PCL Construction Ltd. doesn’t overly rely on data analytics, says Mike Olsson, the Edmonton-based company’s vice-president of human resources and professional development. “We use it as an adjunct. It’s an objective data point. It’s a source of information to then tie to other elements of what’s happening in the world. We rely on our external consultants, as well as our internals. We rely on our human resources managers all across the company to give us their opinions.”

The data does help the company set future directions for its benefits plan, he notes, but “it’s just one piece and it would be a mistake to overly rely on it.”

In numbers

58% of plan sponsors reported receiving claims data analyses identifying the main disease states in their workforce.

Among this group,19% receive the information regularly.

Of the 42% that don’t receive this information, 25% would like to do so.

Among those that do receive it, 91% agreed it helps them understand the use of their benefits plan and 65% are more likely to have specific objectives for their plan, compared to 38% of those that don’t receive claims data analyses.

Of the group that receives claims data analyses, 39% are more concerned about their drug plan compared to 25% of those that don’t receive analyses.

Source: Sanofi Canada health-care survey, 2018

Data analytics is also one piece of a larger puzzle at Niagara Casinos. Stassen notes the company looks at details beyond monthly status reports, “that focus more on the performance of our wellness programs or leading indicators.”

For Smith, data analytics is a big component of doing things better, especially when comparing historical and evolving data. It allows the industry to be more creative, she adds. “Quite often, we look at where we’re experiencing a need and that’s often based on the data we have.”

Read: Using data analytics, AI technology to curb benefits fraud

But the challenge is how to translate different data points into useful insights, says Kennedy. When it’s done effectively, it presents opportunities to intervene earlier, “and ideally, delay or prevent some of those folks from showing up in disability claims. So it’s not necessarily changing disability management, but it speaks to how we’re working with organizations who are taking that big picture approach, and ensuring the data — but, I guess, more importantly, the programming and the work that’s being done with individuals — reflects that larger and more integrated approach.”

Artificial intelligence, machine learning Yafa Sakkejha, chief executive officer at Beneplan Inc., believes the next big thing for any industry is machine learning. While the company currently uses raw data sets, she thinks it would be interesting to see how the industry could use artificial intelligence to predict things before a human or an analytics tool.

“We really want to find the canary in the coal mine, and we’ve only scratched the surface on what we can do here. If you can look at pulling in multiple data sets, and not just at our own direct claims but looking at societal trends. . . . Are we going to continue to trust humans to keep on the pulse or is there a way we can pull other data sets from around the world?”

Read: Opportunities and risks in using machine learning in benefits management

The key with artificial intelligence and machine learning is that, regardless of the model the user starts with, they should ensure they have a feedback loop, says LeFevre. “You’re logging the outcomes as you use the system. One of the biggest benefits of machine learning is a feedback loop rather than creating a model and putting it out on the market. And assuming it’s as good as it’s going to get, actually feed the outcomes data into the system so it improves over time, either validating your initial model or making adjustments to meet the data coming in.”

But this can come with problems, notes Olsson. “In order to get to machine learning and AI, you need a lot of high quantity data, lots and lots of data points. We find on our construction side, especially with safety, it takes millions and millions of data points to actually get the intelligence right. And we’re not there yet.”

There’s a lot of room for data analytics to be better, says Paquette. “For disability management as a whole, I think there are tons of opportunities for us to be better. I’d be really interested if we get to a world where AI can play a role in it.”

Alethea Spiridon is managing editor of Benefits Canada.