iWebCare Project

State of the art in the area concerned Over the years organizations in the health care business chain collected and stored large data sets of audit results on insurance claims. The organizations used historical records and best practice templates (based on the knowledge acquired by domain experts) to identify and prevent probable fraudulent claims in the future. Such fraudulent activity usually includes duplicate claims submission and the sharing of a single patient ID in order to generate billing across multiple providers. Their previous methods of fraud detection often missed opportunities to collect money and adjusters spent too much time reviewing legitimate claims.
By combining well-defined rules (applied by domain experts) with data mining techniques fraud detection and prevention efforts can improve accuracy, decrease manpower and minimize loss.
Currently domain experts inside these organisations use legacy hardware and software infrastructure in order to perform the computational tasks associated with the detection of fraud and abuse. Third party proprietary software (such as SAS and SPSS) is mainly used in these cases.
State of the art in the area of data mining is also addressed. Data mining is now recognized as a key computational technology, supporting traditional tasks such as analysis, visualization, design, and simulation. This field is emerging as a fundamental research area with important applications in science, engineering, education, business, government, and manufacturing. In its typical form, data mining can be viewed as the formulation, analysis, and implementation of an induction process (proceeding from specific data to general patterns) that facilitates the extraction of information from data. The various techniques used differ in terms of the

Types of information that is extracted (e.g., predictive models, association patterns, cause-effect relationships, detection of affinity similarity-based groupings, deviation detection)
Format of the induced information (e.g. rules, decision trees, correlation networks, association patterns, neural networks, matrices, visualization)
Types of data they operate on (e.g., digital images, text, discrete, continuous, sequence, temporal), and
Application domain for which they are developed (e.g., finance, engineering, science, life science, manufacturing, marketing)

Advancement in the state of the art
The project will significantly contribute towards major advances of the state-of-the-art in many interrelated aspects of the problem domain, and in particular methodology, organisational model as well as technical/software architecture.

From the methodological point of view, the project results will include a unified and generic methodological framework for the classification, analysis and design of fraud detection rules that can be applied to the domain of the health care industry. Given that up to now fraud detection is usually tackled in ad-hoc or non-standard way, the lack of a common methodological framework prohibits any homogenisation effort of the fraud detection procedures across member-states.

On the other hand, due to the architecture of the proposed system, major advancements can be accomplished concerning the organisational model. Rules can be maintained at different organisational levels (national health system, social security funds, hospitals), and at the same time be globally accessible by the independent parties; thus a federated, albeit uniform approach to fraud detection can be achieved regardless of differences in technological platforms or organisational models. Finally, from the technical point of view, the proposed web services architecture enables involved organisations to access the fraud detection application in a straightforward and uniform manner by complying with platform independence standards.