Can smart claims technology boost your operational excellence in health insurance?
For insurers, healthcare insurance offerings introduce a set of particular challenges for keeping those portfolios at an acceptable level of profitability. Increasing premiums is not always an option due to regulatory and even legal restrictions, which is why profitability often depends on the insurer’s ability to minimize the operational costs. Therefore, it may not be a surprise that claims leakage prevention and operational efficiency have become the most important areas of strategic concerns.
While claims leakage is difficult to benchmark, sources estimate that leakage accounts for 11% of all claims paid. Some have even witnessed claims leakage of 20% to 30%. Handling a claim often involves several manual interventions, from the first notice of loss through to final settlement. So, there is a lot of room for human error throughout this process and off course there is also much room to declare medical costs here and there that are not part of the policy conditions and should not be paid out.
For insurers, claims operations is a critical function, because it serves as the central hub for intaking, processing, and paying out claims. Unfortunately, it is still a very intensive and manual process.
In addition, we notice that many insurers still struggle with poor data quality and integrity, which is usually the result of an historically grown, complex application landscape. Data is the lifeblood of an insurance company. Calculating claims on false or outdated information could lead to reputation damage and a significant loss of customers on the long-term. In the days of mainframe computing, when applications and data were more centralized, this was less of an issue. During the two last decennia, information technology evolved to a distributed and decentralized environment, while legacy systems are kept alive. The problems arose when new applications were added to these legacy systems, creating multi-layered and potentially redundant IT architectures. This caused a complex tangle of data integrations, and a vast amount of (unstructured) data stored in an abundance of (overlapping) data sources.
Health data is also subject to different laws and regulations such as data privacy (GDPR), to protect personal and sensitive data from cybercrime, as well as to ensure privacy rights. This elevated level of regulation has a significant impact on a companies’ organization and its business processes. For instance, storing health claims data like PII (Personally Identifiable Information) and PHI (Protected Health Information) in the cloud in a secure and safe way can be a challenge. That is why insurance companies still stick to a conservative approach to data processing, just to avoid use cases that might raise privacy issues. Data is still kept on-premises, despite higher operational costs, limited scalability and lower flexibility. As a result of this, they miss out opportunities that can create value. Another example is profiling, which can be used to predict behavior and make automated decisions in Smart Claims Handling (see below). Insurers must be aware that this is only allowed by the GDPR when certain conditions are met, because it involves evaluation of aspects of an individual like personality, behavior, interests, reliability, economic situation, and location.
So, eliminating false and outdated data together with curating validated and compliant information is an essential step, not only for automating the claims process.
Regardless of the technology used, automated decision-making is a data-driven methodology and hence the quality of input data will directly impact the accuracy and STP rate (straight-through processing) of claims. An important characteristic of qualitative data is the extent to which the data is structured and can be interpreted by computer systems without manual intervention. There are multiple options to sourcing the medical costs of a victim preferably in a structured way. Many insurers invest in customer engagement channels (web based customer portals where documents can be uploaded, but also more focused customer smartphone apps) which aim to collect data directly from the victim in a digitalized and structured manner. But the adoption rate of such channels, as well as the quality of data input by the victim, is heavily influenced by the perceived user experience of these applications.
More accurate source of data are data exchanges aiming to bridge the gap between insurers and the medical sector. The most common example in Belgium is AssurPharma, a collaborative effort between Assuralia, APB (the national federation of independent pharmacists) and OPHACO (a recognized cooperative society representing approximately 600 pharmacies). Their platform enables the digital & structured data transmission of BVAC attestations between pharmacists and insurers.
But medical claims go beyond BVAC attestations, and this is where the interpretation of data becomes more complicated. Especially in the area of medical evidence documents outside of the hospitalization, today insurers need to rely on third party service providers to feed this non-structured information into structured form for further handling, but the benefits realized through this approach are not very cost effective.
Lean Claims Handling
Insurance companies still rely heavily on manual claims processing, where documents exist in different formats (machine-printed, scanned, handwritten, etc.), all containing structured, semi-structured, or unstructured data. This process requires subject matter experts to analyze these documents, which is a very costly and time-consuming process. It ultimately inhibits the ability to gain insights and perform analysis in a way that is both economical and timesaving.
These hard-to-disrupt, manual processes drive the use of technology to prevent further claim leaks and to achieve operational excellence. We firmly believe that the first level of optimalisation before using technology is to apply the principles of Lean Six Sigma to spot the right use cases to tackle.
Smart Claims Handling
Cognitive Document Processing (CDP) eases the burden of processing documents manually. CDP leverages deep learning to extract actionable information from structured and unstructured documents to drive decision-making in the claims handling workflow. For example, CDP could make a distinction between claims about disability, hospitalization, and accidents at work, and extract the relevant data elements from it. However, potential use cases should be identified based on the product of the expected benefits (first quantified by applying lean six sigma) and the feasibility to accurately train a deep learning model. The latter requires the availability of well-labeled sample data. The more heterogenous the documents to be processed, the larger this set of sample data must be. Let’s take the example of invoices produced by medical practices. A silver-bullet solution to completely decipher all content and mapped against a canonical medical nomenclature, is today still a challenge. More reasonable would be to train deep learning models to recognize sufficient actionable information and feed the results as input to another deep learning model that is trained to make decisions based on those sets of input. This is especially beneficial in those steps of the process where decision-making is identified as an expensive bottleneck. This could involve the routing of inbound communications, contextual tagging of claims, assignment to assessors and back-office administrators based on required skill level and even calculating the probability of claims leakage due to noncovered loss. CDP, if well-trained, can deliver substantial increases in speed, accuracy, and productivity.
The second pillar is Robotic Process Automation (RPA) to remove laborious and repetitive human tasks in the claims process, which is very error-prone and entails additional costs. Although RPA is a very basic automation process because it automates completely rule-based and programmed processes and is often now preceded by deep learning technology to structure the data before feeding it to the Robot. RPA can be used to move data from one system to another, to fill in forms on a website or on the mainframe, or to open emails and attachments from the claims’ mailbox. RPA can also retrieve and consolidate information from various systems (e.g. claims management, policy information, party information, …) populate fields in standard correspondence and even trigger claims reserve calculations. it is important with RPA to start by identifying the most difficult pain points in the daily tasks of the claims analyst, preferably based on measurements and a cost-benefit analysis.
Finally, never forget the customer. According to Forrester, a leading global market research company, Digital Customer Experience (DCX) is “half of what it means to be a digital business. The other half is Digital Operational Excellence (DOX).” DCX will help insurers to cultivate healthy, long-term customer relationships, potentially increasing lifetime customer value and opportunity through relevant products and offers. A DCX strategy forces insurers to think differently about their customer engagement approach. This means identifying how digital channels like online platforms such as smartphone and desktop, or even social media content, fits into their overall CX journey. Nowadays, consumers of those channels are willing to trade their personal data for more personalized products and services.
One way of improving DCX is Design Thinking, like Baloise did a few years ago. It is a process for solving complex problems in a user-centric way to ensure great experiences and outcomes for the customer. Proven benefits of DCX and Design Thinking are increased brand value, innovation and efficiency, increased customer retention and loyalty, and an increased ROI.
The key is to incentivize customers not only to adopt digital channels but also to help them help you, by submitting the right information at the right time, in the right format. Modern user interaction patterns, like digital claim assistants and conversational forms, have proven to outperform traditional form-style enquiries in perceived pleasantness. Such patterns allow for more dynamic enquiry flows, like branching out to another line of questioning based on the customer’s profile and answers to previous questions. This without exposing the true complexity of the flow to the customer. The Cognitive Document Processing solution would analyze the scan in real-time and immediately feed the results back to the customer engagement layer. The digital claim assistant could then request the customer to confirm a few elements derived from this document, like the name and address of the hospital, the total bill amount, and the date of admission.