compensation medical costs meanwhile have lagged behind the
medical cost index.
Conventional wisdom in the workers’ comp arena says that
20 percent of claims make up 80 percent of losses. Efforts such
as early claims reporting, medical case management, and return
to work have long been proved essential in reducing losses.
Just as predictive modeling is providing more-consistent indicators for potential mispricing than human judgment does, it
is helping to reveal potential claims problems, such as attorney
involvement, unnecessary delays in maximum medical improvement and return to work, and fraud. Predictive modeling
also improves system efficiency, leading to higher productivity
and training opportunities for claims personnel. It eventually
should improve loss reserving as well.
“In a work-related injury, the golden window is 14 days from
time of injury,” said Michael Shor, managing director of Best
Doctors Occupational Health Institute. “If claims advisers do
not engage and get as much information as possible as soon as
possible, workers are less likely to benefit from efforts to improve their care,” he added.
The Best Doctors Predictive Index does not use predictive
modeling per se, but alerts claims examiners at the time of
injury to claims needing special attention. The value of the
predictive index, said Shor, is that it helps the claims examiner
identify the injured worker with risk factors that are correlated
with a compromised medical recovery. Index variables include
questions about smoking habits, height, weight, and previous
injuries. People voluntarily share this information even though
they do not have to under the Health Insurance Portability and
Accountability Act.
Developing predictive modeling tools that identify distinctive patterns of care or sentinel events during the life of the
claim is the next logical step in this process, he added.
While Shor is positive about predictive modeling, he cautions
that “over- engineering” could mean risking the golden window.
Busy claims executives need information they can act on quickly.
“If one tries to develop a tool with 100 percent or even 75 percent accuracy, there is a high probability that the window for
corrective action will already have closed,” he added.
Many of the variables being used in workers’ comp claim
predictive modeling are the same factors that have been benchmarks for decades. These include age, occupation, filing date,
diagnosis code, specific doctors, prescriptions, and types of
therapies. Wu said claims predictive modeling is more difficult
than other applications because the models are severity based,
rather than frequency based, as with underwriting or premium
auditing. There are hundreds of possible variables from claim
and nontraditional insurance sources. That also makes it tricky
to determine the overall cost per claim, he added.
Traditional static variables come from “anything from the
claim file,” said Rong Yi, director of risk adjustment and predictive modeling practice at Milliman Inc. Injury type, job
title, hours worked, and the strenuousness of the job are also
good proxies for medical loss and return to work, she added.
Dynamic variables include medical costs, utilization—such
as number of doctor visits—medical specialties being pursued,
inpatient visits, and drug types, explained Yi. Those applying
predictive modeling to medical care already are seeing positive
results (see “Model Success Story” on Page 39). Since the employer role is so critical to prompt reporting and employee health,
Shor believes that employers in the future will be able to get better
pricing based on work forces with healthier lifestyle risk factor
profiles. “I think what we have done moves it all forward in a
dramatic way,” he said.
Quantifying Traditional Wisdom
The future of workers’ comp predictive modeling is not just exciting for actuaries and insurers. It also gives hope to those at the
heart of the system: employers and injured workers.
By quantifying traditional wisdom—that employers need to
encourage immediate claim filing to facilitate return to work at
medical feasibility—predictive modeling will play a key role in the
system’s elusive goal: stabilizing wayward cost cycles and crises.
ANNMARIE GEDDES LIPOLD has been writing about
workers’ compensation and actuarial topics for
more than 20 years. Find her musings at www.
annmariecommunicatesinsurance/ wordpress.com.
Next Generation Soutions deliver:
▪ Automated processes for high volume tasks
▪ Global coverage of all major economies
▪ A wide spectrum of fixed income, equity and alternative assets
A Moody’s Analytics Company
