Dierckx & Associates

Risk Management, Legal & Financial Investigations, Forensic Services, Christchurch New Zealand

TOPICS OF INSURANCE FRAUD DETECTION AND CONTROL

		Wednesday, 30 April 08 - 06:23 AM (GMT) By John ML Dierckx in Insurance Fraud

Part of an overall anti-fraud strategy is timely detection of fraudulent claims. In order to detect potentially fraudulent claims the organisation will need to have accurate and actual insight in what the “red flags” and or “indicators” of fraudulent claims are and could be. Probably the two most significant developments in this area are data mining techniques and the introduction of Voice Stress Analysis.

In order to identify suspicious claims, the organization needs a methodology that enables them to:

• match data fields within their own systems
• compare data fields with other data sources
• quickly drill down to individual levels

There are literally thousands of tests that can be run to identify possibly fraudulent claims. The decision as to which tests to run depend on several factors, to name a few:

• the type of claim
• the type of claimant
• quality of the data within the systems
• past incidents/claims history
• amounts claimed by claimants
• amounts previously paid to claimants
• specifics on the recipient of settlement
• matching data from unrelated claims

 

 The main criticism on fraud detection tests and systems at the moment is the enormous numbers of so-called “false positives” that are generated, because most of the “positives” have a legitimate explanation. The latest developments in that respect are highly sophisticated  data mining tools that use algorithms. In later paragraphs we will have a closer look at the methodology behind these systems.
 

Reverse Engineering, Profiling, Data Mining

 Whereas there are many theoretical models on the basis of which profiles can be developed, I suggest that the best models for fraud prevention are those that are derived from what is known as “reverse engineering” of past fraudulent claims. Reverse engineering is understood here as the structured approach towards gaining an understanding of the fraudulent claims and how certain red-flags would appear in your systems.

“Reverse Engineering (RE) is the process of discovering the process of discovering the technological principles of a device or object or system through analysis of its structure, function and operation... Reverse engineering is essentially science, using the scientific method... used by military to copy other nations' technology, devices or information, or parts which have been obtained by regular troops ion the fields or by intelligence operations...”

Source: http://en.wikipedia.org/wiki/Reverse_engineering

“What is Reverse Engineering?

Answer: Reverse engineering is the general process of analysing a technology specifically to ascertain how it was designed or how it operates. This kind of inquiry engages individuals in a constructive learning process about the operation of systems and products...

Source: www.chillingeffects.org/reverse/faq.cgi

 
Reverse engineering although often used in the technical context is a broad term that can be used in a wide variety of contexts, especially when used as contextual notion. The purpose of the reverse engineering exercise is to develop so called “red flags” or “indicators”.

The model presented for an anti-fraud strategy is in a substantial part based on what could be described as an intelligence based organization. Reverse engineering in this context is used as generating strategic intelligence that can be used for further data collection and decision making processes. In short reverse engineering is used to develop the answers to one central question: if a claim is fraudulent, what would I see?

In other words you develop a hypothesis (and test that) on the basis of existing case material as to what the indicators would be in the event of a fraudulent claim that can readily seen in your system. On the basis of reverse engineering a set of indicators or potential indicators can be developed that can be used for the purpose of profiling.

The more cases of fraud are “reversely engineered” the more sophisticated the ultimate set of red flags or indicators will be covering a wider range of options and possibilities . Profiling requires as substantial amount of data to be able to generate the “clues” that can readily be extracted from internal and external data sources.

Example

James filed a claim for the theft of property after someone broke into his house. The claim turned out to be a fraudulent claim and was subsequently reverse engineered. The following red flags were identified:

• The ink of James' insurance policy had barely dried, he was only insured for several weeks. Policy Administration/Accounts: red flag, timing

• In the week before the alleged burglary he had enquired about the exact amount for which he was covered and needed some explanation on the circumstances under which coverage was available. Call Centre Log and Client Administration: red flag , enquiry prior to claim

• The burglary had taken place in the Christmas weekend. Claims Administration, red flag: seasonal trend

• The call centre operator had made a note on how nervously James had reacted when she asked specific questions regarding his claim and that he was evasive in providing certain answers. Call Centre: red flag language

• The available personal data of James' showed that he used to work as an insurance broker and was now in between jobs. Client Administration: two red flags apply, insider knowledge and potential financial distress

• The claim was relatively excessive and very near the claims limit of his policy. Client Administration, red flag excessive claim

• When lodging the claim he advised that all his administration was taken as well and that he could not provide any proof of purchase. Claims Administration: red flag, lack of documentation

• During the investigation in which it was established that the claim was fraudulent, it transpired that several items that had allegedly been stolen were also advertised on different websites and in several classifieds papers and some of them were found to be sold upon further enquiries. External: red flag, previously advertised items of property
• James had made six other claims in the past two and a half year all with different companies. External/Internal, Claims Register: red flag, extensive claims history
• The PO  Box used as James' mailing address was identified as associated with several other claims however lodged by different policy holder names. External/Internal, claims register, red flag extensive claims history and inconsistent data

• James credit history showed a substantial indebtedness. Besides that it was found that he had applied for several finances in the period preceding the claim. External, credit registrations, red flag: looking for cash

 
These are but a few of the potential indicators or red flags that might be construed from one single fraudulent claim and were meant as an illustration on how reverse engineering could be used to identify indicators for fraudulent claims. Other examples will be dealt with in the form of an exercise.

It is important to note that none of these indicators by or even as a group are a guarantee that when those indicators are identified there is a case of a fraudulent claim. The red flags are not more than an indication that there is, based on previous experiences, a risk or better yet a chance that fraud is involved in the claim showing one or more of these indicators.

Where reverse engineering is aimed at generating the red flags or indicators, profiling can be described as a data surveillance technique whereby a set of characteristics of a particular class of person or subject is inferred from past experiences and subsequently used to search larger datasets are searched for subjects or persons that show a closer fit to those characteristics.

On the basis of agreed specific characteristics, reversely engineered from older cases, a set of indicators can be developed that are associated with, insurance and other frauds for that matter.

The technique is not used only for risk management purposes. Several marketing forms have utilized the technique as a means to improve efficiency in marketing communications with customers and prospects.that So in sense your anti-fraud efforts and systems may end up being used for non-related purposes and could actually earn themselves back.

Examples of how profiling is used in marketing communications are: Google Adsense and advertisements or Amazon where it creates profiles of book buyers. If you ever visited Amazon you will see the message: “readers that bought this book, also bought....”

To give you a feel of the broad range of the potential of using profiling in the sense as described here, the following could be possible target groups:

• parcels that contain counterfeited music and video;
• patients with a higher likelihood of specific diseases or complications
• companies or individuals that are more likely to commit tax fraud
• travellers that are more likely to be smuggling drugs
• prospects likely to be interested in buying a new product
• job seekers or prospects likely to be suited to fill a certain vacancy
• persons more likely than others to commit violent acts
• persons who are more likely than others to commit suicide
• containers more likely than others to have contraband in them
• chemical transports that are meant to supply drug manufacturing organizations
• claims are potentially fraudulent
• persons that are more likely to make false claims

The process of profiling basically consists of the following steps:

• Describe the class of person, organisation or events the organisation wishes to locate and Create a profile of such persons or events. This profile is created on the basis of reverse engineering described previously. Individuals, organisations or events known to belong to the class to be profiled are identified and their recorded characteristics examined in order to isolate common features. Techniques that can be used to create such profiles are:

• case analysis
• interviews with key staff and experts
• open source research
• statistical analysis of cases

• Formally express the profile, listing the characteristics and the weightings of these characteristics

• Acquire data, concerning the relevant population. In our case this would the available data regarding claims.

• Search the data, for those individuals, organisations or events that comply with the profile.

• Take action, in relation to the identified “matches” or “positives”

This five step process is circular and perpetual. As times goes on and new experience is added to the system the profiles will be come more and more sophisticated and accurate.

Any organization wishing to use profiling as described above will need to have a sound information infrastructure that contains the required data in the required formats. More importantly the systems will need to allow for almost real time monitoring or data surveillance and reporting since the claims handler and managers are usually required to act quickly on claims.

Illustrative are the observations of Tim Macdermid and Jason Burke, both staff members at SPSS, a predictive analytics company in the the Australian IT News on 12 September 2006:

“ Despite the jumble of ageing systems, most insurers are hanging out for a silver bullet remedy to fraudulent claims...

They are concerned their systems would not be able to support such software... The systems are very dated and there is considerable reluctance to modify systems as there is a large amount of risk...

...For insurance companies to take a serious stand against fraud claims, business processes need to change” Source: Kelly Mills, Insurance Fraud evades leaky net, Australian IT News 12 September 2006, www.australianit.news.com.au."

These remarks appear to tie with other experiences of people in the industry, involved in dealing with claims and fraud and are talking from expertience. Whereas the phenomenon of insurance fraud and especially claims fraud is taken quite seriously, the required follow through decisions appear more than once not to be made: decisions to update the data systems, decisions to invest in adequate technology. As a result of this the majority of fraudulent claims still go through the mazes of the net. As one claims fraud manager advised: “we only appear to be picking the easy cherries”.

Depending on the numbers of policy holders and claims to be dealt with there are some good arguments to use modern technologies to mine claims. Modern techniques such as data mining and algorithms provide greater accuracy especially on larger databases.

Highmark claims to have recovered about $11.5 million upon the introduction of data mining systems. Ton Brennan further advised in Insurance Networking News of April 2006, that”that doesn't take into account the insurance cheats that Highmark's SIU scares off. If people or if entities know that you're looking for fraud, and that you'll be on them as soon as it starts, there's not as great a propensity to attempt it.”

The Highmark business case is a good example of how things can evolve and how considerable cost savings can be realized.

There's a direct tie to dollars and cents benefits, unlike a lot of other areas where the benefits are longer term or the direct tie between an analytic technique or a model and the ultimate business outcome is separated by other layers of business intervention.

The Highmark business case evolved through three distinct steps:

• a partnering with a technological company to develop and run models to find patterns of fraud in the data available. Tis already lead to significant and substantial business outcomes in terms of fraud identification and savings
• the second step was to automate some of the manual work done in the first stage and to develop a program that lets investigators run their own queries and models
• step three is developing an active data mining structure that can be used to real time identify potential fraud and uses predictive models to anticipate fraud.

Using data mining and neural networking techniques have proven to be highly successful in the credit card industry with claim loss reductions between 20 and 50%. Remarkable results are claimed in the property and casualty insurance as well as the healthcare insurance.

Voice Stress Analysis

Voice Stress Analysis and Voice Stress Analysis Technology have been implemented on a wider scale in several countries in the past five years in the insurance industry. The technology itself is not new and has been around since the 1970 and 1980's.

Voice Stress Analysis is a technique whereby the psychological stress component associated with deception is measured by extracting micro tremors from the voice of person and measuring their amplitude.

Where traditional and cumbersome lie detectors or polygraphs measure physical responses to lies such as heart beat, blood pressure, perspiration, voice stress analyzers pick up on irregular tensions in the voices of people resulting from lies and deception in conversations. Manufacturers of this technologies which are now available on the open market, claim up to a 95% success rate in picking otherwise inaudible tensions in the voice.

The results booked with these techniques seem impressive with claimed rates of 25% reduction rates in frauds in the UK not withstanding the debated accuracy of the technique. Some studies suggest that the accuracy of Voice Stress Analysis at deterring a subjects truthfulness is not significantly better than chance.

One of the pioneers in this area is Highway, with a telephone based system using Voice Stress Analyser plus a call handler using narrative integrity analysis techniques. From several media reports it transpired that such a system is not used as evidence of fraud per se, it merely gives the call handler a chance to ask additional questions where the systems indicate a potential lie.

According to Highway insurance it saw a substantial part of their  of claims withdrawn when the questioning gets to close to home. Highway claims that the systems are mist advantageous for the honest claimants who can be fast tracked and claims can be settled quicker.

Upon the introduction of voice stress analysis  lead to considerable cost savings in investigation referrals.

 

Insurance Fraud in Florida

		Wednesday, 30 May 07 - 11:16 PM (GMT) By John ML Dierckx in Insurance Fraud

Natural disasters are terrible and leave a lot of people with severe damages. It is good that insurance is there on these moments to help you get back your life. But as always some see this as a possibility to actually gain some funds they are not entitled to. An article from Claims Mag:

Source: http://cms.nationalunderwriter.com/cms/Claims/Breaking+News/2007/05/
30-Fraud+of+the+Week

Fraud of the Week: Hurricane Claim Fraudsters Busted

Alex Sink, Florida ’s chief financial officer, recently announced that 10 people are facing felony fraud charges. The individuals are suspected of submitting $300,000 in fraudulent hurricane claims to their homeowners’ insurance companies.

So far, eight arrests have been made in the investigation by the Florida Department of Financial Services’ Division of Insurance Fraud.

Victor Hernandez, Catherine Abrignani , Martin H. Katz, Brian Friedman, Ridel Valido, Judith Davis, Tim Adams, and Scott Olsen have been arrested. Wilson Uvo and Joseph Castranova were charged, but have not been arrested yet.

 The Fraud division has charged 57 individuals with insurance fraud and related offenses stemming from suspected fraudulent storm claims following the 2004 and 2005 hurricane seasons.

Is it not plain sad to see that people are trying to make money out of what is already  a terrible situation. In New Zealand, what could this mean: recent floods could very well lead to the same temptations.

Accurate claims management and assessments will be needed to prevent damages incurred to false clams. More importantly, when will clamants understand that they are not just ripping off the nsurance company (a "victimless" crime) but also their god friends and neighbours who will be faced with higher premiums resulting from ongoing fraudulent claims

A Case For Incorporating Intelligence in Your Anti-Insurance Fraud Efforts

		Sunday, 29 April 07 - 03:59 AM (GMT) By John ML Dierckx in Insurance Fraud

In the past few weeks and coming period I have been working on several cases, next to providing training in the field of insurance fraud. Especially the latter, without excluding the cases themselves made me think again about the importance of an intelligence function in those organisations that have an institutionalised anti-fraud program.

I asked several of the students to describe the fraud problems they encountered and what structural problems they encountered. I asked them what they expected to be the future problems. I asked the students if their organisations felt hat they had things under control. I then asked if the students and their organisations felt that they understood the problem hey were dealing with.

Why was this question important? Because notwithstanding the efforts being made to counter insurance fraud, it seems most organisations are facing a growing problem.

Me being someone that likes to keep things simple, I tend to start with what I would call trying to understand the problem and subsequently based on the understanding of  that problem start working on a structured approach towards that. Could it just be that decisions being made by managements are based on quick fixes that could very well be adressing:

• not the real problem
• be an outdated problem by the time the quick fix gets incorporated
• a waste of money since simpler (read less costly and interfering) solutions are available

The current trend appears to be especially focussed on detecting fraud in an early stage. Some go as far as claiming that early detection is the best prevention. Whilst it cannot be denied that early detection is an important part of cost containment and can have a preventative effect, it still is in my view a reactive approach. Where a sound underatanding of the problem is aparent, a more pro-actove approach can more than once be adopted.

Five elements of an anti fraud strategy

In my view there are five key components in any anti-fraud strategy, the insurance industry not excluded:

• understanding your current problem and anticipating the future
  
• prevention
  
• detection
• investigation and
• resolution

These components are not seperate entities, the are interrelated. A holistic approach and sound understanding of how all these components influence eachother as well as creative approaches in the operation of them are a key to successful anti-fraud strategies.

This series of articles will especially focuss on strategic intelligence which is especially focussed at the first component: understanding your problem and anticipating the future.

What is Strategic Intelligence?

By definition, strategy is concerned with the development and use of an overal plan that incorporates all the details necessary to  arrive at a main aim. That major aim, in this situation countering fraud, can usually be divided into smaller sets of objectives and goals that have to met along he way in achiveing the overal success.

Intelligence in this context can be described as the sum of what is known, ntegrated with new information and interpreted for meaning. (See Consise Oxford Dictionary and others)

Intelligence refers to both a process and a product. We further distinguish between tactical intelligence (or operational intelligence) which deals with the day to day needs of line management and the decision making processes inherent to that and strategic intelligence, which objective it is to provide accurate, long range intelligence to enable high level planning and management to meet overall perceived threats.

Strategic Intelligence provides the senior manager and executive with insight and understanding in for instance:

• current and emerging trends
• changes in the environment
• new threats
  
• opportunities for controlling action and the development of counter programs
• likely avenues for change to policies, strategies, programs and legisation

Is will be clear from this list strategic intelligence deals with the here and now as well as the future. In this sense the role of strategic intelligence can be descibed as providing input to organisational decision making, forecasting major trends, threats, risks and opportunities and providing support and direction for operational activitiy.

In the coming articles we will have a closer look at the strategic intelligence process and outcomes (the process and product) and how this can serve your organisation in countering fraud.

Insurance Fraud & Why We need To Counter It.

		Sunday, 01 April 07 - 07:13 AM (GMT) By John ML Dierckx in Insurance Fraud

Also concerned about the price of your policy? Perhaps you should give some thought about insurance fraud. In many countries a national sport, like trying to cut the tax man short. The industry itself is concerned. In this first article on insurance fraud some general notions.

Costs to the industry
Several statistics indicate that the average percentage of fraudulent claims is somewhere between five and ten percent of the total of claims, causing the industry billions of dollars in losses world wide. The Insurance Bureau of Canada estimated an annual cost of approximately 1.3 billion annually. A recent report by AIG together with the Economist Intelligence Unit (“Hidden Costs-Insurance Fraud in Australia”) estimated the paid out fraudulent claims around AUS$ 2.1 billion annually: an average of AUS$ 73.-- per insurance policy. The National Insurance Crime Buro in the US estimated a cost of $130 billion per year.

Attitudes to Insurance Fraud
In many countries the most popular sport is “tampering your tax returns”. The smiles on many faces here taken into consideration I take it that this country is no exception. And when asked  many will deny this form of fraud, to be fraud, or at least that they are committing a crime in delivering false tax returns. A similar approach appears applicable towards Insurance Fraud. The report “Hidden Costs-Insurance Fraud in Australia”  summarised some of the public attitudes towards perpetration of Insurance Fraud.

In a survey by the insurance Council of Australia in 1993, 14% of the interviewed people interviewed regarded padding a claim as acceptable. In a 2002 survey, that percentage had grown to 18%. A similar survey by the Association of British insurers carried out in 2002, 7% of the interviewed admitted to having made a fraudulent insurance claim and more significantly 48% did not rule out making a false claim in the future. 40% of the respondents to the survey believed that exaggerating a claim was acceptable or borderline behaviour. 29% felt that making up a claim was acceptable or borderline. US figures were similar, with 24% of the respondents believing that padding a claim was acceptable and 11% of the respondents thinking that making up a claim is acceptable. What can we learn from these figures? Where adequate statistics are not available in respect of false claims and the costs, it appears not unreasonable to suspect a percentage of 10% of the claims filed each year to contain some form of a fraudulent element.

Costs to the industry
Several statistics indicate that the average percentage of fraudulent claims is somewhere between five and ten percent of the total of claims, causing the industry billions of dollars in losses world wide. The Insurance Bureau of Canada estimated an annual cost of approximately 1.3 billion annually. A recent report by AIG together with the Economist Intelligence Unit (“Hidden Costs-Insurance Fraud in Australia”) estimated the paid out fraudulent claims around AUS$ 2.1 billion annually: an average of AUS$ 73.-- per insurance policy. The National Insurance Crime Buro in the US estimated a cost of $130 billion per year.

Why Action is Important

Costs to the Industry
As outlined in previous paragraphs the industry is losing billions of dollars world wide each year as a result a fraudulent claims. But is it really that much. When statistics speak of fraudulent claim figures, they do not always incorporate other less direct costs resulting from these same frauds.

Competition Perspective
Insurance is a highly competitive market. Saving on costs is therefore most important to maintain and expand market position. It is therefore of paramount importance that insurers put effort in anti-fraud policies strategies and actions. At the same time it is perfectly reasonable to expect similar efforts from third parties such as adjusters, experts, brokers and agents who are like the insurance companies themselves in a position where they can actively contribute to keeping the industry clean and cutting the cost of insurance.

Costs to Society
Insurance fraud does not exist in a vacuum and needs to be seen in conjunction with other illegitimate and deviant behaviour. Standing up against insurance fraud therefore serves a higher purpose as well in that it contributes and promotes norm-conform behaviour and stands up against   an attitude whereby personal benefit is more important than legitimacy of one's actions. In that sense I would consider it a communal or societal obligation of insurance companies to stand up against insurance fraud instead of merely incorporating the costs into the premiums.

In addition to this insurance companies are in a position to contribute in the battle against organised crime by being aware of the risks of money laundering but also as victims of organised crime groups  perpetrating organised insurance frauds.

Costs to Insured

Insurance Fraud Results in Expensive Insurance

Insurance fraud results in premiums going up. So ultimately, part of the costs of insurance fraud are covered by innocent policy holders. No one likes to pay for something that ends up in the pockets of those that are not entitled to it.

Next to the direct costs of paying out to those who are not entitled to it, other costs will ultimately reflect in the price of the premiums as well such as detection, investigation and resolution costs. Costs that could be saved i there areas could potentially flow back to the policy holders in the form of premium prices.

Fraud Can Lead to Uninsurability
Amongst others as a result of fraud and the impossibilities to stop fraud, some insurance products were taken of the market in for instance Belgium. It is for instance hard to find insurance for your bicycle in the Netherlands against a reasonable premium price, similarly certain types of travel/holiday insurances were taken of the market. Clients become limited in their choices and some insurable interests can only be insured by those that have enough money for it. In that sense a broad range of products available to the individual also serves a social purpose.

... More items are available in our News Archive