Chubb 4D: Power of Predictive Analytics

Transcription

1 Chubb 4D: Power of Predictive Analytics The State-of-the-art Progresses Claims Management

2

3 Predictive data analytics is coming out of the shadows to change the course of claims management. A new approach, Chubb 4D, provides the tools and expertise to capture and analyze both structured and unstructured claims data. The former is what the industry is used to the traditional line-item views of claims as they progress. The latter, comprises the vital information that does not fit neatly into the rows and columns of a traditional spreadsheet or database, such as claim adjuster notes. Why is predictive analytics important to claims management? Because it finds relationships in data that achieve a more complete picture of a claim, guiding better decisions around its management. This remarkable functionality is now at hand to achieve unparalleled effciencies and cost-effectiveness when managing claims for workers compensation, casualty bodily injury, employment practices liability, and other financial risks. 3

4 The Modeling of Claims The typical claim involves an enormous volume of disparate data that accumulates as the claim progresses. Take the example of a workers compensation claim. The data runs the gamut from the actual claim filing itself to the action plan of the claims examiner; the medical file of the injured or ill worker; his or her specific personal and economic demographics, job tenure and medical history; the different medical services, medications and physical therapy that were prescribed for the individual; the various transactional amounts paid to date; and ongoing progress reports on the individual s condition, to cite just a few data sets. This vast volume of information compiles on a continuous basis over time. Making sense of it all for decision-making purposes is extremely challenging, given the sheer complexity of the data. Predictive analytics alters this paradigm, offering the means to distill and assess detailed line-item claims information. In the hands of insurers and third party claims administrators, such analytical tools can, for instance, identify unrecognized potential claims severity and the relevant contributing factors. Having this information in hand, a claims professional can take deliberate actions to reduce or mitigate the financial losses generated by the claim. To identify and prescribe specific actions that can positively affect claims outcomes, insurers need to undertake robust claims modeling. This has been challenging in the past because the tools available for such modeling had limitations. Unable to capture and assess unstructured claims data, traditional claims models issued predictions based exclusively on structured data. With more of the story now being told, the insurer s ability to reduce the financial impact of a claim is more effective than ever before. Sharper Focus on Predictive Analytics Chubb 4D captures traditional structured data, as well as unstructured data in its advanced analytics models. Structured data includes the traditional information that is collected as part of the claim filing and investigation. Unstructured data focuses on the notes that are routinely written by claims adjusters while the claim remains open. By analyzing both data forms, claims adjusters and supervisors will achieve enhanced visibility into the driving factors of a particular claim. This knowledge can then be aggregated across the larger set of similar claims to develop programs based on more insightful and deliberate actions that target reducing claim costs. 4

5 Making use of the model Information is everything in business. But, unless it is given to applicable decisionmakers on a timely basis for purposeful actions, information becomes stale and of little utility. Even worse, it may direct bad decisions. In the context of predictive analytics, the intelligence provided by a model regarding a claim s relative severity is only as useful as the manner in which this information is received and acted upon. Obviously, the whole point of predictive analytics is to apprise claim teams and company risk managers of something before it occurs. To make good on this value proposition requires the establishment of two processes one ensuring that the right people receive the intelligence produced by the model in a timely manner, and another requiring specific deliberate actions to be taken based on the details. Predictive models just inform; people must intervene to seize upon this information to improve overall program performance. Sharper Focus on Action For claims data to have value as actionable information, it must be accessible to prompt dialogue among those involved in the claims process. Although a model may capture reams of structured and unstructured data, these intricate data sets must be distilled into a comprehensible collection of information. To simplify client understanding, Chubb 4D produces a model score illustrating the relative severity of a claim, a percentage chance of a claim breaching a financial threshold or retention level depending on the model and program. The tool then documents the top factors feeding into these scores. 5

6 6 Chubb 4D: Power of Predictive Analytics

7 Leveraging The Information For predictive analytics to perform as intended there must be consistency in execution. Execution requires the organization to embrace predictive modeling. Actions based on the models must be well defined and supported by technology and procedures that capture and re-enforce the action. Prior planning is critical. Simply adding responsibilities to the many tasks that claims teams must complete as part of their caseloads may not be effective as it creates the risk that proper mitigating procedures will not be performed to their conclusion. The solution for these deliberate actions to be meaningful and not disruptive is to shift the workload of their claims teams appropriately, enhancing their effciency and effectiveness through their use of the model. Sharper Focus on Insight Predictive analytics will identify claim characteristics that drive exposure. These characteristics coupled with claims handling experience create the opportunity to change the course of a claim. To test the effcacy of the actions implemented, a before - after impact assessment serves as a measurement tool. Otherwise, how else can program stakeholders be sure that the actions that were taken actually achieved the desired effects? Say certain interventions are proposed to reduce the duration of a particular claim. One way to test this hypothesis is to go back in time and evaluate the interventions against previous claim experience. In other words, how does the intervention group of claims compare to the claims that would have been intervened on in the past had the model been in place? An analogy to this pastpresent analysis is the insight that a pharmaceutical trial captures through the use of a placebo and an actual drug, but instead of the two approaches running at the same time, the placebo group is based on historical experience. 7

8 Why Unstructured Data is Vital The industry has long relied on structured data to make business decisions. But, unstructured data like claim adjuster notes can be an equally important source of claims intelligence. The diffculty in the past has been the capture and organization of this fast-growing source of information. With Chubb 4D, this is no longer the case. From a claims standpoint, trying to predict an outcome without including unstructured data is like putting together a puzzle with missing pieces. This is not to imply that structured data has less value than its unstructured cousin. Information on a claimant s age, injury type, and occupation are critical elements in predicting the outcome of the claim. But, a far richer story can be told when mining and adding relevant unstructured data to the statistical analyses. Such data, for instance, may include information indicating that a workers compensation claimant recently put on a lot of weight or is under severely stressful conditions at home or at work. Maybe the claimant a short time ago started taking a prescription medication that is unrelated to the claim, but has claimrelated ramifications. The medication could influence the treatment plan, return-to-work options, and claim duration. Perhaps the claimant recently separated or divorced a spouse. This may affect the physical and emotional support he or she was to receive at home, with resulting claims implications. Not all of the above examples will be present on enough claims to include in a predictive model, but the goal is to identify as many as possible and test each as to they relate to claim outcomes. Unstructured data has vital import to the management of a claim. Since this form of information is not static and varies across claims, the ability to mine the text across the data set is critical. By combining this dynamic intelligence with more structured data, the accuracy of the predictive analytics is enhanced. Furthermore, the knowledge gleaned from the combined data sets may be used to assist claims teams to ask more in-depth questions of claimants in the future. Sharper Focus on Data Often buried within a claim adjuster s notes are nuggets of information that can guide better treatment of the claimant or suggest actions that might lower associated claim costs. Adjusters routinely compile these notes from the initial investigation of the claim through subsequent medical reports, legal notifications, and conversations with the employer and claimant. This unstructured data, for example, may indicate that a claimant continually comments about a high level of pain. With Chubb 4D, claims teams can discern how many times the word pain appears in the notes. Algorithms a set of problem-solving rules or instructions inserted in the model can determine the relationship between the number of times the word appears and the likely severity of the claim. Similarly, the notes may disclose a claimant s diabetic condition (or other health-related issue), unknown at the time of the claim filing but voluntarily offered up by the claimant in conversation with the adjuster. These insights are vital to evolving management strategies and improving a claim s outcome. 8

9 A Balancing Act The capacity to mine, process, and analyze both structured and unstructured data together enhances the predictability of data analytics. In other words, the more clues the better the ability to deduce an outcome. But, there is risk in not carefully weighing the value and import of each piece of information before inputting it in the model. Overdependence on text, for instance, or undervaluing such structured information as the type of injury or the claimant s age, can result in inferior deductions. The goal is to continuously sift the wheat from the chaff. To do this, claims teams must be entrusted to capture the same specific, high-quality information both structured and unstructured on a consistent basis. Otherwise, the reliability of the projections will be undependable. Claims teams also must establish postmodel actions to reduce claims duration and/or enhance claimant return-towork goals, and then scrutinize the effectiveness of these actions on a routine basis. Merely running the model is not enough to foster positive change; actions and their impact must be consistently measured and monitored. Sharper Focus on Measurement A major modeling pitfall is measurement as an afterthought. Frequently this is caused by a rush to implement the model, which results in a failure to record relevant data concerning the actions that were taken over time to reduce claim duration and severity. For modeling to be effective, actions must be translated into metrics and then monitored to ensure their consistent application. Prior to implementing the model, insurers need to establish clear processes and metrics as part of planning. Otherwise, they are flying blind, hoping their deliberate actions achieve the desired outcomes. Tools like predictive analytics are only as powerful as the precise processes surrounding their use. 9

10 The Bottom Line While the science of data analytics continues to improve, predictive modeling is not a replacement for experience. Seasoned claims professionals and risk managers will always be relied upon to evaluate the mathematical conclusions produced by the models, and base their actions on this guidance and their seasoned knowledge. The reason is like people predictive models cannot know everything. There will always be nuances, subtle shifts in direction, or data that has not been captured in the model requiring careful consideration and judgment. People must take the science of predictive data analytics and apply their intellect and imagination to make more informed decisions. Extracting the right information from a model also requires inputting the right data. The same can be said for getting the right information at the right time into the hands of the right people. Predictive models must produce meaningful analyses that are easily comprehensible and accessible to all relevant decisionmakers. Finally, the information provided should be as straightforward as the actual analytics is complex. A report or screen with dozens of complicated variables will not be acted on. An approach that filters out the noise and fine-tunes the key variables that have, or support, a causal relationship with a claim s outcome has a better chance of being read, digested, and acted upon. Sharper Focus on Chubb 4D So why consider products embedded with Chubb 4D as your trusted partner in advanced claims analytics? We have made substantial investments in this area that continue. We have put together a team of nearly 40 professionals who are dedicated to modeling, and continue to investigate and deploy the necessary tools to translate data into information. And we see great value in our tools helping our customers reduce costs and increase effciencies. 10

11 11

12 Contact Us Scott Henck SVP Claims O: Chubb is the marketing name used to refer to subsidiaries of Chubb Limited providing insurance and related services. For a list of these subsidiaries, please visit our website at Insurance provided by Chubb American Insurance Company and its U.S. based Chubb underwriting company affliates. All products may not be available in all states. This communication contains product summaries only. Coverage is subject to the language of the policies as actually issued. Surplus lines insurance sold only through licensed surplus lines producers. Chubb is the world s largest publicly traded property and casualty insurance group. With operations in 54 countries, Chubb provides commercial and personal property and casualty insurance, personal accident and supplemental health insurance, reinsurance and life insurance to adverse group of clients. Chubb Limited, the parent company of Chubb, is listed on the New York Stock Exchange (NYSE: CB) and is a component of the S&P 500 index. Copyright 2016 Form (Rev. 6/16)