News / Features

Newsroom

 

Is your health care fraud detection solution ignoring valuable sources?

Julie M. Malida Principal for Health Care Fraud, Enterprise Financial Crimes Practice, SAS

In the health care industry, professional fraud often involves multiple parties that are in collusion, premeditated schemes, identity theft and organized crime. Frequently, the money evaporates in a matter of days. Consequently, it is now more important than ever to be timely in fraud detection (preferably before monies are paid out), employ multiple analytical methods of detection, and use a variety of data sources beyond just claims data.

Let's take a closer look at the most common methods of fraud detection and the frequently ignored data sources that can help improve detection.

Multiple methods of analysis
Fraud is a spectrum of activity that ranges from opportunistic deception to truly premeditated, organized schemes. Each different type of analytics attacks a different issue or type of activity.

Basic rules on prior known schemes have their place and cast a wide net for detecting suspicious activity. An example would be, "Show me all claims where the patient is traveling more than 200 miles for routine care." But rules are very linear in the way they are written, and investigators end up chasing many false positives to weed out the cases that may be explainable. Fraud rules are also very easily gamed, once the fraudster figures out how the rule is written.

Anomaly detection, which casts the net a bit wider and looks for oddities you didn't know enough about to write a specific rule around. However, such oddities show up as outliers in behaviors that don't look normal. Fraudsters can also game anomaly detection, for example, by working with multiple providers to make their schemes harder to detect through this method.

Predictive modeling applies statistical methods like decision trees and neural networks to data from prior cases of known fraud, and looks for statistical similarities. A good example is a chiropractor who treated everyone in the same family for the same low back pain twice per week until all family members' health plan benefits were exhausted. You can build a model that can look for similar characteristics to detect similar fraud schemes before they are paid. Predictive models will vastly reduce the false positives and make analysts laser-like in the claims they pursue for full investigation.

Social network analysis (also called link analysis) builds mathematical models that show the connectedness of different entities and score their statistical significance for fraud, either by looking at their activities or by looking at their personal relationships. Physicians who always send patients to a certain lab or two physicians who attended the same schoolwould be linked entities, for example. Once entities are linked, it becomes interesting to see which grouped entities violate a rule or model from one of the other methods previously described. This method allows analysts to spot organized crime and collusive behavior.

Valuable data sources often ignored
Medical claims data is the first and strongest data source for fraud investigations, but you can improve fraud detection efforts by including some of these additional data sources:

  • Member eligibility data to compare benefit coverage dates, falsification of service dates, or falsification of supply purchases. Past medical provider history, including previous sanctions against providers, their presence on state and federal "watch" lists, and any underlying motives for deception.

  • Ancillary claims like pharmacy billings, lab data and hospital records with revenue can detect whether all the care provided naturally seems to fit together as compared with accepted medical practice.

  • Structured and unstructured text data, including nurses' notes, claim processor notes in claim records, electronic medical records and call center logs, provide a valuable and rich source of investigative data that may be locked in a payer's underlying systems.

Proven value of unstructured text data
To understand the value of text analytics in detecting fraud, consider the text in call center logs. Wouldn't it be interesting to know if the same chiropractor's office called the health plan five times in one week to determine remaining benefit levels on every family member covered under the plan? By mining text data in the call center logs, this could be flagged as an outlier.

And the nurses' notes? Wouldn't it be meaningful if the notes documented a physician saying this patient needed a five-day hospital stay rather than two days due to multiple comorbidities that are not documented in the claim? The physician is either omitting information in one source or fabricating "facts" in another.

Finally, what might an audit of electronic medical records reveal? Perhaps there are far too many patients who all have the same documented height, weight and symptoms to be a normal occurrence. It could alert the investigator that someone is copying and pasting information, instead of filling out medical records accurately.

The punch line
Using multiple analytical methods and all the varied data sources available allows organizations to conduct more efficient investigations, know about fraud and abuse cases sooner, and find out pertinent information quickly. The investigator can also prepare a better bank of evidence if the case ever does proceed to law enforcement or prosecution.

It is becoming a recognized necessity for health plans to find fraud before the money goes out the door. The window of opportunity is shorter than ever due to timely payment guidelines and expectations from regulators and the marketplace.

Investigators and analysts that use anomaly detection, predictive modeling and social network analysis– combined with access to all sources of relevant data–will have the best chance of moving swiftly and accurately to detect fraud, waste and abuse before the loss occurs.

Bio: Julie Malida is the Principal for Health Care Fraud in the Enterprise Financial Crimes Global Practice at SAS. She has devoted 29 years to the health care industry, focusing on managed care, fraud and cost containment in medical claims. Malida is also a Fellow of the Society of Actuaries and a member of the American Academy of Actuaries.