Domestic Actuarial and Internal Insurance Swindle Interventions - Research Paper Example

DOMETIC INSURANCE AND DOMESTIC INSURANCE FRAUD INTERVENTIONS – A SYSTEMATIC REVIEW INTRODUCTION For the social sciences to be meaningful and effective in contributing to the development of society, there is the need for empirical studies and findings to be presented in a way and manner that addresses issues and improves things in the society (Petticrew & Roberts, 2006). The issue of insurance fraud in the domestic insurance context is a major matter that has significance to the insurance industry as a whole. This chapter of the research undertakes a systematic review of authoritative peer-reviewed journals to provide an insight into the detection and prevention of domestic insurance fraud from an academic perspective. To this end, the chapter will review practically conducted research and studies and their key findings and put them in the context of the prevention and detection of risks. Since domestic insurance fraud is a specialised niche in the insurance industry, it seems to have various processes and techniques that are relevant to it. Therefore, this chapter will involve various techniques and tools that have been identified in study processes that have been put into an empirical context. The core findings and explanations of the various journals are critiqued and analysed in the context of the study in order to provide conclusions and a presentation of theories and concepts that are used to deal with domestic insurance fraud. CONTROL OF DOMESTIC INSURANCE FRAUD: THEORETICAL CONTEXT The general framework for containing insurance fraud involves a conscious effort to evaluate and detect insurance risks and deal with them within a given system or framework. According to Furlan et al (2011), this involves the identification of a five-step process which involves the creation of an ideal computer program with the view of detecting and preventing insurance fraud. The following steps are essential: 1. Goal 2. Process (KPI) 3. Activity 4. Resource 5. Computer Program (Furlan, Vasilecas, & Bajec, 2011) By way of goal, the insurance firm will need to define the main end it seeks to meet. This includes the levels of profitability that they seek to attain and how they want to match their insurance premium to their claim pay-outs. This culminates in the creation of a framework for the insurance strategy. Based on the insurance strategy, the key-performance indicators can be formulated and this will include amongst other things, the identification of the main targets and the main factors that must be viewed and evaluated critically. This involves the identification of the methods and ways through which fraud will be deduced and dealt with. The KPI defines the main activities for detecting and dealing with fraud. This leads to the definition of resources and data mining processes and this causes the insurance company to define the computer program that can be used for dealing with suspected fraud and handling it in a way of completing matters and situations. The computer program that is put in place to detect and prevent insurance fraud must have the competency to undertake a large volume of data mining in order to go through all claims and ensure that they are screened through due processes (Ormerod, Ball, & Morley, 2012). The software and data mining system needs to be able to profile information and take the claims and different processes through various stages (Ormerod, Ball, & Morley, 2012). However, a good domestic insurance detection system needs to be supported by specialist investigators who have to select major issues and cases and give it a critical and in-depth analysis and evaluation. This will culminate in the identification and evaluation of matters in order to come up with a thorough and complete assessment of whether they are credible or not. The selection of cases to be evaluated must be connected to the identification of the appropriate and relevant issues based on dominant indicators in relation to fraud (Ormerod, Ball, & Morley, 2012). This involves careful and critical suspicion building with the hope of exploring issues in order to build hypotheses that will enable the insurance company to stand their ground in disputing a dishonest insurance claim (Ormerod, Ball, & Morley, 2012). This therefore implies that insurance fraud control must be handled as a strategic wing of an insurance organisation. This is because the organisation needs to identify the main issues and matters relating to the insurance fraud processes and activities and create a system and strategy for the detection and prevention of fraud through various data matching and data mining techniques in order to deal with the issues that may come up. The journals reviewed in this study provided a critical and comprehensive system through which the different processes and procedures for detecting and preventing insurance fraud is critiqued and evaluated. This provides absolute and relativist views of how each process and procedure can be employed to deal with various aspects and elements of each method and process. DOMESTIC INSURANCE FRAUD PREVENTION Prevention is an important element and aspect of dealing with fraud in the insurance industry. This is because a good prevention system will detect issues and problems and provide a solution to it. Prevention has to do with the creation and institution of important methods and systems that will discourage fraudulent claimants from making their original attempt to take money illegally from an insurance company. One of the approaches put forward for this is to use the naive Bayes theorem approach which creates a system and framework of helping to identify fraud of first instance (Viaene, Derrig, & Dedene, A Case Study of Applying Boosting Naive Bayes to Claim Fraud Diagnosis, 2004). This can be triggered by suspicious integration of figures that are meant to promote and attain a given end in the fraudulent claim system Thus, the Bayes theorem process is used by defining events that come within a given sequence. The sequence alerts the insurance company to identify issues and investigations that can be used to prevent fraud in the first instance by a person or groups of people who have connived to make a claim that is not in good faith. Thus, the system is designed to define a frame of reasonable probabilities for the attainment of claims. Thus, the Bayes theorem is used to create a set of parameters and ceilings and floors within which a reasonable insurance claim can be within. This will be used for the different stages and events in the early phases of a given claim process and system. The naive Bayes theorem model enable the person granting the insurance to conduct preliminary investigations in order to disqualify and eliminate claims that have issues and are not in sync with the basic and fundamental approach and method of making a claim. This is common in the automobile insurance industry. And it could potentially be employed in the domestic insurance system by defining the different approaches and variables that can be used to prevent fraudulent acts and plans. Therefore, the naive Bayes sets up a simple classification of setting criteria and defining the framework within which each events can be dealt with. However, the main limitation of this process is that it needs to be applied with a watchful eye and supervision by the insurance company. Classification of screening options is central and important in the preventive drive and preventive process (Viaene, Ayuso, Guillen, Gheel, & Dedene, 2007). This is done by identifying the variables that are most relevant to the first instance of a given type of insurance fraud. After this is done, there is the need for the insurance company to rate them and integrate them practically into the fraud prevention system and procedure (Belhadji, Dionne, & Tarkhani, 2000). The model will work through the use of logarithms and other processes in probabilities to set the parameters and conduct checks in the system (Belhadji, Dionne, & Tarkhani, 2000). Hence, the insurance company will have to set up a system and process through which the main pointers can be identified and setting the rules and targets for dealing with them. However, the prevention process ought to be done in a way that scenarios will be carefully evaluated and analysed for a solution to be found to each specific case (Viaene S. , Derrig, Baesens, & Dedene, 2002). DETECTION OF DOMESTIC INSURANCE FRAUD In a situation whereby people are involved in insurance fraud, there is the need for some effort to be made to segregate honest claims from dishonest claims. There are numerous methods and processes that are used to analyse claims and deal with insurance fraud in order to provide solutions. There are various statistical methods and approaches that are put together by various insurance companies to detect fraud and deal with it. This can be used in the domestic insurance industry to detect domestic insurance fraud and provide a solution to it. One of the methods is the PRIDIT method which is a combination of Principal Component analysis and RIDIT Scoring (Brockett, Derrig, Golden, Levine, & Alpert, 2002). This is done by using bootstrapping to rank the different levels and different processes of insurance fraud risks in order to create a program and system that can deal with the potential processes and systems that can be used as an excuse or basis for insurance fraud. The PRIDIT method provides a major process and a system for unsupervised learning method that provides a process and a system through which the different risks are put together and defined for checks to be taken. Thus, yields and processes from historical data can be compiled and analysed in order to provide a system and framework for sampling of insurance data and data mining to deal with the needs and expectations of consumers (Tennyson & Salsas-Forn, 2002). PRIDIT utilises statistical methods and simplifies the claim auditing processes and systems. This audits every single feature and every single process in order to provide various levels of checks for the processing of different claims and elements and processes. Where a given process and a given claim is seen to be a high-risk process or procedure, there was the need to take statements and objective elements are used to define the different processes and issues with different requirements and different needs. These risks are examined and reviewed in order to set the risk detection parameters and ensure that cases that fall outside the acceptable scope is identified and audited through the use of various critical methods and approaches to the case and issue at hand. This is complemented by a lot of record keeping in order to dispute and challenge claims that are reasonably viewed to be suspicious (Tennyson & Salsas-Forn, 2002). The PRIDIT method and process is based on fraud rate estimation and this is done to define likely results and likely issues that could come up (Ai J. , Brockett, Golden, & Guillen, 2013). PRIDIT is used by predicting the variables of a given insurance process. This therefore creates a kind of scale in which different rates are ascribed to a given variable and based on the rate estimation, limits and scopes are created to deal with the different claims (Ai J. , Brockett, Golden, & Guillen, 2013). Where something is extreme, that extreme component is evaluated in relation to the results and the processes. The decision tree is used as an aid to the PRIDIT method. This is because there is the need for various decisions and processes to be put together in order to analyse a given claim. This is because the system and software that is used by a given company will be done through the process of analysing and evaluating the different processes and procedures that follow the sequence. This will enable the auditor of a given claim to see whether a claim was conventional or not conventional. Conventional claims will go through the normal channel and procedure of a given claim. The procedure that is out of the normal scope will prompt further analysis and procedures to deal with the processes. The decision tree provides different options and different possibilities that enables the company to evaluate and screen each insurance claim and procedure (Gepp, Wilson, Kumar, & Bhattacharya, 2012). The decision tree provides different methods of data mining and this is complemented with the Logit analysis and procedure which analyses different processes and procedures that comes with some degree of success (Gepp, Wilson, Kumar, & Bhattacharya, 2012). The logit process involves multinominal logit model (MNL) and this involves classifying different claims in order to investigate them further and provide an important method and procedure for further analysis and evaluations (Caudlill, Ayuso, & Guillet, 2005). However, the decision tree process is used to provide a higher degree of responsive review and analysis of a larger data set (Gepp, Wilson, Kumar, & Bhattacharya, 2012). This is because it can be used to deal with a set of large data sources and large data information that is compiled and put together by a company. This allows a large sample to be covered and through this, different data types can be classified for further evaluation and analysis in order to provide an insight into whether there are some fraudulent claims or aspects in the claim. Also, another approach and method that can be viewed as a strong tool for detecting claim fraud is the auditing of claims and analysis and evaluation. This auditing is done by evaluating the different needs and expectations of a given class of transactions. This involves the identification of various needs and processes. However, in this process, there are so many issues with the selection of samples. This is because some samples are somewhat not representatives. This includes selection biases and other difficulties in order to close in on suspicious issues and problems that could lead to different findings and requirements. Thus, the sampling process and the selection of claims to audit is important in defining whether false claims will be detected or not (Pinquet, Ayuso, & Guillen, 2007). However, a biviriate model that is put forward by Pinquet et al prevents bias and helps the insurance company to go through the sampling process in a way and manner that enhances the detection potentials and possibilities of a given set of data. RESPONSE TO DETECTED ISSUES Where a given form of fraud is detected, there is the need to subject the sample or the class of transactions through various forms of audits and various kinds of critical revaluations and analysis. This will enable the insurance company to analyse and evaluate the different components of the claim. The end of this will involve identifying the methods and processes for dealing with the claim and also provide the right results and the right attitudes for dealing with fraudulent claims and demands. Automatic fraud detection systems and processes are not sufficient in providing solutions in themselves (Laegreid, 2007). This implies that there is the need for further revaluation of different variables and processes in a given case and situation. This must be carefully developed and done in a way and manner that deals with the issues and problems that comes with it. The process of critiquing claims that are identified to be issue-oriented or potentially fraudulent will involve the proper classification and further investigations based on the variables detected (Ai, Brockett, & Golden, Assessing Consumer Fraud Risk in Insurance Claims: An Unsupervised Learning, 2012). CONCLUSION The insurance process and system can be streamlined through the presentation of an appropriate method and system for preventing fraud and detecting fraud. This is done through the presenting a method and a procedure for evaluating the system and creating an electronic system for the evaluation of different claims in order to prevent them from becoming an issue. The different processes and procedures that are used include the different detection procedures that can statistically be used to deduce the different claims that are not accurate and appropriate. Most of these requirements and these systems include studies from the automobile industry and most of them have been adapted by the domestic insurance industry. Bibliography Ai, J., Brockett, P. L., & Golden, L. L. (2012). Assessing Consumer Fraud Risk in Insurance Claims: An Unsupervised Learning. SSRN , 1 - 41. Ai, J., Brockett, P., Golden, L., & Guillen, M. (2013). A Robust Unsupervised Method for Fraud Rate Estimation. Journal of Risk and Insurance , 121 - 143. Artis, M., Ayuso, M., & Gillien, M. (2002). Detection of Automobile Insurance Fraud with Discrete Choice Models and Misclassified Claims. The Journal of Risk and Insurance , 325 - 340. Belhadji, E. B., Dionne, G., & Tarkhani, F. (2000). A Model for the Detection of Insurance Fraud. The Geneva Papers on Risk and Insurance , 517 - 538. Brockett, P. L., Derrig, R. A., Golden, L. L., Levine, A., & Alpert, M. (2002). Fraud Classification Using Principal Component Analysis of RIDITs. Journal of Risk and Insurance , 341 - 371 . Caudlill, S., Ayuso, M., & Guillet, M. (2005). Fraud Detection Using a Multinomial Logit Model with Missing Information. Journal of Risk and Insurance , 539 - 550. Furlan, S., Vasilecas, O., & Bajec, M. (2011). Method for Selection of Motor Insurance Fraud Management System Components Based on Business Performance. Technological and Economic Development of Economy , 535 - 561. Gepp, A., Wilson, J. H., Kumar, K., & Bhattacharya, S. (2012). A Comparative Analysis of Decision Trees Vis-a-Vis Other Computational Data Mining Techniques in Automotive Insurance Fraud Detection. Journal of Data Science , 537 - 561. Laegreid, I. (2007). Automatic Fraud Detection – Does it Work? Annals of Acturial Science , 271 - 288. Ormerod, T. C., Ball, L. J., & Morley, N. J. (2012). Informing the development of a fraud prevention toolset through a situated analysis of fraud investigation expertise. Behaviour and Information Technology , 371 - 381. Petticrew, M., & Roberts, H. (2006). Systematic Reviews in Social Sciences: A Practical Guide. London: Blackwell Publishing. Pinquet, J., Ayuso, M., & Guillen, M. (2007). Selection Bias and Auditing Policies for Insurance Claims. The Journal of Risk and Insurance , 425 - 440. Tennyson, S., & Salsas-Forn, P. (2002). Claiming Auditing in Automobile Insurance: Fraud Detection and Deterrence Objectives. Journal of Risk and Insurance , 289 - 308. Viaene, S., Ayuso, M., Guillen, M., Gheel, D. V., & Dedene, G. (2007). Strategies for detecting fraudulent claims in the automobile insurance industry. European Journal of Operational Research , 565 - 583. Viaene, S., Derrig, R. A., Baesens, B., & Dedene, G. (2002). A Comparison of state-of-the-art classification techniques for expert automobile insurance claim fraud detection. Journal of Risk and Insurance , 373 - 421. Viaene, S., Derrig, R., & Dedene, G. (2004). A Case Study of Applying Boosting Naive Bayes to Claim Fraud Diagnosis. IEE Transactions on Knowledge and Data Engineering , 612 - 620. APPENDIX Article Title: A Case Study of Applying Boosting Naive Bayes to Claim Fraud Diagnosis (Viaene, Derrig, & Dedene, 2004) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X The empirical data is presented as an example, and not as an actual discussion or transaction. Country USA & Belgium Initial Validity Type of Research Design Derivation of formula and principles for the use of the Baye’s theory in detecting and dealing with insurance fraud What is the study population and sample size? N/A Was a control group used? X Measurement Has fraud been clearly defined? X Not really, a background of claim fraud is giving in the beginning of the paper though Have the samples been clearly defined? i.e. where did the data come from? N/A What was the outcome measurement? i.e. statistical, qualitative? What statistic was used? X Was the intervention found to be effective? X It needs to be applied for clear analysis Bias Was the detection measurement objective? X Are the results unbiased? X Outcome Were the statistics used correctly? X Are the results significant? X Yes, it proposes a method of dealing with the detection system and improving it to handle issues and matters Are the results clearly reported? X Have the limitations been discussed? X Article Title: A Model for the Detection of Insurance Fraud (Belhadji, Dionne, & Tarkhani, 2000) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X Country Canada Initial Validity Type of Research Design Quasi-Experimental What is the study population and sample size? X Insurance firms in Canada were studied in the research. A sample of 20 insurance firms in Quebec was studied. Was a control group used? X Measurement Has fraud been clearly defined? X Have the samples been clearly defined? i.e. where did the data come from? X What was the outcome measurement? X Statistical What statistic was used? X Was the intervention found to be effective? X Yes, the intervention is effective but it ought to be carried out with the traditional approaches to improve the detection and prevention levels Bias Was the detection measurement objective? X Are the results unbiased? X Outcome Were the statistics used correctly? X Are the results significant? X Are the results clearly reported? X Have the limitations been discussed? X Article Title: A Comparison of state-of-the-art classification techniques for expert automobile insurance claim fraud detection (Viaene S. , Derrig, Baesens, & Dedene, 2002) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X Country United States of America (Massachusetts) Initial Validity Type of Research Design Longitudinal with control group. What is the study population and sample size? X Was a control group used? X Measurement Has fraud been clearly defined? X Have the samples been clearly defined? i.e. where did the data come from? X What was the outcome measurement? Statistical and Qualitative What statistic was used? X Was the intervention found to be effective? X Bias Was the detection measurement objective? X The detection measurement could have included all methods used but there were some that were not analysed Are the results unbiased? X Results could be said to exclude some methods Outcome Were the statistics used correctly? X Are the results significant? X Are the results clearly reported? X Have the limitations been discussed? X Article Title: Fraud Classification Using Principal Component Analysis of RIDITs (Brockett, Derrig, Golden, Levine, & Alpert, 2002) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X Country United States of America Initial Validity Type of Research Design Longitudinal Research What is the study population and sample size? X Was a control group used? X Measurement Has fraud been clearly defined? X Have the samples been clearly defined? i.e. where did the data come from? X What was the outcome measurement? Statistical and Qualitative What statistic was used? X Was the intervention found to be effective? X Bias Was the detection measurement objective? X Are the results unbiased? X Outcome Were the statistics used correctly? X Are the results significant? X Are the results clearly reported? X Have the limitations been discussed? X Article Title: Detection of Automobile Insurance Fraud with Discrete Choice Models and Misclassified Claims (Artis, Ayuso, & Gillien, 2002) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X Country American Writers, but Study was on the Spanish Insurance Industry Initial Validity Type of Research Design Data Mining What is the study population and sample size? The Spanish Insurance Markets and a total of 1,995 claims between 1993 and 1997 in the Spain were studied Was a control group used? X Measurement Has fraud been clearly defined? X Have the samples been clearly defined? i.e. where did the data come from? X What was the outcome measurement? Statistical What statistic was used? Mean and Standard Deviation in the context of cumulative probability function Was the intervention found to be effective? X Bias Was the detection measurement objective? X Are the results unbiased? X Outcome Were the statistics used correctly? X Are the results significant? X Are the results clearly reported? X Have the limitations been discussed? X The limitation involves the use of other methods and processes to analyse similar cases and issues. Binary choices do not effectively evaluate the other aspects of fraudulent claims. Article Title: Claiming Auditing in Automobile Insurance: Fraud Detection and Deterrence Objectives (Tennyson & Salsas-Forn, 2002) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X Country United States of America Initial Validity Type of Research Design Control Group What is the study population and sample size? Auto insurance claims in Massachusetts Was a control group used? 1207 Personal Injury claims Measurement Has fraud been clearly defined? X Have the samples been clearly defined? i.e. where did the data come from? X What was the outcome measurement? X Statistical and Qualitative What statistic was used? Distribution of Personal Insurance Claims; Audit Method Distribution Was the intervention found to be effective? X Bias Was the detection measurement objective? X There were numerous methods and approaches that was used by others which yielded different results Are the results unbiased? X Outcome Were the statistics used correctly? X Are the results significant? X Are the results clearly reported? X Have the limitations been discussed? X Article Title: Fraud Detection Using a Multinomial Logit Model with Missing Information (Caudlill, Ayuso, & Guillet, 2005) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X Country Initial Validity Type of Research Design Data Mining What is the study population and sample size? X Car Insurance clients Was a control group used? X AAG Car insurance clients in Spain Measurement Has fraud been clearly defined? X Have the samples been clearly defined? i.e. where did the data come from? X What was the outcome measurement? X Statistical What statistic was used? X Logit Model Was the intervention found to be effective? X Somewhat Bias Was the detection measurement objective? X Quite subjective Are the results unbiased? X Outcome Were the statistics used correctly? X Are the results significant? X Are the results clearly reported? X Have the limitations been discussed? X Article Title: Informing the development of a fraud prevention toolset through a situated analysis of fraud investigation expertise (Ormerod, Ball, & Morley, 2012) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X Country United Kingdom Initial Validity Type of Research Design Longitudinal Study with Ethnographic Studies What is the study population and sample size? X Four Insurance companies in the UK Was a control group used? X General Analysis of information was conducted Measurement Has fraud been clearly defined? X Have the samples been clearly defined? i.e. where did the data come from? X What was the outcome measurement? Qualitative Was used? X Was the intervention found to be effective? X Bias Was the detection measurement objective? X Are the results unbiased? X Outcome Were the statistics used correctly? X Are the results significant? X Are the results clearly reported? X Have the limitations been discussed? X Article Title: A Robust Unsupervised Method for Fraud Rate Estimation (Ai, Brockett, Golden, & Guillen, 2013) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X Country Spain and the United States of America Initial Validity Type of Research Design Deduced from longitudinal studies What is the study population and sample size? X Was a control group used? X Measurement Has fraud been clearly defined? X Have the samples been clearly defined? i.e. where did the data come from? X What was the outcome measurement? Qualitative What statistic was used? PRIDIT Model Was the intervention found to be effective? X Bias Was the detection measurement objective? X Are the results unbiased? X Outcome Were the statistics used correctly? X Are the results significant? X Are the results clearly reported? X Have the limitations been discussed? X Article Title: A Comparative Analysis of Decision Trees Vis-a-Vis Other Computational Data Mining Techniques in Automotive Insurance Fraud Detection (Gepp, Wilson, Kumar, & Bhattacharya, 2012) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X Slight inputs of empirical studies were integrated Country USA Initial Validity Type of Research Design Longitudinal What is the study population and sample size? X N/A Was a control group used? X N/A Measurement Has fraud been clearly defined? X Have the samples been clearly defined? i.e. where did the data come from? X What was the outcome measurement? Qualitative and Statistical What statistic was used? Decision Tree and other Computational Tools Was the intervention found to be effective? X Bias Was the detection measurement objective? X Are the results unbiased? X Outcome Were the statistics used correctly? X Are the results significant? X Are the results clearly reported? X Have the limitations been discussed? X Article Title: Method for Selection of Motor Insurance Fraud Management System Components Based on Business Performance (Furlan, Vasilecas, & Bajec, 2011) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X Country Slovenia Initial Validity Type of Research Design Interviews and Longitudinal Studies What is the study population and sample size? 40 Insurance Companies in Slovenia Was a control group used? X Measurement Has fraud been clearly defined? X Have the samples been clearly defined? i.e. where did the data come from? X What was the outcome measurement? Qualitative What statistic was used? X N/A Was the intervention found to be effective? X Bias Was the detection measurement objective? X Are the results unbiased? X Outcome Were the statistics used correctly? N/A Are the results significant? X Are the results clearly reported? X Have the limitations been discussed? X Article Title: Selection Bias and Auditing Policies for Insurance Claims (Pinquet, Ayuso, & Guillen, 2007) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X Country Spain Initial Validity Type of Research Design Longitudinal Study What is the study population and sample size? Spanish Auto Insurance Companies Was a control group used? Audit Selection in the year 2000 of the selected companies. Measurement Has fraud been clearly defined? X Have the samples been clearly defined? i.e. where did the data come from? X What was the outcome measurement? Qualitative What statistic was used? Bivirate Model Was the intervention found to be effective? X Bias Was the detection measurement objective? X Are the results unbiased? X Outcome Were the statistics used correctly? X Are the results significant? X Are the results clearly reported? X Have the limitations been discussed? X Article Title: Assessing Consumer Fraud Risk in Insurance Claims: An Unsupervised Learning Technique Using Discrete and Continuous Predictor Variables (Ai, Brockett, & Golden, Assessing Consumer Fraud Risk in Insurance Claims: An Unsupervised Learning, 2012) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X Country USA Initial Validity Type of Research Design Longitudinal Study What is the study population and sample size? Insurance entities in Massachusetts. Was a control group used? X Measurement Has fraud been clearly defined? X Have the samples been clearly defined? i.e. where did the data come from? X What was the outcome measurement? Statistical and Qualitative What statistic was used? PRIDIT Was the intervention found to be effective? X Bias Was the detection measurement objective? X Are the results unbiased? X Outcome Were the statistics used correctly? X Are the results significant? X Are the results clearly reported? X Have the limitations been discussed? X Article Title: Automatic Fraud Detection – Does it Work? (Laegreid, 2007) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X Country United States of America Initial Validity Type of Research Design Data Mining What is the study population and sample size? Insurance Claims in Household and Motor Insurance Claims Was a control group used? 640,000 cases were reviewed Measurement Has fraud been clearly defined? X Have the samples been clearly defined? i.e. where did the data come from? X What was the outcome measurement? Statistical What statistic was used? Probabilistic Model Was the intervention found to be effective? X Bias Was the detection measurement objective? X Are the results unbiased? X Outcome Were the statistics used correctly? X Are the results significant? X Are the results clearly reported? X Have the limitations been discussed? X Article Title: Strategies for Detecting Fraudulent Claims in the Automobile Insurance Industry (Viaene, Ayuso, Guillen, Gheel, & Dedene, 2007) Question Yes No Unclear Comment Initial Screening Is the study written in English? X Is the study concerned with Domestic Insurance? X Is the study concerned with Domestic Insurance Fraud Interventions? X Does the study contain empirical data? X Country Spain Initial Validity Type of Research Design Data Mining What is the study population and sample size? Spanish Insurance Industry Was a control group used? 2,400 Claims Measurement Has fraud been clearly defined? X Have the samples been clearly defined? i.e. where did the data come from? X A group of selected Spanish insurance firms What was the outcome measurement? Statistical What statistic was used? Statistical Probability Was the intervention found to be effective? X Bias Was the detection measurement objective? X Are the results unbiased? X Outcome Were the statistics used correctly? X Are the results significant? X Are the results clearly reported? X Have the limitations been discussed? X Read More