Since the late 1990s, frequency for lost-time claims—which
occur when injured workers lose more than seven days of
work—has been on the decline and severity has been on the
rise. Accident-year 2010 figures, however, could mean a change
in trend. Frequency rose 3 percent while wage replacement
costs, which have been going up consistently since 1994, saw
a 3 percent decline in 2010, according to NCCI results.
Enter predictive modeling—the data analytics tool revolutionizing workers’ compensation for large pioneering
carriers and self-insurers that have the necessary access to
data and expertise.
Insurance predictive modeling applies statistical techniques and algorithms in considering traditional insurance and
non-insurance data to develop variables that predict the likelihood of a particular situation. Think of it as generalized linear
modeling up the expediential notch, verifying and creating new
predictive variables thanks to greater data availability and technology. A form of data mining (the analysis of data sets to reveal
new data relationships), predictive modeling techniques
include logistic regression and decision
tree analysis.
