Physician's Ambulance Serv., Inc. v. Ohio Dept. of Medicaid

[Cite as Physician's Ambulance Serv., Inc. v. Ohio Dept. of Medicaid, 2020-Ohio-6842.]
IN THE COURT OF APPEALS OF OHIO
TENTH APPELLATE DISTRICT
Physician's Ambulance Service, Inc., :
Appellant-Appellant, :
No. 20AP-32
v. : (C.P.C. No. 18CV-2071)
Ohio Department of Medicaid, : (REGULAR CALENDAR)
Appellee-Appellee. :
D E C I S I O N
Rendered on December 22, 2020
On brief: Webster & Associates Co., LPA, and Geoffrey E.
Webster, for appellant. Argued: Geoffrey E. Webster.
On brief: Dave Yost, Attorney General, and Cheryl R.
Hawkinson, for appellee. Argued: Cheryl R. Hawkinson.
APPEAL from the Franklin County Court of Common Pleas
SADLER, P.J.
{¶ 1} Appellant-appellant, Physician's Ambulance Service, Inc. ("PAS"), appeals
from a judgment of the Franklin County Court of Common Pleas affirming an adjudication
order issued by appellee-appellee, Ohio Department of Medicaid ("ODM"). For the reasons
that follow, we affirm.
I. FACTS AND PROCEDURAL HISTORY
{¶ 2} PAS is a licensed ambulette service that provides transportation services to
Medicaid recipients under the terms of a provider agreement with ODM that requires ODM
to provide Medicaid reimbursement to PAS pursuant to applicable Medicaid regulations.
In 2016, the Auditor of State ("AOS") conducted an audit at the request of ODM of all
services provided by PAS from January 1, 2012 to December 31, 2014. After examining
No. 20AP-32 2
documentation PAS provided to AOS, conducting interviews, and performing a statistical
analysis, AOS concluded PAS was overpaid $107,120.99. In January 2017, ODM issued a
proposed adjudication order requiring PAS to repay that amount plus interest. PAS timely
requested administrative review and a hearing on the proposed adjudication order.
{¶ 3} An administrative hearing examiner conducted an evidentiary hearing on the
matter on September 18 and 19, 2017. At the hearing, ODM presented the testimony of
three AOS employees: Kristi Erlewine, Chief Auditor of the Medicaid Contract Audit
Section; Sherri Couts, Senior Audit Manager in the Medicaid Contract Audit Section; and
Norman Hofmann, AOS Senior Data Analyst. PAS presented the testimony of an expert
witness, Dr. Douglas Wolfe, former Chairman of the Department of Statistics at The Ohio
State University.
{¶ 4} Erlewine testified the AOS audit team met with PAS, provided PAS with a
Medicaid questionnaire, and obtained all relevant records pertaining to a total of 26,727
Medicaid service events during the audit period. Because of a change in the applicable
Medicaid rules, however, it was necessary to segregate certain records pertaining to services
provided to nursing home residents prior to January 1, 2014 and audit them separately
under what AOS referred to as the "exception test." (Tr. Vol. I at 26.) The entire population
of the segregated records were audited and the result was a $591.08 overpayment, an
amount which is not disputed by PAS.1
{¶ 5} Couts testified she and a team of analysts reviewed the data collected from
PAS for the remaining Medicaid service events. According to Couts, "[o]nce we've selected
a provider and the examination period and the service codes that we want to test, we ask
[Hofmann] to pull the statistical sample. And then when we're finished with our work, we
ask him to perform a projection or extrapolation of the errors." (Tr. Vol. I at 21-22.) To
that end, Couts provided Hofmann with the information contained in the audit summary.
(State's Ex. 12.) The types of overpayments revealed by the audit summary included
services provided without a certificate of medical necessity, trip documentation, and/or
qualified drivers.
1Under the new rule, Medicaid made payments for ambulette services directly to the nursing home and the
nursing home was to pay the ambulette company.
No. 20AP-32 3
{¶ 6} Hofmann testified he has worked for AOS for 22 years, and he currently holds
the position of Senior Analyst. Hofmann is a certified public accountant in Maryland and
Ohio, and he is also a chartered global management accountant. He holds professional
certifications as an information systems auditor, fraud examiner, information technologies
professional, and government financial manager. Hofmann identified ODM Exhibit 38 as
the current AOS Audit Manual.
{¶ 7} According to Hofmann, after he was contacted by Couts, he downloaded the
relevant PAS information and utilized a computer software program known as RAT-STATS
to determine the sample size needed to make an accurate projection of the Medicaid
overpayment for the 26,727 service events in the audit period. RAT-STATS revealed that
the audit team would need to examine 214 service event records in order to yield a
statistically valid projection of the total overpayment for the audit period. Hofmann then
used the simple random sample function in RAT-STATS to select the 214 service event
records which were then examined by the AOS audit team. (State's Ex. 46.) The audit
revealed that 76 of 214 service records contained overpayments to PAS in the total amount
of $2,923.15, an amount which is not disputed by PAS.
{¶ 8} According to Hofmann, the AOS Audit Manual contains a cascading
approach to sampling for purposes of projecting overpayments. The three methods
authorized by the Audit Manual are Method I, direct projection, Method II, indirect
projection, and Method III, one-tailed lower limit projection.2 Method I estimates the
overpayment by taking the average overpayment (difference values) and multiplying it by
the sample population and then projecting the overpayment to the total population.
Hofmann testified he could not use Method I to project the overpayment in this case
without substantially increasing the sample size. In other words, Method I would not meet
the "goal of achieving 10 percent precision at the 95 percent confidence level." (State's Ex.
38 at 001371.)3 Hofmann's calculation reveals that "[d]irect projection of overpayment
amounts results in a Point Estimate of $103,430; however the precision on this estimate
does not meet the AOS requirements for use of a point estimate." (State's Ex. 47 at 3.)
2A fourth method, known as a regression analysis, is not authorized by the Audit Manual.
3 The evidence shows that statistical precision is a measure of how close two or more estimates are to each
other. A confidence level is the chance the statistical mean of the sample is within one standard deviation of
the statistical mean for the entire population.
No. 20AP-32 4
{¶ 9} Hofmann testified, however, by using Method II he was able to accurately
calculate the Medicaid overpayment to PAS without increasing the sample size. According
to the Audit Manual, "Method II[] projects the audited or correct payment amount per
sampling unit across the population and then subtracts the sum from the actual amount
paid." (Emphasis sic.) (State's Exhibit 38 at 001371.) Hofmann testified Method II,
indirect projection, is authorized by the Audit Manual when three conditions are met: 1)
the sample size is sufficient to meet Cochran's approximate test for normality;4 2) the
projected payment amount from the sample is within one standard error of the actual
recorded payment amount for the population; and 3) the precision for the point estimate
of audited or correct payment amount is within 10 percent at the 95 percent confidence
level. (State's Ex. 38 at 001371.) According to Hofmann, in 2001, AOS added a chapter to
the Audit Manual authorizing the use of Method II in calculating overpayments. Hofmann
stated he was one of the co-authors of the new chapter.
{¶ 10} Hofmann acknowledged that "[i]f the criteria for using Method [II] were not
present, [AOS] would have used Method [III]." (Tr. Vol. I at 143.) Hofmann maintained
AOS went with Method II because that method yielded a better/lower precision level of
8.96 percent at the higher confidence level of 95 percent. Hofmann believed the estimated
overpayment calculated pursuant to Method III would be a "conservative estimate." 5 (Tr.
Vol. I at 141.)
{¶ 11} Utilizing Method II and the sample overpayment of $2,923.15, Hofmann
projected a total overpayment to PAS for the audit period of $106,529.91. Hofmann then
added the $591.08 overpayment under the exception test to the projected overpayment of
$106,529.91 for the remaining events population to reach a total overpayment for the audit
period of $107,120.99. See State's Ex. 47.
{¶ 12} On cross-examination, Hofmann testified he input an average error rate of
70 percent into RAT-STATS in order to come up with the required sample size of 214,
because a 70 percent error rate is the historical error rate for ambulette services audited by
4 Cochran's test is a measure of sample skewness: how close the distribution of values is to the desired "bell-
shaped curve," also known as a "Gaussian" distribution. (Tr. Vol. I at 97; Tr. Vol. II at 12.)
5 Hofmann's calculation of the overpayment using Method III reveals a lower limit of $85,715 and an upper
limit of $121,144 at a 90 percent confidence level with 17.13 percent precision. (State's Ex. 47 at 2.)
No. 20AP-32 5
AOS over the last 5 years. (Tr. Vol. I at 92-93.)6 Hofmann explained that he has compiled
a database containing results of prior Medicaid audits conducted by AOS, and he can
recover historical audited error rates for several different types of Medicaid services
including ambulette, ambulance, schools, personal care assistance, nurses, physicians, etc.
Hofmann acknowledged he could not use Method I to project the overpayment without
substantially increasing the sample size because the 214 audited event records contained
too many zero values (event records with no overpayment). Hofmann admitted he did not
know what the actual error rate in the sample would be when he entered the data into RAT-
STATS to determine the minimum sample size.
{¶ 13} Hofmann testified Method II could be used within AOS guidelines because
Method II considers projected audited values for the sample (correct Medicaid payment),
rather than the projected difference values (the amount of the overpayment). Hofmann
maintained that Method II "indirectly" considers difference values only to the extent that
payment made for the zero value event records are used to project the audited values for
the sample. (Tr. Vol. I at 184.) Hofmann stated that, in general, a sample with a significant
number of zero values negatively impacts precision at a 95 percent confidence level under
Method I but enhances precision at that confidence level under Method II.
{¶ 14} PAS called Dr. Wolfe, an expert statistician, as their only witness. Dr. Wolfe
is the retired Chair of the Department of Statistics at The Ohio State University. He holds
a Ph.D. in mathematical statistics, and he is a Fellow of the American Statistics Association,
the Institute of Mathematical Statistics, and the International Institute of Statistics. He has
published more than 100 peer reviewed journal articles, primarily in the field of statistical
methodology. Dr. Wolfe admitted that he is not an expert in the field of auditing and that
he has no criticism of the work performed by the auditing team in this case.
{¶ 15} Dr. Wolfe had three basic criticisms of the AOS's findings in this case. First,
he opined the sample was too small given the fact that only 76 of the 214 audited event
records in the sample revealed an overpayment. Dr. Wolfe testified AOS should have
employed a preliminary sample before determining the required sample size, and he did
not believe the zero values should have been counted as part of the sample. Second, he was
critical of the data AOS chose to examine. Dr. Wolfe believed the projected difference
6 Hofmann was not asked to identify the various audit periods represented in the database.
No. 20AP-32 6
values should have been used in the calculation rather than the projected audited values.
Third, he disagreed with the statistical analysis employed by AOS to determine the
overpayment. The crux of Dr. Wolfe's opinion testimony was that Method II is not a
statistically valid method for projecting Medicaid overpayments and that Method I should
have been used in this case. In Dr. Wolfe's opinion, even if Hofmann had increased the
sample size, it would not have been appropriate to use Method II. Dr. Wolfe explained how
AOS should have determined the overpayment as follows:
Q. So when we're using the indirect method, when the State
– when the Auditor's Department used the indirect method
here, they only project the audit values. They didn't – they
ignored the paid values, basically?
A. That's right.
Q. Okay. So when we have a situation where the precision is
so high, so far beyond the accepted precision level, how do you
know – how do you properly go about correcting an
unacceptably higher precision?
A. Take a larger sample size.
And easy to do, in process. I don't know whether it is in
practice in terms of amount of money involved. But I have a
set unit in my sample. All you have to do is run the same seed
in the program, and you'll get that same first 214 and it will
add up to 120, or 220 or whatever is necessary for it.
A good process would be to go back and use the 35 percent
error rate that was found for this particular client, put that
into his RATSTATS program and let it tell you how large a
sample size you would have to have.
And I believe [Hofmann] alluded a guess that would have met
700-and-some, and they decided not to do that.
Q. Had he done that – if you increase your sample size would
you still use Method II?
A. No.
I wouldn't ever use Method II.
(Emphasis added.) (Tr. Vol. II at 26-27.)
{¶ 16} On January 8, 2018, the hearing examiner issued a report wherein it was
recommended "the Director implement the audit findings and proposed adjudication order
dated January 20, 2017, with a finding that [PAS] owes ODM $107,120.99 for overpayment
of Medicaid charges billed during the period of January 1, 2012 through December 31, 2014,
No. 20AP-32 7
plus interest." (Report & Recommendation at 10.) On February 21, 2018, ODM issued the
recommended adjudication order. On March 8, 2018, PAS filed an appeal pursuant to R.C.
119.12 and 5164.38.
{¶ 17} On December 18, 2019, the trial court issued a judgment entry affirming the
adjudication order on finding it "is supported by reliable, probative and substantial
evidence and is in accordance with law." (Trial Court Decision at 7.) PAS timely appealed
to this court from the December 18, 2019 judgment.
II. ASSIGNMENTS OF ERROR
{¶ 18} PAS assigns the following as trial court error:
[1.] The trial court erred as a matter of law by failing to find
the prima facie presumption drops out of the case once the
department offers evidence, i.e. Couts and Hofmann's
testimony.
[2.] The trial court erred as a matter of law by relying on
Omnicare to universally validate Method II when the state
used the "known error rate" in Omnicare and here, the state
used an arbitrary determined, unsupported "anticipated error
rate".
[3.] The trial court erred in failing to require the State to
comply with its own manual in contravention to Medcorp Inc.
v. ODJFS.
[4.] The trial court erred in failing to recognize or distinguish
years of case law invalidating Indirect Projection.
[5.] The trial court erred as a matter of law by recognizing
Hofmann's testimony as expert opinion when Hofmann was
never identified as an expert.
[6.] The lower court erred as a matter of law when it based its
opinion on a weight of the evidence given there was only one
expert opinion offered by Appellant, the state failed to offer
any expert opinion and the lower court cited to "lack of written
authority" as a bases for wholly discounting Dr. Wolfe's expert
testimony.
III. STANDARD OF REVIEW
{¶ 19} Review of an adjudication order issued by ODM is governed by R.C. 119.12.
Meadowwood Nursing Facility v. Ohio Dept. of Job & Family Servs., 10th Dist. No. 04AP-
732, 2005-Ohio-1263, ¶ 8. Pursuant to R.C. 119.12, a court of common pleas must
determine whether an agency's decision is supported by reliable, probative, and substantial
No. 20AP-32 8
evidence and is in accordance with law. R.C. 119.12(M); Pons v. Ohio State Med. Bd., 66
Ohio St.3d 619, 621 (1993). Reliable, probative, and substantial evidence is defined as
follows:
(1) "Reliable" evidence is dependable; that is, it can be
confidently trusted. In order to be reliable, there must be a
reasonable probability that the evidence is true.
(2) "Probative" evidence is evidence that tends to prove the
issue in question; it must be relevant in determining the issue.
(3) "Substantial" evidence is evidence with some weight; it
must have importance and value.
Bartchy v. State Bd. of Edn., 120 Ohio St.3d 205, 2008-Ohio-4826, ¶ 39, quoting Our
Place, Inc. v. Ohio Liquor Control Comm., 63 Ohio St.3d 570, 571 (1992).
{¶ 20} The Supreme Court of Ohio in Univ. of Cincinnati v. Conrad, 63 Ohio St.2d
108 (1980), expounded on the reliable, probative, and substantial evidence standard as
follows:
[D]etermining whether an agency order is supported by
reliable, probative and substantial evidence essentially is a
question of the absence or presence of the requisite quantum
of evidence. Although this in essence is a legal question,
inevitably it involves a consideration of the evidence, and to a
limited extent would permit a substitution of judgment by the
reviewing Common Pleas Court.
In undertaking this hybrid form of review, the Court of
Common Pleas must give due deference to the administrative
resolution of evidentiary conflicts. For example, when the
evidence before the court consists of conflicting testimony of
approximately equal weight, the court should defer to the
determination of the administrative body, which, as the fact-
finder, had the opportunity to observe the demeanor of the
witnesses and weigh their credibility. However, the findings of
the agency are by no means conclusive.
Where the court, in its appraisal of the evidence, determines
that there exist legally significant reasons for discrediting
certain evidence relied upon by the administrative body, and
necessary to its determination, the court may reverse, vacate or
modify the administrative order. Thus, where a witness'
testimony is internally inconsistent, or is impeached by
evidence of a prior inconsistent statement, the court may
properly decide that such testimony should be given no weight.
Likewise, where it appears that the administrative
determination rests upon inferences improperly drawn from
No. 20AP-32 9
the evidence adduced, the court may reverse the administrative
order.
Id. at 111-12, citing Andrews v. Bd. of Liquor Control, 164 Ohio St. 275 (1955).
{¶ 21} " '[A]n administrative appeal to the common pleas court does not provide a
trial de novo * * * but rather "the Court of Common Pleas must give due deference to the
administrative resolution of evidentiary conflicts," [a]nd where the agency's decision is
supported by sufficient evidence and the law, the common pleas court lacks authority to
review the agency's exercise of discretion, even if its decision is "admittedly harsh." ' "
Beauchamp v. Petit, 10th Dist. No. 17AP-591, 2018-Ohio-1164, ¶ 23, quoting Capital Care
Network of Toledo v. Ohio Dept. of Health, 153 Ohio St.3d 362, 2018-Ohio-440, ¶ 25,
quoting Conrad at 111, and Henry's Cafe, Inc. v. Bd. of Liquor Control, 170 Ohio St. 233,
236-37 (1959).
{¶ 22} " ' "In reviewing an order of an administrative agency, an appellate court's
role is more limited than that of a trial court reviewing the same order." ' " Bartchy at ¶ 41,
quoting Rossford Exempted Village School Dist. Bd. of Edn. v. State Bd. of Edn., 63 Ohio
St.3d 705, 707 (1992), quoting Lorain City Bd. of Edn. v. State Emp. Relations Bd., 40 Ohio
St.3d 257, 260-61 (1988). See also Hand & Hand MRDD Residential Servs., Inc. v. Ohio
Dept. of Dev. Disabilities, 10th Dist. No. 17AP-116, 2017-Ohio-9287, ¶ 12. An appellate
court does not weigh the evidence but, rather, only determines whether the trial court has
abused its discretion. Bartchy; Hand & Hand MRDD. The term "abuse of discretion"
connotes more than an error of law or judgment; it implies that the court's attitude is
unreasonable, arbitrary, or unconscionable. Hand & Hand MRDD at ¶ 12, citing
Blakemore v. Blakemore, 5 Ohio St.3d 217, 219 (1983). "Absent an abuse of discretion, an
appellate court must affirm the common pleas court's judgment, even if the appellate court
would have arrived at a different conclusion than that of the common pleas court."
Edmands v. State Med. Bd., 10th Dist. No. 16AP-726, 2017-Ohio-8215, ¶ 11, citing Bartchy
at ¶ 41-42. For issues of law, including whether the common pleas court applied the proper
standard of review, an appellate court's scope of review is plenary. Bartchy at ¶ 43.
No. 20AP-32 10
IV. LEGAL ANALYSIS
A. First Assignment of Error
{¶ 23} By the first assignment of error, PAS contends the trial court erred in
affirming the adjudication order because the hearing examiner misapplied an evidentiary
presumption in favor of ODM. We disagree.
{¶ 24} Ohio Adm.Code 5160-70-06(A)(4) provides in relevant part: "Any * * *
document that supports the issuance of the notice of intended action issued by ODM, if
offered into evidence, constitutes, regardless of consent of any party, prima facie evidence
sufficient to establish the facts contained therein and that, if not rebutted by the party, is
sufficient to sustain a determination that ODM has met its burden of proof. The party
carries the burden of production to rebut the prima facie evidence." The hearing examiner's
report contains the following factual finding:
9. The evidence presented by Respondent through the
testimony of Dr. Douglas Wolfe, was not sufficient to rebut the
prima facie presumption of the validity of Method II
established through the evidence presented by ODM. O.A.C.
5160-70-06(A)(4).
(Emphasis sic.) (Report & Recommendation at 8.)
{¶ 25} PAS argues the above-cited factual finding indicates the hearing examiner
inappropriately applied the evidentiary presumption found in Ohio Adm.Code 5160-70-
06(A)(4), even though ODM presented evidence at the hearing in support of the
adjudication order. We disagree.
{¶ 26} In Meadowwood Nursing Facility, 2005-Ohio-1263, this court noted that
"[p]rima facie evidence has been defined as that which is 'sufficient to support but not to
compel a certain conclusion and does no more than furnish evidence to be considered and
weighed but not necessarily accepted by the trier of the facts.' " Id. at ¶ 14, quoting
Cleveland v. Keah, 157 Ohio St. 331, 337 (1952). "But such evidence is not conclusive,
rather, 'the term denotes evidence which will support, but not require, a verdict in favor of
the party offering the evidence.' " Id. at ¶ 14, quoting Krischbaum v. Dillon, 58 Ohio St.3d
58, 64 (1991).
{¶ 27} In Cotterman v. Ohio Dept. of Pub. Welfare, 28 Ohio St.3d 256 (1986), the
Supreme Court examined a prior version of Ohio Adm.Code 5160-70-06(A)(4) and
discussed the impact of the rule in a case involving alleged overpayments received by a
No. 20AP-32 11
Medicaid service provider. The Ohio Department of Public Welfare, now ODM, had
determined the overpayment by auditing 123 sample cases collected randomly from PAS's
Medicaid records and, during the hearing, the state presented evidence concerning 17 of
the cases. The issue for the Supreme Court on appeal was whether the prima facie
presumption was lost with respect to all 123 cases or just the 17 cases to which evidence was
presented. The court evaluated former Ohio Adm.Code 5101:3-50-22 and determined the
presumption placed the burden of production on the provider to rebut each sample case.
The court noted, however, the 17 cases for which evidence was presented by the agency,
"the presumption would ab initio be inapplicable." (Emphasis sic.) Cotterman at 258,
citing Ayers v. Woodard, 166 Ohio St. 138 (1957), paragraph three of the syllabus. See also
Meadowwood Nursing Facility at ¶ 14-15.
{¶ 28} Here, the language of Finding of Fact 9 references a prima facie presumption
but that reference is modified by the phrase "the validity of Method II established through
the evidence presented by ODM." (Emphasis added.) (Report & Recommendation at 8.)
Accordingly, we do not read Finding of Fact 9 as a clear expression of the hearing
examiner's application of an evidentiary presumption in favor of ODM. Moreover, the
report and recommendation also contains the following finding of fact:
10. From the evidence presented in the administrative
hearing ODM established by a preponderance of the evidence
that the overpayment for services under the Medicaid
program was [$107,120.99] for the audit period of January 1,
2012 to December 31, 2014.
(Emphasis added.) (Report & Recommendation at 9.)
{¶ 29} Similarly, the hearing examiner's conclusions of law provide in relevant part:
3. Where the State establishes by a preponderance of the
evidence, as is the case here, that its audit and proposed
adjudication order conform to the provisions of R.C. Chapter
5160 and are otherwise supported by the evidence and are in
accordance with law, the State has met its burden of proof.
The State demonstrated that PAS received Medicaid
payments to which it was not entitled because it provided
services without having Certificates of Medical Necessity, trip
documentation, and qualified drivers. The State of Ohio is
entitled to an Order directing Respondent to comply with the
audit findings, the terms of which are incorporated by
reference.
No. 20AP-32 12
(Report & Recommendation at 9.)
{¶ 30} In our view, the above-cited findings of fact and conclusions of law show that
the hearing examiner applied the appropriate evidentiary burden in reaching the decision
in this case. Meadowwood Nursing Facility at ¶ 14-15. Thus, the trial court did not adopt
an erroneous evidentiary presumption in reviewing the ODM adjudication order.
{¶ 31} For the foregoing reasons, PAS's first assignment of error is overruled.
B. Second and Sixth Assignments of Error
{¶ 32} Because the second and sixth assignments of error are related, we will
consider them together. In the second assignment of error, PAS argues the trial court erred
when it relied on this court's prior decision in Omnicare Respiratory Servs. v. Ohio Dept.
of Job & Family Servs., 10th Dist. No. 09AP-547, 2010-Ohio-625, in affirming the
adjudication order. In the sixth assignment of error, PAS contends the trial court erred, as
a matter of law, when it determined Hofmann's testimony provided reliable, probative, and
substantial evidence to support the projected overpayment even though PAS's expert
witness opined the methodology employed by AOS was flawed.
{¶ 33} In Omnicare, the provider, Omnicare, sought review of a decision by the
common pleas court affirming an Ohio Department of Job and Family Services' ("ODJFS")
determination of a Medicaid overpayment. To determine the amount of the overpayment,
ODJFS asked AOS to conduct a detailed examination of a sample of 246 audited claims, in
which 44 overpayment errors were found amounting to $6,097.92.7 By means of an
indirect method of statistical analysis, Method II, AOS projected the known errors onto the
population of 22,000 claims and derived an overpayment of $418,730.87. This amount
was included in the total overpayment of $1,978,108.65.
{¶ 34} At the administrative hearing, Omnicare presented the testimony of Dr.
Wolfe who challenged the expert presented by ODJFS, Dr. Melvin Moeschberger, on a
number of points including the validity of indirect projection as a means to accurately
determine the total overpayment. Dr. Wolfe testified about the flaws in Method II and
claimed Dr. Moeschberger did not review the statistical work done in the audit before
concluding the method produced an accurate result. The common pleas court concluded
7The Omnicare decision does not reveal the error rate AOS used to arrive at a sample size of 246. Nor is there
any indication in Omnicare that AOS used a preliminary or pilot sample in arriving at the sample size of 246.
No. 20AP-32 13
ODJFS violated due process in determining the proper Medicaid reimbursement rate for
certain services but also found that Dr. Moeschberger's testimony provided support for the
statistical analysis employed by AOS in determining the overpayments made for certain
other services. ODJFS appealed to this court, and the provider cross-appealed.
{¶ 35} One of the issues raised in Omnicare's cross-appeal to this court was whether
the common pleas court abused its discretion when it deferred to the hearing examiner's
determination that AOS used a statistically valid method of determining the overpayment.
This court found as follows:
The hearing examiner credited the testimony of Dr.
Moeschberger over that of Dr. Wolfe. The hearing examiner
noted that the projection method used by the auditor is
described in elementary statistical textbooks and the
calculations were standard in the field. The hearing examiner
further found that Dr. Wolfe did not cite to authority other
than his own professional opinion in challenging Dr.
Moeschberger's approach. The trial court found substantial,
reliable, and probative evidence to support the statistical
analysis used in the audit. We find no abuse of discretion in
this finding.
Id. at ¶ 34.
{¶ 36} Though the universe of claims in Omnicare was lower than this case and the
sample size was a bit larger, the sample in Omnicare contained a far greater percentage of
zero values than the sample in this case, and the projected overpayment was much larger.
Yet this court concluded the trial court did not abuse its discretion when it found the
calculated overpayment reliable.
{¶ 37} The Omnicare case is squarely on point both factually and legally as there is
no question the statistical methodology applied by AOS to determine the Medicaid
overpayment in Omnicare is the same methodology employed by AOS in this case. PAS also
called the same expert statistician to discredit Method II, Dr. Wolfe. The primary difference
between this case and Omnicare is that ODM relied on Hofmann's testimony to explain and
defend AOS's findings rather than retaining an outside expert witness.
{¶ 38} The trial court's decision in this case provides in relevant part as follows:
Here, like in Omnicare, the Hearing Officer considered Mr.
Hofmann's education, experience and certificates. The
Hearing Officer also noted that "Method II" is recognized by
the American Institute of Certified Public Accountants. In
No. 20AP-32 14
contrast, the Hearing Examiner stated that although Dr.
Wolfe was certainly a qualified expert in the field, he failed to
cite to any independent authority supporting his conclusion
that "Method II" is invalid. As such, the Hearing Officer, like
the one in Omnicare, credited the testimony of Mr. Hofmann
over Dr. Wolfe.
Based upon a thorough review of the record, the parties'
briefs, and the relevant law, this Court finds that Adjudication
Order of the Ohio Department of Medicaid is supported by
reliable, probative and substantial evidence and is in
accordance with law.
(Trial Court Decision at 7.)
{¶ 39} Here, Dr. Wolfe compared Method II to the study of a weight loss program
where the examiner projects weight loss over the entire population using only the end
weight of the sample subjects. Dr. Wolfe posited that such a study would not be a valid
measure of the efficacy of the weight loss regimen because it ignores the beginning weight
of the sample subjects and excludes data from subjects who lost no weight. The hearing
examiner, however, put several questions to Dr. Wolfe that evidenced disagreement with
Dr. Wolfe's analogy. (Tr. Vol. II at 48-59.) At the close of the exchange, the hearing
examiner stated: "Okay. I think I understand what you're saying." (Tr. Vol. II at 59.) The
hearing examiner subsequently issued a report and recommendation wherein he discussed
Dr. Wolfe's testimony:
Wolfe used an analogy of a group of people placed on a diet
who were later weighed to check weight loss. He contended
that the critical data was how much each person lost, not what
they weighed before or after the diet by itself. He used this
analogy in an effort to undermine the value of the "paired
data," meaning the difference between starting weight and
finishing weight. The Hearing Officer finds this analogy to be
unconvincing. In a large group of people who start a diet,
those in the group who neither gain nor lose weight—the
''zeroes"—have just as much predictive value as those who
gain or lose weight.
(Report & Recommendation at 6.)
{¶ 40} The transcript of the administrative hearing also shows the hearing examiner
questioned Dr. Wolfe extensively about his criticism of Method II and his assertion that
Method I was the only statistically valid method to determine the PAS overpayment. In his
report and recommendation, the hearing examiner concluded: "Dr. Wolfe, while highly
No. 20AP-32 15
qualified, offered no authority to support his opinion that Method II is 'wrong' other than
his own professional opinion." (Report & Recommendation at 6.) In our view, the record
shows the hearing examiner understood the logic underlying Dr. Wolfe's opinion but was
not convinced by his testimony.
{¶ 41} Though Dr. Wolfe also criticized Hofmann for applying an historical error
rate of 70 percent when using RAT-STATS to determine the correct sample size, he opined
Method II should never be used to project a Medicaid overpayment, regardless of the
sample size. Consequently, his criticism of Hofmann's decision to use a greater error rate
is largely immaterial. The same can be said of Dr. Wolfe's claim that the sample contained
too many zero values to be useful, as Dr. Wolfe opined he would not use Method II to
calculate an overpayment under any circumstances.
{¶ 42} Moreover, the record shows that the hearing examiner was made aware of
the conflicting evidence on this issue. PAS cross-examined Hofmann extensively about his
use of the 70 percent error rate in determining sample size, and the hearing examiner was
made aware the audited error rate of the sample turned out to be 35.1 percent. Hofmann
testified sampling additional PAS data could change the audited error rate of the sample,
and he has observed audited error rates of 100 percent in samples of other ambulette
providers "within the last two or three years." (Tr. Vol. I at 175.) Hofmann maintained the
current average error rate for ambulette service providers is "a little bit over 73 percent."
(Tr. Vol. I at 173.) Dr. Wolfe testified an historical error rate of 70 percent is "very high,"
but he did not opine Hofmann's historical data was incorrect. (Tr. Vol. II at 19.)
{¶ 43} The hearing examiner was persuaded by Hofmann's testimony that Method
II was "developed on the recommendation of Dr. Melvin Washberger who was then
chairman of the Department of Statistics at The Ohio State University" and that Method II
"is recognized by the American Institute of Certified Public Accountants, based upon a
textbook titled 'Statistical Auditing' written by Donald Roberts in 1978." (Report &
Recommendation at 6.) The hearing examiner also found that Omnicare presented a
"nearly identical situation" and that Dr. Wolfe asserted a "nearly identical criticism of
Method II" in that case. (Report & Recommendation at 6.)
{¶ 44} We acknowledge that Dr. Wolfe's challenge to the AOS methodology, if
accepted, casts doubt on the accuracy of the estimated Medicaid overpayment in this case.
No. 20AP-32 16
There is also no dispute Hofmann's use of a 70 percent error rate in determining the
required sample size, rather than the actual error rate observed in the sample, resulted in a
smaller sample size than was needed for a Method I projection.8 However, if actual error
rates were required to determine sample size, the Audit Manual would require the use of a
preliminary or pilot sample in all cases, but it does not. The evidence shows the sample size
of 214 was large enough for use in a Method II projection because all 214 audited values in
the sample could be used to project the audited value of the population. The projected
audited value of the population was then subtracted from the total actual payments for the
population to arrive at the amount of the overpayment. There is no dispute the Audit
Manual permits the calculation of a Medicaid overpayment by this method, provided the
criteria are met. The hearing examiner found the criteria were met, and the evidence
supports that finding.
{¶ 45} In an R.C. 119.12 appeal, the court of common pleas must give due deference
to the administrative resolution of evidentiary conflicts, and where the agency's decision is
supported by sufficient evidence and the law, the common pleas court lacks authority to
reverse the agency's exercise of discretion even if its decision is admittedly harsh.
Beauchamp, 2018-Ohio-1164, at ¶ 23, quoting Capital Care Network of Toledo, 2018-
Ohio-440, at ¶ 25, quoting Conrad, 63 Ohio St.2d at 111, and Henry's Cafe, 170 Ohio St. at
236-37. See also Bartchy, 2008-Ohio-4826, at ¶ 37. Absent an abuse of discretion, this
court must affirm the common pleas court's judgment, even if the appellate court would
have arrived at a different conclusion than that of the common pleas court. Edmands,
2017-Ohio-8215, at ¶ 11, citing Bartchy at ¶ 41-42.
{¶ 46} Here, the hearing examiner was presented with conflicting opinions
regarding the validity of the statistical analysis known as Method II. The underlying data
was not in dispute. The hearing examiner chose to believe Hofmann's testimony that the
methodology used by AOS to estimate the overpayment was statistically valid and that the
result was accurate. Though Dr. Wolfe's testimony was certainly relevant, and his
credentials as an expert statistician were impressive, the hearing examiner was not required
to agree with Dr. Wolfe's opinion in light of the evidence presented by ODM. The trial court
8 Hofmann admitted that a sample size of 715 would have been needed at the 35.1 percent error rate observed
in the sample of 214.
No. 20AP-32 17
found the evidence presented by ODM was reliable, and our prior decision in Omnicare
supported the hearing examiner's conclusions. Absent an abuse of discretion, our standard
of review does not permit us to reverse the judgment of the rial court simply because we
would have reached a different conclusion, as does the dissent. On this record, we do not
agree that it was unreasonable, arbitrary, or unconscionable for the trial court to defer to
the hearing examiner's determination of the relative weight and credibility of the conflicting
testimony and affirm the adjudication order. Accordingly, we hold the trial court did not
abuse its discretion when it found ODM's adjudication order was supported by reliable,
probative, and substantial evidence and was in accordance with law.
{¶ 47} For the foregoing reasons, we overrule PAS's second and sixth assignments
of error.
C. Third Assignment of Error
{¶ 48} In the third assignment of error, PAS argues the trial court erred when it
affirmed the ODM decision because AOS did not comply with the Audit Manual in
calculating the overpayment for the audit period. We disagree.
{¶ 49} PAS relies on this court's prior decision in Medcorp, Inc. v. Ohio Dept. of Job
& Family Servs., 10th Dist. No. 07AP-312, 2008-Ohio-464, as support for the third
assignment of error.9 We find Medcorp to be factually and legally distinguishable.
{¶ 50} In Medcorp, the provider pursued an R.C. 119.12 appeal to the court of
common pleas from an adjudication order of ODJFS finding an overpayment of Medicaid
reimbursement between March 1, 1996 and September 30, 1997. At the hearing before the
ODJFS hearing examiner, ODJFS presented the testimony of an expert witness, Dr.
Moeschberger, who explained that even though the method utilized by AOS to calculate the
overpayment, known as a "second methodology," was not expressly provided for in the
Audit Manual, it was authorized by implication. Id. at ¶ 19. Acknowledging it is rare to use
the "second methodology," Dr. Moeschberger explained it was also "rare to have an
instance, such as this, where all of the claims at issue in the preliminary sample are entirely
disallowed." Id.
9 Our decision in Medcorp was initially reversed on other grounds and dismissed by the Supreme Court in
Medcorp, Inc. v. Ohio Dept. of Job & Family Servs., 121 Ohio St.3d 622, 2009-Ohio-2058 ("Medcorp I").
However, in Medcorp v. Ohio Dept. of Job & Family Servs., 124 Ohio St.3d 1215, 2009-Ohio-6425, the court
reconsidered Medcorp I and affirmed the decision of this court.
No. 20AP-32 18
{¶ 51} The provider countered with an expert statistician by the name of Dr. Warren
Bilker, who opined the methodology employed by ODJFS was not provided for in the Audit
Manual and the sample of 48 was too small to be used to accurately project an overpayment
for the entire universe of 10,462 claims. The hearing examiner agreed with the provider's
expert and substantially reduced the overpayment, but ODJFS's director reinstated the full
overpayment. The provider filed an R.C. 119.12 appeal to the common pleas court.
{¶ 52} ODJFS argued in the common pleas court that Dr. Moeschberger's testimony
provided reliable, probative, and substantial evidence to support the adjudication order.
The common pleas court disagreed and overturned ODJFS's order. ODJFS appealed to this
court.
{¶ 53} This court affirmed the judgment of the common pleas court on finding that
the methodology used to determine the overpayment was not authorized by the Audit
Manual. The Medcorp decision provides in relevant part:
As found by the trial court, the manual does not suggest it
would be appropriate to apply the results of a preliminary
sample to the entire universe of claims without using an
expanded sample, and Dr. Bilker's testimony supported this
reading of the manual. Though Dr. Moeschberger testified to
the contrary, given the trial court's determination that the
manual itself refuted Dr. Moeschberger['s] testimony, this is
not, as ODJFS suggests, merely a matter of deciding which
expert opinion to follow. Rather it is the trial court reviewing
the administrative record and finding that, based on the
record as a whole, the agency order is not in accordance with
law or supported by reliable, probative and substantial
evidence.
Id. at ¶ 25.
{¶ 54} PAS argues the Medcorp decision requires this court to reverse the trial
court's decision. We disagree.
{¶ 55} Here, the evidence establishes Method II is now firmly ensconced in the AOS
Audit Manual as a statistically valid method for projecting Medicaid overpayments.
Hofmann testified the Audit Manual was revised in 2001, and it now expressly authorizes
the use of indirect projection as a method of estimating Medicaid overpayments provided
certain criteria are met. At the time of the audit in Medcorp, Method II was not expressly
authorized by the Audit Manual.
No. 20AP-32 19
{¶ 56} Though Medcorp stands for the proposition that a Medicaid overpayment
may be challenged by a provider on the basis that the statistical methodology used by AOS
in projecting the overpayment is contrary to the methodology authorized by the AOS Audit
Manual, the evidence in this case establishes that indirect projection, Method II, is
expressly authorized by the AOS Audit Manual. (State's Ex. 38.)10 Dr. Wolfe conceded this
point on direct examination when he testified "[t]he Audit Manual of the Auditor of State
allows for Method II to be used." (Tr. Vol. II at 34.)
{¶ 57} PAS next claims that AOS failed to comply with the Audit Manual by
eschewing a preliminary or "pilot" sample before selecting a sample size for the audit. (Tr.
Vol. I at 175.) Dr. Wolfe's testimony on cross-examination shows the use of a preliminary
sample is discretionary:
10 The Audit Manual provides as follows:
.52 Most MCA [Medicaid Contract Audit] statistical projections are
intended to estimate an overpayment that is repayable to the State. MCA
employs three "cascading" approaches when estimating population
overpayments (findings) from sample results. All projected findings have
the first and most common approach called the "direct projection method"
or Method I, projects the amount overpaid in the sample (the difference
between what should have been paid and what was paid) across the
population. This is generally done by calculating the average amount of
overpayment per sampling unit and then multiplying the average
overpayment by the number of sampling units in the population being
reviewed.
.53 If a 10 percent or less precision is not achieved, a second approach is
used. The second approach, called the "indirect projection method" or
Method II, projects the audited or correct payment amount per sampling
unit across the population and then subtracts the sum from the actual
amount paid. The indirect projection method will only be used if three
conditions are met as follows: 1) the sample size is sufficient to meet
Cochran's approximate test for normality; 2) the projected payment
amount from the sample is within one standard error of the actual
recorded payment amount for the population; and 3) the precision for the
point estimate of audited or correct payment amount is within 10 percent
at the 95 percent confidence level. If either the approximation of
normality requirement is not met or the precision calculated is above 10%,
the third approach, called the "lower limit method" or Method III will be
used, yielding a single lower estimate at the 95% Confidence Level. A
single limit (1-tailed) lower estimate at the 95% Confidence Level is exactly
the same as the lower limit of a 90% Confidence Level two limit (2-tailed)
estimate; which is the current audit standard mandated by CMS [Centers
for Medicare and Medicaid Services] for Medicare audits.
(Emphasis sic.) (State's Ex. 38 at 001371-72.)
No. 20AP-32 20
Q. Do all audits require the use of a preliminary sample size
determination?
A. The audit as dictated by the manual for the State of Ohio
says they should have a preliminary sample to look at to
decide on the standard deviation of the differences.
You wouldn't have to have -- I mean, it is not required in some
auditing settings.
Q. But it is here?
A. It is suggested in the manual, let's put it that way.
(Emphasis added.) (Tr. Vol. II at 11.)
{¶ 58} Because the claim that AOS did not comply with the Audit Manual is not
supported by the language of the Audit Manual or the evidence, we overrule PAS's third
assignment of error.
D. Fourth Assignment of Error
{¶ 59} In the fourth assignment of error, PAS argues the trial court erred in "failing
to recognize or distinguish years of case law invalidating Indirect Projection." (Appellant's
Amended Brief at viii.) In support of this assignment of error, PAS cites a common pleas
court decision from another appellate district, an ODM decision that was appropriately
distinguished by the hearing examiner, and a decision from Centers for Medicare and
Medicaid Services on reconsideration of a claim denial.11 As previously discussed,
Omnicare supports the trial court's determination in this case. Thus, to the extent that any
of the rulings citied by PAS are in conflict with Omnicare, they are not persuasive authority
in this district.
{¶ 60} For these reasons, we overrule PAS's fourth assignment of error.
E. Fifth Assignment of Error
{¶ 61} In the fifth assignment of error, PAS contends the administrative hearing
examiner and the trial court erred by recognizing Hofmann as an expert statistician and
permitting him to testify as such even though ODM failed to identify Hofmann as an expert
witness, as required by the relevant administrative procedures. ODM claims PAS waived
11See Ohio State Dept. of Job & Family Servs. v. Van Enterprises, Inc., Richland C.P. No. 09-CV-1425
(Apr. 30, 2012); In the Matter of Michael Matson, LPN, Docket No. 15-ODMSUR-10 (Apr. 12, 2016); and In
the Matter of Life Ambulance Serv., Inc., Medicare Appeal No. 1-77790952 (July 22, 2011).
No. 20AP-32 21
this argument by failing to raise it at the administrative hearing where the issue could have
been resolved and a record made. We agree with ODM.
{¶ 62} Ohio Adm.Code 5160-70-05(A) provides for pre-hearing discovery in R.C.
Chapter 119 proceedings as follows:
(1) Unless otherwise ordered by the hearing examiner, pre-
hearing discovery is allowed. Unless otherwise ordered by the
hearing examiner or as set forth in this rule, pre-hearing
discovery shall be conducted and used at the hearing in
accordance with the Ohio Rules of Civil Procedure.
***
(5) At a time required by the hearing examiner, a party shall
provide a report prepared by every witness identified as an
expert by the party. * * * An expert witness who is an employee
of ODM shall not be required to prepare an expert witness
report.
{¶ 63} Civ.R. 37(D) permits the exclusion of an expert witness' testimony when the
proponent fails to comply with procedural rules or orders requiring pretrial identification.
Kolidakis v. Glenn McLendon Trucking Co., 7th Dist. No. 03 MA 64, 2004-Ohio-3638,
¶ 30. We agree Hofmann's education, experience, and training arguably qualify him as an
expert statistician, and the record arguably shows he provided opinion testimony in this
matter. (Tr. Vol. I at 82, 162.) However, even if we were to conclude ODM violated Ohio
Adm.Code 5160-70-05(A), PAS never raised the issue in the administrative proceedings,
either prior to the hearing or during Hofmann's testimony.
{¶ 64} As a general rule, failure to specifically object to the testimony of an expert
witness at trial waives all but plain error on appeal. Willoughby v. Kasabwala, 11th Dist.
No. 2011-L-116, 2012-Ohio-2005, ¶ 41. The Supreme Court has stated "a party's failure to
raise an issue at the administrative level precludes the party from raising it before a
reviewing court." State ex rel. Schlegel v. Stykemain Pontiac Buick GMC, Ltd., 120 Ohio
St.3d 43, 2008-Ohio-5303, ¶ 17, citing State ex rel. Quarto Mining Co. v. Foreman, 79 Ohio
St.3d 78 (1997).
{¶ 65} We find PAS waived the argument in the fifth assignment of error by failing
to raise it at the administrative hearing. Moreover, the record shows PAS had full
knowledge of Hofmann's work in this matter from the outset of this litigation and should
have known the substance of his testimony and any opinions he held. PAS presented Dr.
No. 20AP-32 22
Wolfe's expert testimony for the specific purpose of rebutting AOS's findings and
Hofmann's expected testimony. Thus, it is not evident on this record that PAS was
prejudiced by any violation of prehearing discovery rules.
{¶ 66} For the foregoing reasons, we overrule PAS's fifth assignment of error.
V. CONCLUSION
{¶ 67} Having overruled PAS's six assignments of error, we affirm the judgment of
the Franklin County Court of Common Pleas.
Judgment affirmed.
DORRIAN, J., concurs in judgment only.
NELSON, J., dissents.
NELSON, J., dissenting.
{¶ 68} So ODM ran a sample of some 214 events billed by PAS and found errors in
76. That's an error rate of roughly 35 percent. But in deciding what sample size to use for
extrapolating what overpayments had been made on the 26,727 service events actually
provided by PAS, ODM used an error rate of 70 percent (twice that attributed to PAS), based
on an official's experience with various unidentified other providers. That inflated error rate
allowed ODM to extrapolate from a sample size of about 0.8 percent, if my math is correct,
to balloon $2,923.15 in discovered overpayments into a liability of $106,529.91.
{¶ 69} Because I don't see provision in the Audit Manual to double the PAS actual
error rate so as to permit the "Method II" calculation as based on the very small sample size
used, and because this approach seems either contrary to or at least not justified by the
methodology we sanctioned in Omnicare, 2010-Ohio-625, at ¶ 4, 34 (referencing only
"known errors" rate and upholding "projection method * * * described in elementary
statistical textbooks"), I would find that the trial court abused its discretion in finding that
the resulting sample size provided a "reliable" basis for the global assessment at issue. I
would return the matter to the trial court to be sent back to the administrative process for
further evaluation based on "dependable" evidence that can be "confidently trusted" to
supply a basis that I do not think these calculations supply. Compare supra at ¶ 19.
Therefore, and to that extent, I respectfully dissent.
{¶ 70} In a nutshell, I don't agree with the lead opinion's view that "Hofmann's
decision to use a [100 percent] greater error rate [than that shown for PAS] is largely
immaterial." Compare supra at ¶ 41. The only reason the lead opinion gives for that
No. 20AP-32 23
conclusion is that PAS's expert had further criticisms of the Hofmann approach and "opined
[that] Method II should never be used * * *." Id. But I don't see why it necessarily follows
that any sample size is fine even if not justified by the provider's own error rate: Method II
may well be appropriate in certain circumstances, but it does not appear to me to
contemplate substitution of double the provider's known error rate in assessing appropriate
sample size. And the Auditor's Manual does not contain or even advert to the purported but
here wholly unquantified average error "database" that apparently generated the
hypothesized 70 percent figure on which the sample size depended.
{¶ 71} To me, use of the conjured 70 percent error rate matters. As the lead opinion
candidly notes, "Hofmann's use of a 70 percent error rate in determining the required
sample size, rather than the actual error rate observed in the sample, resulted in a smaller
sample size than was needed for a Method I projection * * * * Hofmann admitted that a
sample size of 715 would have been needed at the 35.1 percent error rate observed in the
sample of 214." Id. at ¶ 44 and fn. 8. Here, as in Medcorp, "the manual does not suggest it
would be appropriate to apply the results of a preliminary sample to the entire universe of
claims without using an expanded sample[.]" 2008-Ohio-464, at ¶ 25. And at least without
further explanation of why use of the error rate ascribed to others, but not to PAS, was
justified, I would not stretch administrative discretion so far: the infinitely forgiving concept
of 'good enough for government work' is not at this point the prevailing standard for
administrative review.
{¶ 72} Auditor employee Hofmann conceded that there is nothing in the manual
that speaks to using the "database" for a projected error rate key to determining sample size.
Tr. Vol. I at 171 ("Probably not. I don't believe I put something in there like that"). Indeed,
he testified that auditors are not required to use the "database" at all, although others "tend
to at least look at it". Id. at 192. And his description of what the "database" is does not
inspire confidence: he says that it consists of his own compilation "[o]ver the last five or so
years" of error rates from other providers, id. at 93, "to get an idea of what is the current
average error rate for the providers that we have looked at," id. at 94, but he is "not sure how
many [ambulette provider audit figures] were in the last two years. I'm not sure if * * * any
of them were in the last two years," although one seems to have been, id. at 174-75. Compare
No. 20AP-32 24
id. at 108 (reiterating that anticipated 70 percent error rate was based on those [apparently
years' old] "prior examinations dealing with ambulette services").
{¶ 73} Yet, Hofmann's use of his "database" from outside the manual had enormous
significance here. It reduced the sample size taken by more than two-thirds, from 715 to just
214. Id. at 189. And without a proper sample size, "you can't assume normality." Id. at 98.
Asked on direct examination, "[w]hen you find out the error rate is less than what was used
to project the sample, * * * what do you do?" Hofmann replied: "Well, I first go through and
see what my projection results are, but it would indicate my point estimate for a difference
amount is probably going to have problems, since I was calculating the sample size to give
me a plus or minus 10 percent or better at – assuming 70 percent – the fact that [in actuality]
it's half of that is going to indicate that my precision at 95 percent confidence level for the
point estimate is probably going to be worse than 10 percent." Id. at 138-39. That's why he
couldn't use Method I without taking a larger sample size. Id.; see also id. at 178 ("Method
1, if you're expecting a low error rate, you're going to take a lot larger sample size, because
you have a lower error rate and to get a plus or minus 10 percent, you're going to need a
much larger sample size."); 189 (to use Method I, "would have had to increase the sample
size from 214 to 715"); 192 (Q. "So was your sample large enough here to use direct
projection?" A. "If you're talking about Method 1, it was not.").
{¶ 74} Without relying on the "database," Hofmann probably would have had to
take a "pilot" sample from PAS, and that would have led him to increase the sample size.
See id. at 175 (absent database, "if I could convince the audit staff to do it, [would] build a
pilot sample into it, so they would look at it, and say, 'Okay, what is the error rate coming
out to be, and how do we – how do we tweak the sample size as we're going along?' "). Pilot
samples are done, but take more work. Hofmann had been told not to increase sample size
in the past. Id. at 189.
{¶ 75} Hofmann's candid testimony was instructive:
[A]s a general rule, [the audit staff] don't particularly want to
do [pilot samples], which puts me into somewhat of an
unenviable spot to try to figure out, okay, what the heck is the
error rate which I use to calculate the sample size; so if I had no
idea of what the --.
Id. at 175-76; see also id. at 193 ("I had the ability that we could, if we wanted -- if we decided
it was necessary, we could expand the sample size"). But the fact that government
No. 20AP-32 25
employees "don't particularly want to" use data specific to the provider they are examining
doesn't necessarily relieve them of a need to do so or make their resulting product reliable
and dependable. And I do not think Hofmann's "unenviable spot" somehow provided the
trial court with a basis that could be "confidently trusted."
{¶ 76} To me, the government's logic is somewhat circular. Hofmann says that he
used the "estimated error rate of 70 percent to find the sample size of 214." Id. at 187. And
that sample size would permit use of Method II. Id. at 159. But my understanding is that
the only explanation given for using the 70 percent figure, and hence arriving at the 214 size,
was that the calculations would work "if a 70 percent error rate was found. 70 percent error
rate was used because audits between 2015, that was an average error rate." Id. at 103. The
record still does not divulge, I believe, why years' old figures on error rates from other
providers should have salience in determining the appropriate sample size with regard to
this (apparently or potentially less error prone) provider. And again, that "database"
approach seems nowhere to be mentioned in the audit manual. Id. at 171.
{¶ 77} I'm not sure that the confidence of reviewing courts can be enhanced by
Hofmann's acknowledgement that under Method III, "a method that's often used by
HHSOIG and Medicare audits," overpayments to PAS would have been assessed at $85,715
(rather than the roughly 25 percent higher figure imposed here). Compare id. at 141 with
supra at ¶ 10, fn.5. But it's for the reasons outlined in the paragraphs above that I
respectfully dissent.
_____________