Prior to the selection
of the Statistically Valid Random Sample (SVRS), the claims universe
shall also be properly focused and analyzed to determine the sampling
plan and methodology. Focusing the universe is performed by targeting
specific claims which match the approach and/or allegations of the case,
and removing unnecessary low dollar claims. The overall sampling
plan and methodology may include a stratified sampling approach
consisting of one or more SVRS(s) and/or 100% claims audit(s).
|
In order to determine
the probable scope and extent of overpayments, regardless of how
the overpayment was incurred, a SVRS shall be drawn from each identified
stratum (or the entire universe of claims in some cases). Denied
claims or claims where TRICARE paid zero dollars shall always be removed
from the Universe prior to stratification and sampling. Only netted
records shall be used.
|
This primary sample
shall be selected using a random number generator with a known seed
number. Using a known non-zero seed number is critical, as it will
provide for the reproduction of the same set of random numbers with
the same sample and universe.
|
The sample size
shall be calculated using the following parameters:
|
• 90% confidence level
• 10% precision level
• 50% occurrence rate (if there is no established
rate of occurrence) or an estimate of the occurrence rate from a
previously documented statistically valid analysis (by a Federal
health care entitlement program) of the units of audit (e.g., same
provider, same procedures, same time period) of the possible fraud.
|
An oversample
of 20% shall always be randomly selected from the specific stratum
or universe, and audited with the primary sample at the beginning
of an audit.
|
In all claim audits
using statistical techniques to extrapolate findings of a sample
to a specific stratum or a universe of claims, the audit addresses
the average overpayment per claim as the single unit of measurement.
The claim and the explanation of benefits are the evidentiary documents
which demonstrate the billed services submitted by a provider or
beneficiary and the payments made to a provider or beneficiary.
The claim is compared to the contents of the medical record to validate whether
a service was provided, whether it was provided at the level billed,
whether it was provided by the authorized provider shown on the
claim, or any other information which may be relevant to identify a
dollar loss to the Government. This information shall be recorded
on a summary spreadsheet generated by Microsoft® Excel, or compatible
software, with a .xls file extension for compatibility with other
widely used spreadsheet software (typically, the DHA Random Sample
Audit Worksheet shall be included for each SVRS). Each claim in
the sample shall be listed on the summary spreadsheet and the overpayment
totaled. When no overpayment exists, the claim shall appear on the
summary spreadsheet with zero listed as the overpayment.
|
Each claim of
the audited oversample shall also be included with the case, either
as part of the summary spreadsheet or as part of a separate spreadsheet.
|
The overpayments
shall be expressed in dollars and cents. The total shall then be
summed and divided by the number of claims in the sample (remembering
that claims with no overpayments are shown in the column to be summed
as zero). The product is the average mean overpayment per claim
in the sample. The average mean overpayment per claim in the sample
shall be multiplied by the number of claims in the specific stratum
or universe from which the sample and oversample was taken, and
this product expressed in dollars and cents is the extrapolated
dollar loss to the Government for that stratum or universe.
|
|
DETERMINING
EXTRAPOLATION AMOUNT AND VALIDATING THE AUDIT FINDINGS
|
It is necessary
to calculate the standard deviation, standard error of the mean,
and sampling error. The contractor shall have the electronic capability
to accomplish these calculations and shall execute the computations
according to the methodology provided in the following paragraphs.
|
In the sample
technique discussed in the previous section, if the sample has been
properly designed and selected, and the specific stratum or universe
approximates a normal distribution appropriately, there are 90 chances
in 100 that the claim overpayments will fall within the range of
the arithmetical mean plus or minus 1.645 times the calculated standard
deviation. Additional values shall be calculated as well, to determine
the validity of the overpayment estimates.
|
Calculating
the standard deviation of the sample: The standard deviation,
which is expressed in dollars and cents, shall be determined using
the following steps:
|
1. Calculate the difference
between each claim observation and the average mean overpayment.
|
2. Square each of the
calculated differences.
|
3. Sum the Squares of
the differences for all of the claim observations.
|
4. Divide the Sum of the
Squares by the number of observations in the sample.
|
Note: When the sample
size is less than 40, Divide the Sum of the Squares by the number
of observations minus one.
|
5. Take the Square Root
of the Divided Sum of the Squares.
|
|
Calculating
the standard error of the mean: The standard error of the
mean shall be calculated by dividing the standard deviation by the
square root of the sample size.
|
Calculating
the sampling error and overpayment estimate range: The sampling
error shall be calculated by multiplying the standard error of the
mean by the “Z” score (for a 90% confidence level the “Z” score
is 1.645). The “Z” score changes as the confidence level changes.
|
Calculating
the precision value: The precision value, expressed in dollars
and cents, shall be calculated by multiplying the sampling error
by the number of claims in the stratum or universe.
|
Calculating
the overpayment estimates: The overpayment point estimate
was calculated above by multiplying the average mean of overpayment
per claim by the number of claims in the specific stratum or universe.
The high and low (plus or minus) estimates of overpayments shall
be calculated respectively by adding and by subtracting the precision
value from the overpayment point estimate. The overpayment estimates
shall be expressed in dollars and cents.
|
Calculating
the sample precision percentage: The sample precision percentage
shall be calculated by dividing the precision value by the overpayment
point estimate. The desired precision percentage is 10% or less
for tight precision, with approximately 20% or more representing
low precision.
|
Testing the validity
of the sample and the overpayment estimates:
|
1. If the standard deviation
is greater than two times the arithmetic mean, this is an indicator
that the sample does not demonstrate the confidence level required
for validity.
|
2. If the high estimate
of overpayments is greater than the specific stratum or universe
amount or the low estimate of overpayments is less than zero, then
the computed overpayment amount shall not be used.
|
3. When high precision
is not achieved, the lower overpayment estimate shall be used as
the amount of overpayment demanded, as opposed to the point estimate.
This procedure yields a conservative demand amount for recovery
that is very likely less than the true amount of overpayment, and
it allows a reasonable recovery without requiring the tight precision
that might be needed to support a demand for the point estimate.
|
Alternate
Sampling Methods
|
If the tests for
the validity of the sample and overpayment estimates are not met,
it may be an indicator that the universe should have been stratified
appropriately, or other techniques should be used. If this is the
case, consult with DHA PI. For example, if there are services subjected
to audit where there are large differences in payments (e.g., surgical
and medical), there will likely be a need to stratify the universe
into two or more separate categories for separate sample selection.
When stratification is necessary and after consulting with DHA PI,
please seek consultation for such sample techniques from a qualified
statistician.
|
The standard reference
for auditing with samples is the Handbook of Sampling for Auditing
and Accounting, Third Edition, by Herbert Arkin, McGraw-Hill Book
Company, copyright 1984.
|