By Dan Reynolds
The use of predictive modeling is growing in workers' compensation and is attracting keen interest from investors and insurance carriers.
Like any tool, experts said predictive modeling will work best when applied by sophisticated risk managers who recognize that the quality of intervention on the part of claims managers, nurses and physicians will still play a big part in getting injured workers back to work and reducing their dependence on addictive pain killers.
As models become more popular in workers' comp, they will bring into play cross-motivations.
Vendors want to grow and make a profit, on the one hand. On the other is the broader social and economic interest in helping injured workers heal and get back to work. The two motivations aren't mutually exclusive, but they do run on different tracks.
"Unfortunately what you have in this industry is a lot of people whose finances aren't aligned with getting the injured worker back to work," said Mark Pew, a senior vice president of business development with Prium, a medical cost management company based in Duluth, Ga.
"You've got a treating physician whose car payment is based on how many office visits he sees from how many patients. You've got pharmacy manufacturers who have invested hundreds of millions of dollars in drugs and who need to get their money back," Pew said.
In the battle to control drug use and other costs in workers' comp, carriers are seeing the potential of predictive models, according to a recent survey.
A Towers Watson study released in February found that personal lines insurers are ahead of commercial insurers in using predictive modeling to manage risk and create better combined ratios. But the study points to optimism among commercial lines workers' comp carriers that predictive modeling can be a useful tool.
Among standard commercial lines insurers that cover workers' comp, Towers Watson researchers found that 41 percent of them use predictive analytics to manage exposures. The researchers also found that 31 percent planned to use predictive analytics, and that 27 percent do not use predictive analytics and had no plans to use them. (See chart on Page 48.)
Towers Watson describes predictive modeling as "the application of statistical techniques and algorithms to individual risk data to better understand the behavior of a target variable based on how multiple variables interact."
In workers' comp that technique boils down most urgently to how we identify, intervene with and treat injured workers that are most likely to abuse pain medications.
A co-author of the Towers Watson report said that's where predictive modeling shows some of the most immediate promise.
"I think there is a place for data and analytics, sort of top-down when you look at things like what percentage of your prescription drugs are generic versus designer, what percentage of the time do you overprescribe the volume of pharmaceuticals needed for an individual," said Brian Stoll, a Towers Watson director based in Simsbury, Conn,
Carriers, employers and vendors are scrambling to address a national epidemic of painkilling medication abuse and addiction. Along with other health care costs, pharmacy costs are driving up workers' comp costs and creating losses for workers' comp carriers. Herein lies a business opportunity for companies who can come up with a solution, but the task isa daunting one.
Analysts with Oldwick, N.J.-based A.M. Best Company Inc. project that the combined ratio for workers' comp carriers in 2011 would be 121.5. Higher claims frequency is also adding to losses, if what The Hartford reported in its 2011 year-end results is being experienced by other carriers.
An obstacle to a more widespread and more effective use of predictive modeling is the fragmented nature of the workers' comp marketplace. Not only are there 50 different state jurisdictions with their own regulations and reporting requirements, but the quality of data on losses or factors that would indicate an injured worker's propensity to become drug addicted, for example, vary widely among companies and their clients.
"I can tell you that based on the types of businesses that we take over, the pharmacy history data show up in various forms of quality," said Todd Pisciotti, a vice president of sales and marketing for Healthesystems, a Tampa, Fla.-based pharmacy benefit management and consulting company.
There are some companies that might have good electronic records of loss histories, pharmacy use and other factors, and other companies that still have stacks of paper records, much of which are incomplete or useless.
"Occasionally there is no consistency in how they are storing that data, so now you have a whole patient profile population that is really questionable at best," Pisciotti said.
Opinions vary as to how cooperative or effective insurance carriers are in sharing and integrating their data with vendors.
Jim Andrews, the vice president of pharmacy for Healthcare Solutions, a Cypress Care subsidiary based in Duluth, Ga., said carriers are most concerned about rising pharmacy costs, and so that is where his company is putting its energy in entering and tracking data. But there are other types of data that are not so easily codified.
"The carrier is sitting on an incredible amount of data and it is not easily fed into a system," Andrews said. But if that data can eventually be corralled effectively, it can have multiple uses, another pharmacy benefit management executive said.
"It is my belief that most payers have the data necessary to perform significant analytics both from a reporting perspective as well as a predictive modeling perspective," said Matt Schreiber, the vice president of marketing with myMatrixx, another Tampa-based pharmacy benefit management company.
In November, myMatrixx launched myRisk Predictor, a tool that would give clients the ability to predict which employees are at the highest risk for abusing prescription medications.
The tool alerts carriers, third-party administrators and self-insureds when a worker may be in need of an intervention. With the release, myMatrixx is stepping into a field that is attracting a lot of attention from private equity.
"Almost every private equity group that has figured out that workers' comp is a potentially profitable area, they are all looking at the predictive analytics play," said Michael Gavin, chief marketing officer with Prium.
"They are all looking at that player that they can put money behind," said Gavin.
And those players are out there.
"I certainly think there is a competitive nature out there and I think there are companies that are starting to distinguish themselves in the market," Schreiber said.
Gavin, however, said Prium is not focused on predictive modeling at this point.
"We are not a software shop. We are data driven. We are as data-driven as any company but our focus has been on fixing prescription abuse, fixing claims and making sure that things don't run off the rails, and where the claims are off the rails they are brought back," Gavin said.
a question of quality
Gavin said it's the quality of the intervention that matters, and managers working for companies that have predictive modeling tools agree.
A simple checklist in the hands of a doctor who can probe a patient about their family history, their history of drug and alcohol abuse and other psycho-social factors can do a lot of good, Gavin said.
Although there is a lot of interest in predictive modeling on the part of carriers and other investors, the science as a claims management tool has a way to go, experts said.
The ideal would be to create a system that can pull in all factors, from durable medical equipment use to psychological profiles of patients to pharmacy records to physician quality rankings, to giving payers the best possible data in real time, and then tying it in real time to a transaction that could trigger an earlier intervention. But most admit that the industry isn't there yet.
"What has happened over time is that the number of statistical formulas has continued to grow," said Andy James, an assistant vice president of enterprise reporting for Healthesystems.
"I think the change that is starting to come is the ability to broaden the number of statistical methodologies available to anyone to apply to their data set," James said. Open source models are also part of the landscape.
There is widespread hope that modeling, coupled with intervention, can defeat the pain killing addiction enemy. "It is far better to try and predict risk going forward rather than identifying the effectiveness of the intervention," said Jay Krueger, chief strategy officer with Tampa-based PMSI. "If you can intervene much earlier in the claims cycle you get the benefit of reduced medical cost, which far outweigh the costs of clinical intervention."
DAN REYNOLDS is managing editor of Risk & Insurance®. He can be reached at firstname.lastname@example.org.
March 1, 2012
Copyright 2012© LRP Publications