Paul Carroll
I’m hoping to ask you questions about a host of issues, but let’s start with what artificial intelligence is doing in workers’ compensation.
Bill Zachry
The amount of change that’s happening right now is awe-inspiring.
For instance, these large language models can take a lot of disparate records, put them in logical order, summate them, and give the information to the doctors in a way that they can turn into a medical-legal report. One of the problems the system has had is that there is a lack of consistent quality in the medical-legal process. AI can improve the consistency and quality of the reports. I'm very optimistic about that piece of the puzzle.
I'm a former claims adjuster, and I think of the system in terms of a linear progression: Start with prevention, move on to benefit provision, and then close/settle the claim.
Starting with injury prevention, I’m an adviser to Voxel, a fascinating AI company that takes videos within store distribution centers and uses AI to determine whether employees are lifting safely, whether they're wearing PPE equipment, whether they're driving the forklift safely, and whether they're following other safety-related guidelnes. AI has so many applications.
Paul Carroll
At The Institutes, of which ITL is an affiliate, we've been talking a lot about the potential for the insurance industry to move toward a Predict & Prevent model, so I'm interested to hear more about the prevention piece of the puzzle.
Bill Zachry
With Voxel, for instance, they provide a safety score and give the safety people specific examples of opportunities for improvement. They’ll show when people aren’t wearing their safety hats or gloves. Then the safety people go back and work with the front line. One of their big clients had a 64% drop in claims frequency, and that technology is only getting better.
Some areas, like construction sites, are harder than distribution centers, warehouses, and retail facilities. The change in exposure is so rapid. But I think we’ll eventually get to those tougher areas, too.
Paul Carroll
I assume a focus for the future will be on getting the feedback to be in real time, or at least near real time.
Bill Zachry
It’s already there. The real challenge is making sure everyone is engaged so they will not only get the information but also be motivated to act on it.
The next focus after prevention is claims administration. As the U.S. has gone from a manufacturing to a service economy to a tech economy, claims frequency has been declining for decades. That trend is going to continue. But there’s been a slight uptick in severity. Severity is actually not what used to be considered severity. Severity used to be quadriplegics, burns, amputations, head injuries, hospitalizations, and things like that, which are decreasing. Now we’re experiencing claims that drag on for reasons beyond the physical injuries.
At Safeway, we had managed care nurses call every injured worker and run them through a questionnaire to gauge the risk of delayed recovery. Identifying and intervening with the at-risk employees cut our claims costs by 40%. These are employees who had experienced what are called “adverse childhood experiences.”
My experience was that the bottom 50% of the claims only accounted for 10% of my loss dollars, while, the top 3% accounted for 60% of my dollars. I found that if I identified that 3% within the first couple of weeks and set a program to get the cases closed within the year, I could cut more than 40% of my claims costs.
Doing that is one of the big opportunities in this industry.
Paul Carroll
Is this mostly an issue of staying away from litigation?
Bill Zachry
We found that if we identified and intervened appropriately with the “at risk” injured worker, they would not litigate. These were employees who responded well if you told them that you were going to take great care of them; and if we did, it was amazing how they responded. We also found that if they had already gone to an attorney, they'd already jumped off the cliff, and there was nothing we could do to help them. They were not interested in recovering and returning to work.
You have to be careful because every incentive has an unintended consequence. For instance, when you have safety programs, people don't want to report claims, because there are usually financial incentives for maintaining a low claims frequency.
One of the biggest challenges in the workers’ compensation system is how injured workers’ representation is paid. That drives behaviors and drives results. An injury's optimum result is getting the right care at the right time and getting the person healthy and back to working with zero disability. Well, zero disability means the applicant's attorney gets no money. So, time and again, I see the applicant’s attorneys maximize the disability to maximize their revenue, even though that’s to the detriment of the injured worker, who can’t return to their normal job even though everything else means they could have. If you focus on returning to work instead of maximizing disability, you get a lot more people back to work.
At Safeway, we realized that there were these unintended consequences and that you have to adjust your incentives every year to account for the problems of misplaced incentives. For example, we charged the facility for every injury. Then we offered to reduce the “chargeback” if the facility provided light or modified duties. Early return to work maximizes recovery and reduces lost time. We also did not charge the store for the hours that the employee was working while doing lighter duties. On the other hand, if the store did not report the injury within 24 hours, we added $5,000 to their charge. You have to be aware of each unintended consequence and adjust the incentives as you go along.
Another impact of AI will be on promptly approving the right care for the injured workers.
I believe many claims administrators will use AI to improve treatment based on the diagnosis. They will be approving the care automatically. However, there are potential unintended consequences. For instance, there is no study on the percentage of claims misdiagnosed in workers' compensation. For the 50% of the claims that account for only 10% of the loss dollars, it doesn't really matter. You could diagnose a stubbed toe, and the issue could be the right thumb, but if you just leave the worker alone, let them get treatment and go back to work, you'll be fine. It's that top 3%, 5%, 8%, or 10% of the claims where, if you have a misdiagnosis, then the treatment can really go in the wrong direction. We don’t have the guardrails to deal with a misdiagnosis, so how do you identify a misdiagnosis?
If the claims administrator is using AI to approve treatment based on the diagnosis, they should also use AI to determine if they are getting the results that they want to get in an appropriate time frame. And if we aren't, what's going on here that we should be looking at?
Paul Carroll
Where else do you see room for progress?
Bill Zachry
We have opportunities to reduce the time between requests and treatment approval; I think that can be automated.
We also need to use evidence-based treatment guidelines nationwide. For instance, according to the American Medical Association, most diagnosed carpal tunnel syndrome issues are not usually caused by work. They're caused by genetics and other problems, such as age and sex. But in California, where you need only one iota of exposure, it's 100% compensable, while in Colorado and Texas, it's very rarely considered compensable. I think getting a standard for compensability is one of the opportunities.
There is also the expansion of work presumption injuries for public safety officers. I truly appreciate what the public safety officers do and how they do it. They are the ones who are running toward the gunshots and toward the fire. But the comp system is being used inappropriately. It’s evaluating them as though they should be able to run as fast, jump as high, and lift as much when they’re 45 or 55 as when they were 25 or 35. That’s just not realistic, and generally, they already have medical care 24/7.
I think there’s an opportunity with gig workers based on a black cab company program in New York City. Uber and Lyft drivers are considered independent contractors in California, but if we added a 3% surcharge to everything they do and used those funds to create a program for them, we could get them covered for workers’ comp through the State Fund, or whatever.
Another impact of AI is that much of the compliance work currently done by examiners or assistants will, I think, disappear. The examiner's new focus will be on the relationship between the examiner and the injured worker, trying to find out what the injured worker needs and wants. I think that's something that can't be done by AI. I think it has to be done through creating a relationship.
In the Medicare Set Aside process, which is focused on making sure that the cost is not shifted from workers’ comp to the Social Security system or to disability, it costs several hundred dollars to have a nurse pull all the data together and generate a report. But I’ve seen prototypes where you press a button and get the report in two seconds. The cost is eight cents. The piece of the industry that generates those reports will go away.
If you look out to the future, based on technologies like CRISPR, I'm very optimistic about what we can do for severely injured people. You can even see a day when we won’t have disability.
Paul Carroll
That's a fascinating idea that never occurred to me, but I love your optimism, especially in light of all the other promising ideas you’ve presented. Thanks for the time and the insights, Bill.
About Bill Zachry
 | William M. Zachry is a board member of the California State Compensation Insurance Fund, appointed by governors Arnold Schwarzenegger and Jerry Brown. He is currently program co-chair of National Comp. He served three years as a senior fellow at the Sedgwick Institute. Zachry was awarded the Summa Comp Laude award in November 2020, the RIMS Risk Manager of the Year 2014, the CCWC Workers Compensation Professional of the Year 2016, Co-Chair AMICUS Committee California Chamber of Commerce. He is the former GVP risk management at Safeway /Albertson's, former board member California Self Insurers' Security Fund, former co-chair California Chamber of Commerce AMICUS committee, chair California Fraud Assessment Commission, Zenith Insurance VP claims, HIH (C.E.Heath) (Care America) S.V.P. claims. references. |
