Here’s how Blue Cross obtains samples for non-hospital facility audits

We occasionally get questions from our providers about our audit process for non-hospital facility audits and how we obtain samples for our audits. To help answer these questions, we’ve outlined our process below:

Why we do what we do
We obtain audit samples in preparation for extrapolating refund recoveries for non-hospital audits. These audits are conducted for:

• End-stage renal disease
• Home health care
• Hospice
• Skilled nursing facility
• Freestanding outpatient physical therapy facilities

According to our participation agreements, Blue Cross Blue Shield of Michigan may project refund recoveries from statistically valid samples involving issues other than medical necessity. These include, but are not limited to, procedure or revenue code billing errors.

Sampling for non-hospital facility audits
The goal of sampling is to get a portion of the patient case population that represents the characteristics of the entire population. The two following fundamental sampling concepts ensure the sample is a good representation of the population:

• The process of random audit case selection
• The size of the sample

Random sampling
The sampling methodology used for non-hospital audits is computer-generated and random. The process places an equal probability of selection on every patient at a facility within a specific timeframe.

Sample size
This is determined by two values: confidence and precision. It’s these values that determine the risk level associated with the sample estimates. The acceptable values are 95 percent for confidence and 10 percent for precision. This means that Blue Cross wants to be 95 percent confident that the sample compliance error rate is not different from the population compliance error rate by more than 10 percent. This produces a sample size large enough that the estimates obtained through the audit accurately represent population characteristics.

Extrapolated overpayment
After the sample is selected and the audit is complete, the sample overpayment per patient is calculated. This number is the sum of the payments associated with compliance errors in the sample, divided by the number of patients in the sample. Since the sample is a good representation of the population, it follows that the overpayment per patient in the sample will not differ much from the overpayment per patient in the population. So the extrapolated overpayment for the population is the average overpayment per patient in the sample, multiplied by the number of patients in the population.

If you have any questions regarding this process, call Rhonda Thomas at 313-448-1286.