An article in the July 2010 issue of HR Magazine —“Overseeing Audits of Your Health Plans”— pointed out the increasing use of various types of health plan audits, including the retrospective audit of medical and prescription drug claims.
For self-insured employers considering such an audit — many possibly for the first time — the analysis below drills a little deeper to point out that there are essentially three different types of retrospective audits available, and that expectations for results should be guided by the type of audit HR professionals chose to do.
Each type can be — and often is — promoted as a "100-percent-of-claims" audit, and each type can be employed to scrutinize paid claims going back one or two years. But that’s where the similarity ends.
All Claims Audits Are Not the Same
Among the different approaches to 100-percent-of-claims audits to keep in mind are:
• The commonly used “random sample audit,” which relies on an analysis of a 300 to 400-claim sample derived from 100 percent of paid claims for the audit period.
• The “focused audit,” which scans 100 percent of the claims for the audit period, looking for claim errors that fit a predetermined profile, such as incorrect co-pays collected during office visits in an outpatient setting.
• The true “100-percent-of-claims audit,” which is based on analyzing 100 percent of claims, and which scrutinizes 50 or more components of every claim. In doing so, the audit electronically compares each component to what is called for in the administrative services only (ASO) agreement and the summary plan description (SPD) — documents that detail the processes and policies of the plan’s third party administrator (TPA).
Different Methods, Different Results
Some would argue that random sampling produces an acceptable result, so Healthcare Data Management Inc. (HDM), a King of Prussia, Pa.-based provider of audits and health plan analytics, decided to put it to the test.
HDM commissioned an independent comparison study of its methodology and random sampling by two members of the faculty at Saint Joseph’s University in Philadelphia: Ronald Klimberg, a professor in the Decision and System Sciences Department of the University’s Haub School of Business, and George P. Sillup, an assistant professor, Department of Pharmaceutical Marketing, and fellow in the University’s Pedro Arrupe Center for Business Ethics.
To conduct the study, Professors Klimberg and Sillup used actual claims data from two Fortune 100 HDM client companies, anonymously identified as Companies A and B. Company A data spanned over two years and contained over 54,000 claims records. The dollar amounts of the claims paid were $12.8 million in year one and $12.5 million for year two.
Company B was larger and had about 464,000 claims that totaled $118.4 million in claims paid.
All the data sets had been subjected to a "true" 100-percent-of-claims analysis, resulting in errors-only data sets. The analysis is actually one step of a five-step protocol that includes an on-site audit to confirm the logic of the analysis. These error-only data sets were then used to conduct random sampling simulations — 100 simulations each of 300- and 400-claim random samples.
The result: random sampling missed over 90 percent of the errors caught by the methodology based on the 100-percent-of-claims analysis.
In terms of dollars, random sampling missed about $200,000 to over three quarters of a million dollars in overpayment and underpayment errors that were clearly identified by the true 100-percent-of-claims methodology.
The results from 100 random simulations of sample size 300 were compared to the “population of errors” determined by a true 100-percent-of-claims methodology. On average, the 300 random sampling missed from $1.2 million to $5.4 million of the total amount of erroneous claims paid. (HDM refers to erroneous claims as “exception” claims.) Additionally, the random samples missed overpayment claims ranging from about $150,000 to $700,000, as well as underpayment claims ranging from approximately $24,000 to $85,000.
Professors Klimberg and Sillup similarly analyzed 100 random simulations of sample size 400. The 400-sample size simulations missed from $1.2 million to $5.4 million of the dollar amount of the erroneous claims paid. For overpayments, the amount missed was $145,000 to $688,000. For underpayments, it was about $24,000 to $58,000.
Increasing the size of the random sample did not result in the identification of significantly more errors.
Other Findings
The study proved that random sampling was inferior to the other methodology on other counts. For example, because the true 100-percent-of-claims audit looks at all components of every claim, it depicts employee health care utilization more accurately.
All of the findings of the study are provided in the study white paper, Health Plan Auditing: 100-Percent-of-Claims vs. Random-Sample Audits — Look at What You’re Missing, available on the HDM web site (registration required).
David McSweeney, chief operating officer of Healthcare Data Management Inc., has over 30 years of experience as a financial and operations executive for a variety of health care organizations. He previously served as vice president for Blue Shield of New Jersey and New Jersey Delta Dental, regional vice president for United Healthcare, and president and COO for Alternative Dental Care Inc., president and CEO for Vienna Corp., and co-founder and president of American Health Fund. McSweeney additionally served as president and COO of Claims Administration Corp., a wholly owned subsidiary of CNA, where he led a $2.2-billion group healthcare and benefits enterprise.
Related Articles:
Overseeing Audits of Your Health Plans, HR Magazine, July 2010
Data Mining Claims: The Benefits of Digging Deeper, SHRM Online Benefits Discipline, July 2010
Quick Links:
SHRM Online Benefits Discipline
SHRM Online Health Care Reform web page