It’s easy to become trapped in assumptions. One such assumption is that predictive modeling is exclusively achieved by scouring data using advanced mathematical computations. While that is the generally accepted method, and a valid one, it need not be the only method. Extraordinary outcomes and substantial savings can also result from simpler initiatives.
While more commonly used in other industries, predictive modeling is gathering interest in workers compensation. Predictive modeling is a process used to create a statistical model of future behavior. It is the area of data mining and business intelligence concerned with forecasting probable outcomes or trends.Multiple methods of testing assumptions and uncertainty are involved while looking for patterns in the data. Statistical modeling uses advanced mathematics to find correlations, look for consistent causation, develop a theory — apply it, validate it, adjust the theory and continually retest it. (WCxKit)
Answers are sought for basic questions using predictive modeling. If X is true, then what is the probability Y will occur? Or when Y occurs, what are the factors that could have predicted it?
A familiar example of how predictive modeling is used is auto insurance companies taking into account potential driving safety predictors in the data such as age, gender, and driving record. The predictions are not guesses. Instead, indicators are found using huge amounts of data and are based on the idea that consistent historic behavior found in costly claims is a predictor of future claims with similar conditions. Auto insurance premium costs are rated by this actuarial intelligence applied to the data.
Similarly, when predictive modeling is applied to workers compensation, the objective is to identify claims likely to be complex and costly based on historical data. The goal is also to identify those claims early so damage control can be implemented such as focused claim and medical management or early settlement. Regression analysis and other advanced methods of statistical mathematics are applied to the data to find key indicator data in those claims.
Nevertheless, advanced statistical modeling is not for everyone. Studying the data in this way requires huge amounts of data to achieve statistically significant results. Experts must be contracted and financial investments made. Such efforts are well founded, encouraged and ultimately lead to refined intelligence in workers comp. Still, there are lesser, yet valid, achievements to be made.
Notably, most payer organizations have an untapped predictive resource: internal wisdom. Claims adjusters, nurse case managers, and medical directors all know their claimant population and instinctively know what claims are likely to be problematic. Moreover, many organizations utilize the three-point contact methodology where vital data are collected about the claim that can be predictive. All this information should be collected in data form and structured procedurally to strengthen claim management functions.
Another way to find predictors in data is to leverage workers comp medical research as a guide. Edward Bernacki, MD and his team at Johns Hopkins published a study in the Journal of Occupational and Environmental Medicine in January of this year describing Cost Intensive Physicians. (Bernacki, et.al. “Impact of Cost Intensive Physicians on Workers Compensation” JOEM. Vol. 52. No. 1. January, 2010) Using five years’ closed claim data from the Louisiana Workers Compensation Corporation, they studied claims that began with reserves less than $15,000, but migrated to reserves of +/-$50,000. Of those claims, 3.8% of physicians involved were responsible for 72% of the costs. Their information about cost intensive physicians can be applied to identify predictors.
Cost intensive physicians, as labeled in the study, were those who had higher medical costs, longer medical treatment duration, longer claim durations, and higher indemnity costs. Therefore, one can conclude that identifying, avoiding or managing the cost intensive physicians is one way to contain costs. Look for those features associated with specific physicians in the data along with other characteristics highlighted in the study.
The Bernacki study also named certain injury types and procedures high cost predictors. Those injury types or diagnoses that do not have clearly defined treatment pathways are often problematic. Whereas a fractured tibia has a predictable treatment path, injuries of joints and back strains do not, leaving a wide berth of treatment options and opportunities for abuse. Monitor the data proactively to isolate injury types and procedures identified in the study and manage them aggressively.
Another study, “Long-term Outcomes of Lumbar Fusion Among Workers Compensation Subjects: An Historical Cohort Study”, found at http://www.ncbi.nlm.nih.gov/pubmed/20736894. This study concluded, “Lumbar fusion for the diagnoses of disc degeneration, disc herniation, and/or radiculopathy in a workers comp setting is associated with significant increase in disability, opiate use, prolonged work loss, and poor RTW status.” Clearly, these conditions, when spinal fusion is the selected option, are predictors of complexity and cost. (WCxKit)
Identifying and naming predictors using the knowledge from professionals and from research is a simple, yet valid approach to cost control through predictive methods. Call it “predictive modeling-light.” Monitor the data concurrently and continually to identify claims containing combinations of data that portend risk and focus on those. While the process is not sophisticated or complex, it will structure claims and medical management procedures, making them more proactive, effective, and replicable.
Author Karen Wolfe, BSN, MA, MBA, President/CEO, MedMetrics®, LLC. Karen is founder and president of MedMetrics® LLC, an Internet-based Workers Compensation medical analytics company. She applies her medical knowledge to gathering, understanding and applying Workers Compensation data to the operational process. MedMetrics imports, integrates, and analyzes its clients’ medical billing and claims level data. MedMetrics uses several tools such as Predictive Intelligence Profiling and Medical Provider Performance Assessment to gather and analyze data. Contact: Phone: 541-390-1680; Karenwolfe@medmetrics.org; www.medmetrics.org.
TRY COMP CALCULATOR: http://www.LowerWC.com/calculator.php
MODIFIED DUTY CALCULATOR: http://www.LowerWC.com/transitional-duty-cost-calculator.php
JOIN WC GROUP: http://www.linkedin.com/groups?homeNewMember=&gid=1922050/
GET NEWSLETTER: Workers Comp Resource Center Newsletter
Do not use this information without independent verification. All state laws vary. You should consult with your insurance broker or agent about workers comp issues.
GET NEWSLETTER: Workers Comp Resource Center Newsletter
Do not use this information without independent verification. All state laws vary. You should consult with your insurance broker or agent about workers comp issues.
©2010 Amaxx Risk Solutions, Inc. All rights reserved under International Copyright Law. If you would like permission to reprint this material, contact Info@WorkersCompKit.com.