Interactive learning for efficiently detecting errors in insurance claims

Abstract

Many practical data mining systems such as those for fraud detection and surveillance deal with building classifiers that are not autonomous but part of a larger interactive system with an expert in the loop. The goal of these systems is not just to maximize the performance of the classifier but to make the experts more efficient at performing their task, thus maximizing the overall Return on Investment of the system. This paper describes an interactive system for detecting payment errors in insurance claims with claim auditors in the loop. We describe an interactive claims prioritization component that uses an online cost-sensitive learning approach (more-like-this) to make the system efficient. Our interactive prioritization component is built on top of a batch classifier that has been trained to detect payment errors in health insurance claims and optimizes the interaction between the classifier and the domain experts who are consuming the results of this system. The goal is to make these auditors more efficient and effective as well as improving the classification performance of the system. The result is both a reduction in time it takes for the auditors to review and label claims as well as improving the precision of the system in finding payment errors. We show results obtained from applying this system at two major US health insurance companies indicating significant reduction in claim audit costs and potential savings of $20-$26 million/year making the insurance providers more efficient and lowering their operating costs. Our system reduces the money being wasted by providers and insurers dealing with incorrectly processed claims and makes the healthcare system more efficient.

A Review of User Interface Design for Interactive Machine Learning

“…The system builds a probabilistic model based on a Naïve Bayes classifier of contacts likely to be relevant to a particular group and makes suggestions and provides new filters for identifying new members. Ghani and Kumar [2011] apply an IML process to greatly assist the work of auditors processing health insurance claims. In addition to learning to highlight fields of likely interest to support the auditor, the classification method can also help by grouping similar claims in the processing queue to reduce inefficiencies of context switching.…”

“…Yimam et al [2015] highlight the fact that a text classification system is likely to reach a point of diminishing returns for user annotations. Ghani and Kumar [2011] also argue for a distinction between system performance and classification accuracy. The task overview interface should provide visibility of the global objectives, but also contextualise these with other information relevant to the task such as the target application of the model and the availability of training data.…”

“…Importantly, some applications such as abstrackr [Wallace et al 2012] and the assisted claims processing system of [Ghani and Kumar 2011] make no real distinction between training and deployment. The model learns and becomes more helpful as its predictive power increases.…”

“…The value of the approach is perhaps most evident when one considers the potential to allow a domain expert to train a model. Researchers have demonstrated IML systems that leverage domain expertise in a range of specific applications including animal behaviour annotation [Kabra et al 2013], insurance claims auditing [Ghani and Kumar 2011] and conducting systematic reviews of published medical evidence [Wallace et al 2012]. In these applications, the effort required by the user is significantly reduced as the model improves its understanding of the concept being trained.…”

Interactive Anomaly Detection Based on Clustering and Online Mirror Descent

“…In [9], a method for detecting errors in insurance claims was proposed. The method aims at reducing the expert effort required to verify the insurance claim errors.…”