The new age of analytics has led to the development of a varietyof solutions that leverage internal and external claims-relateddata to enable claims handling resources to take preventativeactions earlier in the life cycle of a claim. From first notice ofincident (FNOI), and resource assignment, to the identification ofpotentially fraudulent claim activity, models are changing theclaims handling world. However, one critical component of helpingorganizations achieve maximum efficiency using predictive models is the upfront investment claim thatorganizations make to ensure that the models are implementedeffectively.

|

Modeling Basics

|

Claims predictive modeling combines internal claimcharacteristics and external third-party data to calculate amathematical score that allows claims to be segmented at FNOI andthroughout the life of the claim. Several hundred variables takenfrom background information about the employee, the work injury,external public databases, medical data, and other sources arestatistically tested to identify the 50 to 100 candidate variableswith the greatest predictive power. The final model’s variables aredetermined by leveraging a number of model development techniques(for example, correlation analysis, principal components analysis,variable prioritization, exploratory data analysis, and so on),iterative training, and testing of candidate predictive models toevaluate the statistical significance/confidence, robustness, andbusiness reasonability of the candidate variables.

|

Models for workers’ compensation can take many shapes, butseparate models are typically developed for lost time indemnity,lost time medical, and medical-only claims. The output from thesemodels can be applied in many ways including:

  • Initial routing to the most appropriate unit (low-touchstraight through processing/high-touch).
  • Initial assignment to an appropriately skilled adjuster.
  • Early identification of claims with a high propensity for fraudor that require medical management. Early escalation of claims tomore senior resources (supervisors, regional technicalspecialists).

Throughout the lifecycle of a claim, modelscan be run on a periodic basis (monthly, quarterly, and so on) inorder to consider any additional information that may help toimprove the segmentation power of the model. For example, thereceipt of medical bills and pharmaceutical information, along withthe claimant’s medical history, can be factored into the model inorder to provide medical management guidance throughout the claim.To the extent that the data indicates a potential for higherseverity claims development, the case can be reassigned to a moresenior adjuster or escalated to a specific technician. If the dataindicates a potential for lower severity claims development thanoriginally predicted at FNOI, then the skill level of the adjusterand involvement of medical professionals can be appropriatelymodified.

|

Bringing Model Scores to Life

|

In order to bring claim models to life, it is critical for anorganization to invest time upfront to help ensure that the modelsare implemented into a production environment as smoothly andefficiently as possible. With the right amount of proactiveplanning, the organization can transition from model development,to claim scoring and the delivery of high impact business workflowon the claims adjuster’s desk top. Traditionally, business rulescan be developed in a format that is easy to interpret. Instead ofsimply being called a score of 950, typically three or fourdescriptive, rationale-oriented sentences can be used to explainroughly 85 to 90 percent of the drivers behind the score.

|

In order for this to happen, it is critical to address thefollowing six areas while the models are being developed andtested:

|

1. Process/Workflow Changes Implementing aclaims predictive model requires more than just generatingmathematical equations and the resultant scores. Assessing thenumerous impacts of the model on current business processes helpsto integrate the new claim scoring methods in an effective andefficient manner. An initial assessment of potential impacts on theprocess will help to identify areas where efficiencies can begained and risks can potentially be mitigated. For example,referrals to key resources can be automated to reduce the manualburden and shorten each phase of the claim lifecycle. However, itis important to ensure that automated referrals are sent toresources with the appropriate skills and capacity to handle theclaim. The risk of inappropriate referrals can be mitigated throughthe development of a skills matrix and online calendar to guide theautomated referral workflow.

|

|

2. Business Rules and WorkflowManagement An analysis of the existing business processrules assists the organization in assessing the expected impact ofimplementing a claims predictive model. The impact assessmentfocuses on enabling the optimal alignment of current claims routingand handling the new rules that incorporate claims predictive modelscores. The new rules, which reflect lessons learned from thepredictive modeling exercise, also help the organization to enhancetheir existing approach with a strong focus on achieving, thenexceeding current industry-leading practices. For example,predictive modeling can assist with the automation of referrals tonurse case managers. By revisiting existing referral rules andincorporating model output into those rules, a company can focustheir medical resources on the most complex medical claims. At thesame time, these valuable resources can be directed away from losttime only claims.

|

3. Technology Changes A claims predictive modelimpacts business processes as well as the supporting technology.Assessing the existing technology infrastructure and how modelingoutputs such as the score and reason messages, and modeling inputssuch as data feeds from third parties, can be incorporated helps toestablish an integrated business and technical solution strategy.For instance, enhancing the integration to an existing claimsadministration system screen by supplementing select hub userscreens with model outputs and reason codes could help ensure thatnew workflows would be implemented more seamlessly. In addition toenhancing systems, improved data capture and storage in existingsystems may be required to support the model calculations andscoring engine. As part of the modeling exercise, it is notuncommon for organizations to identify valuable new data and keyperformance indicators that were not captured before, or if theywere, were captured sporadically. To the extent that data can becaptured going forward, the impact of the model can beimproved.

  • Given that many claims organizations are already consideringupgrading or replacing their core claims management capabilities,it might be the perfect time to integrate advanced analytics whilethe “hood to the car” is still open.

4. User Testing As with anytechnology implementation, testing by process owners and day-to-dayusers is essential in order to fold the new functionality intoexisting business processes more efficiently. It is not out of theordinary for this phase of the development lifecycle to command 30to 40 percent of the time and resources, particularly when modelingis involved. Test strategies, plans, and scripts should bedeveloped in accordance with the expected processes and businessrules. Aligning testing resources by key areas (by line ofbusiness, authority level, level of expertise, and so on can helpensure the appropriate level of scenario testing before the systemgoes live to minimize any potential disruptions.

|

5. Training Any major process change ortechnology implementation should be accompanied by a trainingprogram that takes the end-user environment into account. The mosteffective training programs include system demonstrations usingsample claim scenarios that are realistic and relevant to the usersof the system. Appropriate training helps to ensure that claimspersonnel fully understand, accept, and then embed the newprocedures into their day-to-day claim adjusting activities.Without the appropriate training and buy-in from claims personnel,organizations run the risk of inconsistent adherence and use of themodels.

|

6. Change Management Managing the changeresulting from model implementation is important to the successfuladoption of the enhanced processes, rules, and systems ultimatelytargeted at achieving improved business results. Having a small,yet dedicated team of resources (ideally supplemented with claimsorganization change agents) to judge the impact to the existingorganization, engage key stakeholders, and communicate strategicmessaging aids the successful adoption of the system, and moves theorganization towards full productivity throughoutimplementation.

  • The combination of developing advanced predictive models, usinga variety of traditional and non-traditional information frominternal and external sources, and then undertaking a thoroughimplementation addressing all aspects of change, can createsignificant measurable benefit across the entire workers’compensation claims organization and reduce overall claimscosts.

With the right amount of planning upfront and the coordinationof all parties involved, a claims organization can effectivelyimplement predictive modeling solutions to realize both short andlonger-term benefits. It is important to remember that modeldevelopment should never be conducted in a vacuum, though. Byconsidering changes to processes, user testing, and the opensharing of business rule impacts throughout the model development,the organization will be much more likely to end up with aneffective, productive solution, instead of a very expensivebusiness calculator that sits on adjuster’s desks collectingdust.

|

Want to continue reading?
Become a Free PropertyCasualty360 Digital Reader

  • All PropertyCasualty360.com news coverage, best practices, and in-depth analysis.
  • Educational webcasts, resources from industry leaders, and informative newsletters.
  • Other award-winning websites including BenefitsPRO.com and ThinkAdvisor.com.
NOT FOR REPRINT

© 2024 ALM Global, LLC, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to [email protected]. For more information visit Asset & Logo Licensing.