Relationships Between Central Quality Assurance Criteria for the

Assessment of Statutory Health Insurance Patients for Further

Development Using Medical Informatics Methods

Mareike Burmester

1,5

, Paul-Ulrich Menz

, Vera Ries

, Klaus-Peter Thiele

, Bernhard van Treeck

and Reinhard Schuster

1,5

Medical Advisory Board of Statutory Health Insurance in Northern Germany (MD Nord), 23554 L

ubeck, Germany

Medical Advisory Board of Statutory Health Insurance in Westphalia-Lippe (MD Westfalen-Lippe), 48153 M

unster,

Germany

Medical Advisory Service Institution of the Statutory Health Insurance in North Rhine (MD Nordrhein), 40212 D

usseldorf,

Germany

Federal Joint Committee (G-BA), 10587 Berlin, Germany

Institute of Mathematics, University of L

ubeck, 23562 L

ubeck, Germany

Keywords:

Medical Quality Assurance, Statutory Health Insurance in Germany, Linear Optimization, Correlations and

Partial Correlations.

Abstract:

In recent years, a comprehensive system has been developed both in terms of content and IT for quality

assurance in the assessment for the granting of beneﬁts for persons with statutory health insurance by the

Medical Advisory Boards, which has also been enshrined in German legislation. In addition to not insigniﬁcant

formal criteria and criteria relating to speciﬁc assessment areas, four criteria relevant to the entire assessment

spectrum are evaluated in detail. One- and two-dimensional criteria provide an overview as an introduction to

the topic. Similar to the procedure in image processing, linear optimization methods are used to infer relevant

intervals of the detailed parameters from row and column totals. Using correlations and partial correlations, the

relationship between the central quality criteria is shown. Methods of spherical trigonometry are generalized.

For each of the three sides of the quadrilateral of the four central criteria, it is of central importance that the

partial correlations are greater or smaller than the correlations overall. This is determined by the modulus

value, which in the application under consideration produces the same results on all sides of the tetrahedron

under consideration.

1 INTRODUCTION

The Medical Advisory Board of Statutory Health In-

surance (SHI) in Germany is responsible for conduct-

ing assessments of SHI-insured individuals in order

to ascertain their medical and socio-medical require-

ments for beneﬁt decisions by the statutory health

and long-term care insurance funds. Additionally, the

Medical Advisory Board provides advice to the afore-

mentioned insurance funds on a case-by-case basis.

The assessments of the Medical Advisory Board out-

side of long-term care insurance in the area of health

insurance are predominantly carried out by doctors.

They are impartial in their assessment and are only

bound by their conscience; however, they are required

to adhere to the legal framework and current medical

knowledge to a signiﬁcant extent, see (Gemeinsamer

Bundesausschuss, 2022). The relationship between

law and medicine is a key topic in the ﬁeld of so-

cial medicine. The abstract legal claims of insured

persons are concretised by multi-layered committees,

resulting in binding national regulations. Medical

ﬁndings, legislation and individual case decisions by

courts in various instances resulting from insufﬁcient

concretisation set precedents for subsequent similar

decisions, which may be cited as such in future in-

stances. These developments and the further develop-

ment of regulations by the committees as a framework

for decision-making are naturally delayed to a lesser

or greater extent. This has direct consequences for

the ongoing necessary further training of all those in-

volved as experts in this interdisciplinary context, see

712

Burmester, M., Menz, P.-U., Ries, V., Thiele, K.-P., van Treeck, B. and Schuster, R.

Relationships Between Central Quality Assurance Criteria for the Assessment of Statutory Health Insurance Patients for Further Development Using Medical Informatics Methods.

DOI: 10.5220/0013255600003911

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 18th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2025) - Volume 2: HEALTHINF, pages 712-719

ISBN: 978-989-758-731-3; ISSN: 2184-4305

(Altenstetter and Busse, 2005).

In 2016, Medical Advisory Board initiated a joint

evaluation of existing quality assurance initiatives and

the development of a comprehensive, standardised

quality assurance process for all areas of assessment.

This process was developed with the involvement of

all regional Medical Advisory Boards and specialist

competence units, which set the standards for assess-

ment in each specialist area. The Medical Advisory

Boards have trialled the new national quality assur-

ance plan in three test runs throughout Germany, pi-

loted it in the largest assessment area and then gradu-

ally implemented it in practice in all assessment areas.

Finally, the national quality assurance plan was even

taken up by the legislature, incorporated into the Ger-

man Social Code and, following a broad national dis-

cussion among all stakeholders and interested parties,

adopted in a guideline conﬁrmed by the Federal Min-

istry of Health with sub-legal binding force. Quality

assurance follows the generally recognised principles

of the quality assurance cycle for continuous eval-

uation and further development, see (Gemeinsamer

Bundesausschuss, 2023).

The assessment is carried out by the Medical Ad-

visory Boards, which are essentially organised by fed-

eral state. Accordingly, an initial quality check is car-

ried out in the regionally responsible Medical Advi-

sory Board. For quality assurance purposes, the as-

sessments are randomly drawn on a nationwide basis.

The special feature of the newly implemented system

is that a partial sample is sent to another Medical Ad-

visory Boards for a second independent assessment.

If there are noticeable differences between the self-

assessment and external assessment, a further qual-

ity check is carried out by a third Medical Advisory

Board.

The entire process of drawing up and providing

expert opinions, along with all the stages of the audit

described above, is conducted via a central server on

which all the expert opinions speciﬁed for the audit

are accessible in anonymised form, see (Medizinis-

cher Dienst Bund, 2023), (Ries et al., 2020) and (Ries

et al., 2023).

Should discrepancies remain following the third

stage of the examinations, a discussion and consen-

sus are reached on the speciﬁc case construction at

nationwide consensus conferences. After consensus

conferences were held for the ﬁrst time in 2018 as

part of the feasibility study, they are now established

as a nationwide quality assurance instrument. They

promote mutual learning and take place twice a year

in each of the nine assessment areas, once in person

and once online via video conference. Representa-

tives of all regional Medical Advisory Boards take

part in each consensus conference, as well as at least

one representative of the assigned specialist compe-

tence unit. This makes it possible to promptly im-

plement the need for further development in the regu-

lations identiﬁed in the consensus conferences, as the

competence units set the standards for the assessment.

A total of 20 criteria have been established for the

evaluation of the assessment, which are standardised

across all assessment areas. Furthermore, additional

technical criteria are currently being harmonised with

the general criteria. In addition to the aforementioned

formal criteria, there are four criteria of central im-

portance for standardised assessment processes in the

regional Medical Advisory Boards, cf. (Gaertner and

van Essen, 2024), (Gaertner and Gnatzy, 2011).

Three of these central criteria are operationalised

as three test aspects for the accuracy of the expert

opinion, namely criteria 14, 15 and 16. Triggered by

limitations on the transfer of the case ﬁle due to data

protection regulations that vary in strictness from re-

gion to region, the national quality assurance plan has

created the new requirement for expert opinions with

the fourth central criterion 9 that they must be con-

vincing with a concise presentation of the character-

istics of the individual case that are relevant to the

assessment.

2 MATERIALS AND METHODS

The following analysis focuses on the medical ﬁelds

(cf. (Anja Dippmann, 2024)):

• 100: Incapacity of work

• 200: Hospital care

• 400: New and unconventional examination and

treatment methods / pharmacy

• 500: Prevention and rehabilitation

• 700: Medical supplies

In the year under review 2023, 16942 internal and

4953 external evaluations were carried out in these

medical ﬁelds. In addition, only the central criteria

• Criterion 9: The report contains the information

necessary to assess the facts of the case

• Criterion 14: The expert opinion takes into ac-

count current medical knowledge

• Criterion 15: The expert opinion takes into ac-

count the socio-medical requirements

• Criterion 16: The result of the expert opinion is

plausible and understandable in the context of the

facts presented

Relationships Between Central Quality Assurance Criteria for the Assessment of Statutory Health Insurance Patients for Further

Development Using Medical Informatics Methods

713

taken into account.

The results of the assessments of the expert opin-

ions are presented on a three-level scale (green, yel-

low, red). The colour red means that the criterion to be

assessed has not been met. Green means that the cri-

terion has been fully met, while yellow indicates that

there is room for improvement (cf. (Gerlach, 2001)).

The ﬁrst step is to compare the internal and ex-

ternal ratings, differentiated by cause group. The fre-

quency of occurrence of each scale value is analysed.

This is followed by a speciﬁcation where the distribu-

tion of criteria 14 and 16 is analysed separately.

The aim of consistent quality assurance across all

assessment regions is to achieve the same assessments

both within and between regions. The ﬁrst step is

to compare the internal and overall ratings. This re-

sults in a total of nine different combinations of rat-

ings with the three possible scale values (green/green,

green/yellow, ... red/yellow, red/red). This can also

be visualised as a nine-box table, in the same way as

a four-box table. The marginal sums can be calculated

from the available values. In the present constellation,

a total of six proportional values can be identiﬁed,

with the sum of the vertical and horizontal marginal

totals having the same value. This situation is typi-

cal of medical image processing, where the object is

to be inferred from certain sums per viewing direction

(see (Zeng, 2010), (Jan, 2005)). Compared to medical

imaging, the dimension here is small, but it is suitable

for demonstrating basic aspects.

The extent to which the actual combination de-

viates from the possible extreme values is of inter-

est for standardised quality assurance. In the follow-

ing, we will determine the interval in which the val-

ues in the nine-ﬁeld table lie if the boundary values

are speciﬁed. A linear optimisation problem is solved

for this purpose. It is assumed that the modelling is

carried out using linear optimisation and that mono-

tonicity is given. In addition, the proportions of the

joint assessment are assumed to be between 0% and

100%. The target value is deﬁned as the interval in

which either the largest valuation differences or iden-

tical valuations occur. If there are no values of com-

mon valuations in the practical application, but only

the marginal totals, intervals for deviations can thus

be determined. A separate analysis is also carried out

for the individual criteria.

The comparison of internal and external assess-

ments is not the only application that enables this ap-

proach. It is also possible to analyse the interrelation-

ships between the criteria. In the following, medical

ﬁeld 100 for incapacity for work is analysed, whereby

the assessments of criteria 14 (medical content) and

16 (comprehensibility of the results) are compared.

Furthermore, an analysis is carried out for medical

ﬁeld 200 (hospital care).

Finally, the correlations, the partial correlations

and the modulus value are analysed. For this purpose,

the random variables X

, X

are deﬁned as the

four central criteria of quality assurance. The correla-

tion ρ

i j

= ρ

describes the relationship between two

variables X

and X

, where i, j ∈ {1, 2, 3, 4}, i ̸= j ap-

plies. The partial correlations are calculated accord-

ing to the following formula:

i j,k

i j

− ρ

1 − ρ

where i, j, k ∈ {1, 2, 3, 4} are pairwise different. The

modulus value

m =

1 − ρ

i j,k

1 − ρ

i j

is always identical for all three combinations of partial

correlation and the corresponding correlation. If the

modulus value m > 1, the following applies

m > 1

⇔

1−ρ

i j,k

1−ρ

i j

> 1

⇔ 1 − ρ

i j,k

> 1 − ρ

i j

⇔ ρ

i j

> ρ

i j,k

⇔ |ρ

i j

| > |ρ

i j,k

The calculations described above are performed sepa-

rately for the internal and overarching assessment. In

addition, the internal valuation is analysed separately

between the medical ﬁelds considered.

Furthermore, calculations in linear optimization

with Mathematica from Wolfram Research are used.

3 RESULTS

Table 1 shows the differences between the internal

and the overarching assessment of the expert opinions

depending on the reason group for the four central cri-

teria 9, 14, 15 and 16. Only those expert opinions that

were subject to both an internal and an overarching

assessment were taken into account.

Both the red and amber assessments show a more

critical tendency in the overarching assessments. This

may be due to the fact that the approach to the ap-

praisal varies from region to region and that the struc-

ture of the appraisals is different and therefore less

well known. Both are reasons for standardisation.

When assessing the medical requirements in ac-

cordance with criterion 14, the best results are shown

HEALTHINF 2025 - 18th International Conference on Health Informatics

714

Table 1: The results of the internal and external assessment

of all Medical Advisory Boards in relation to criteria 9 and

14-16 for the year 2023 are presented herewith. The values

are expressed as a percentage (%).

Internal External

Medical

green yellow red green yellow red

Field

100 84.5 11.5 4.0 77.9 16.8 5.3

200 95.3 3.7 1.0 93.5 4.4 2.2

400 89.8 8.3 1.9 85.3 10.0 4.7

500 82.9 12.3 4.8 82.6 12.6 4.8

700 93.7 4.6 1.7 88.6 6.8 4.6

Table 2: The results of the internal and external assessment

of all Medical Advisory Boards in relation to criteria 14 for

the year 2023. The values are expressed as a percentage

(%).

Internal External

Medical

green yellow red green yellow red

Field

100 97.4 1.8 0.8 95.6 3.8 0.6

200 99.4 0.5 0.1 99.1 0.7 0.2

400 89.6 9.0 1.4 82.8 13.4 3.8

500 95.8 3.1 1.0 93.9 4.0 2.1

700 97.5 1.5 1.0 95.1 2.5 2.4

Table 3: The results of the internal and external assessment

of all Medical Advisory Boards in relation to criteria 16 for

the year 2023. The values are expressed as a percentage

(%).

Internal External

Medical

green yellow red green yellow red

Field

100 75.0 19.5 5.4 68.3 26.8 5.0

200 91.6 6.8 1.6 88.6 8.4 3.0

400 92.3 5.4 2.4 85.7 9.1 5.2

500 78.1 17.3 4.6 78.7 16.4 4.9

700 88.7 9.4 1.9 82.3 11.5 6.2

even more clearly in Table 2 in medical ﬁeld 200

‘Hospital care’.

The plausibility of the results in criterion 16 in Ta-

ble 3 is assessed more critically than in the previously

analysed criterion 14. Here too, the results for hospi-

tal care show the best results.

Equal internal and overarching assessments are

desirable for standardised quality assurance across all

regions of the assessment. In this context, the ques-

tion arises as to what extent the actual combination of

criteria deviates from possible extreme values.

Linear optimisation as a processing method ﬁrst

requires the meaningful formulation of inequalities

and target values, which are then compared with re-

gard to their results. It seems sensible to maximise the

number of equal evaluations. Alternatively, it would

be possible to minimise the greatest differences. In a

Table 4: The percentage of identical or dissenting votes

for the identical appraisals in the internal and external ap-

praisals.

Internal evaluation

green yellow red

Sum

external

External

evaluation

green 80.0 5.0 1.5 86.5

yellow 7.2 1.6 0.5 9.3

red 3.0 0.7 0.5 4.2

Sum

90.2 7.3 2.5 100

internal

Table 5: Results of the optimisation problem separately for

the four main criteria. In the left-hand column are the actual

values for the red/green deviation case, with the correspond-

ing calculated intervals in which the value could lie with

the speciﬁed limits. On the right, these values for the case

where green/green, yellow/yellow and red/red were chosen.

Crit.

Maximum Difference Consensus

Real Interval Real Interval

9 5.9 [2.1 , 10.1] 77.2 [69.1 , 86.6]

14 2.2 [0.5 , 2.4] 91.9 [90.8 , 96.5]

15 4.8 [1.4 , 6.4] 82.2 [78.3 , 91.4]

16 4.8 [1.5 , 7.5] 77.0 [68.4 , 86.4]

combination of both considerations, it would ﬁrst be

necessary to determine which weights would lead to

a sensible solution.

Another option would be to analyse the equally

problematic ratings. The assumption that a com-

bined analysis would not result in equally poorly as-

sessed appraisals is not based on empirical experi-

ence. Therefore, a minimum level can be assumed

for other optimisation criteria. Alternatively, the real

value with the minimum and maximum possible val-

ues without further target values can be considered for

empirical determination.

With regard to the differentiation between internal

and overarching audits with regard to the four deﬁned

criteria, the following can be stated:

With the exception of a small deviation, the values

in each row and column are monotonically decreas-

ing. For the linear optimisation problem with the fur-

ther restriction that there are expert opinions that are

rated red (criterion not fulﬁlled) both internally and

across the board, an example of 0.4% with 0.5% in

reality is assumed. This means that in reality there

is complete agreement (green/green, yellow/yellow,

red/red) for 82.1% of the expert opinions. Assum-

ing that the minimum of common criteria assessed in

red is not restricted in the expert opinion, the interval

ranges from 76.6% to 90.2%. With the restriction, the

minimum increases slightly to 77.8%.

For the largest valuation differences, which are

present in 4.5% of cases in real terms, the interval

Relationships Between Central Quality Assurance Criteria for the Assessment of Statutory Health Insurance Patients for Further

Development Using Medical Informatics Methods

715

Table 6: Describes the percentages of how often green, yel-

low and red were selected in criteria 14 and 16 in medical

group 100, and how often which combination of values oc-

curred.

Criteria 16

green yellow red Sum

Criteria

green 73.6 19.5 4.3 97.4

yellow 0.4 1.2 0.4 2.0

red 0.1 0.1 0.4 0.6

Sum 74.1 20.8 5.1 100

Table 7: Describes the percentages of how often green, yel-

low and red were selected in criteria 14 and 16 in medical

group 200, and how often which combination of values oc-

curred.

Criteria 16

green yellow red Sum

Criteria

green 90.6 6.8 1.8 99.2

yellow 0.1 0.4 0.1 0.6

red 0.0 0.0 0.2 0.2

Sum 90.7 7.2 2.1 100

ranges from 1.4% to 6.6% without the additional re-

striction. With the additional restriction, the upper

limit is reduced to 5.4%.

With regard to the individual criteria, the follow-

ing result is obtained for the variant without additional

restrictions (Table 5):

The comparison of the ratings between criteria 14

and 16 in the medical ﬁeld 100 results in the following

nine-ﬁeld table (Table 6).

When looking at criteria 14 and 16, it is initially

noticeable that there is a greater asymmetry in the

evaluation matrix. In addition, the monotony in the

rows and columns is more limited in the medical ﬁeld

200 than in the medical ﬁeld 100. This is related to the

dependency between the criteria, which is considered

next with the correlations.

In the medical ﬁeld 100, the interval from 4.4% to

5.7% (i.e. outside the real value) lies at 4.3% max-

imum valuation differences with monotonicity as-

sumption and without monotonicity assumption wider

between 2.5% to 5.7%. The complete match is 75.3%

in real terms with monotonicity assumption we un-

derestimate the interval from 71.5% to 74.1%. With-

out the monotonicity assumption, on the other hand,

a wider interval of 71.5% to 76.7% is achieved.

In the medical ﬁeld 200, higher levels of agree-

ment can again be observed (Table 7).

The largest valuation differences in real terms

are 1.8%, with the intervals [2.4%, 6.9%] exhibiting

monotonicity and [1.4%, 2.4%] lacking a monotonic-

ity assumption. In real terms, there is complete agree-

ment at 91.1% with the intervals [89.8%, 90.6%] with

monotonicity and [89.8%, 91.4%] without a mono-

tonicity assumption.

Finally, the measures of correlation, partial corre-

lation and modular value of the four central criteria

are analysed in more detail.

Table 8: Describes the correlation between two of the four

central criteria. All evaluated reports were analysed inter-

nally for this purpose. The matrix is symmetrical due to the

symmetry property of correlation.

Second Criteria

9 14 15 16

First

Criteria

9 0.295 0.393 0.601

14 0.295 0.371 0.348

15 0.393 0.371 0.496

16 0.601 0.348 0.496

Table 9: The correlations between criteria 9, 14, 15, 16

for all externally assessed reports, regardless of the medi-

cal ﬁelds.

Second Criteria

9 14 15 16

First

Criteria

9 0.290 0.444 0.623

14 0.290 0.458 0.433

15 0.444 0.458 0.458

16 0.623 0.433 0.458

Table 10: Describes the difference between internal and ex-

ternal ratings and the correlation between the four central

criteria.

External

9 14 15 16

Internal

9 0.252 0.345 0.529

14 0.252 0.392 0.370

15 0.345 0.392 0.490

16 0.529 0.370 0.490

In order to obtain an indication of how strongly

the four central criteria correlate in the internal assess-

ment of the expert opinions in the individual cause

groups, the cause groups are analysed separately once

again.

Table 11: Correlation of the criteria based on internal re-

ports of the medical ﬁeld 100. In this case, n = 1576 expert

opinions were analysed.

Second Criteria

9 14 15 16

First

Criteria

9 0.211 0.499 0.567

14 0.211 0.203 0.261

15 0.499 0.203 0.548

16 0.567 0.261 0.548

The highest correlation between the four criteria

analysed can be seen in the expert opinions on medic-

inal products and on new and unconventional treat-

ment methods (medical ﬁeld 400), see Table 13, cf.

(Schuster, 2022). The highest number of expert opin-

HEALTHINF 2025 - 18th International Conference on Health Informatics

716

Table 12: Correlation of the criteria based on internal re-

ports of the medical ﬁeld 200. Here n = 8040 expert opin-

ions were analysed.

Second Criteria

9 14 15 16

First

Criteria

9 0.227 0.137 0.600

14 0.227 0.238 0.292

15 0.137 0.238 0.354

16 0.600 0.292 0.354

Table 13: Correlation of the criteria based on internal re-

ports of the medical ﬁeld 400. Here, n = 1404 expert opin-

ions were the subject of analysis.

Second Criteria

9 14 15 16

First

Criteria

9 0.513 0.525 0.568

14 0.513 0.507 0.512

15 0.525 0.507 0.542

16 0.568 0.512 0.542

Table 14: Correlation of the criteria based on internal re-

ports of the medical ﬁeld 500. Here, n = 1375 reports were

analysed.

Second Criteria

9 14 15 16

First

Criteria

9 0.366 0.487 0.575

14 0.366 0.399 0.386

15 0.487 0.399 0.595

16 0.575 0.386 0.595

Table 15: Correlation of the criteria based on internal re-

ports of the medical ﬁeld 700. In case of medical ﬁeld 700,

n = 1380 expert opinions were analysed.

Second Criteria

9 14 15 16

First

Criteria

9 0.334 0.377 0.639

14 0.334 0.490 0.405

15 0.377 0.490 0.458

16 0.639 0.405 0.458

ions on hospital care is found in medical ﬁeld 200

in Table 12 and thus also in the total number of ex-

pert opinions (Table 8). It is also worth noting that

the differences between internal (Table 8) and over-

all assessment (Table 9 and, moreover, the differences

between these assessments vary only moderately. If

internal and overall assessment were essentially the

same, the difference in table 10 would not differ in

this way. The smallest differences between the med-

ical assessments (criterion 14) and the socio-medical

assessments in the individual medical ﬁeld analysed

are for medicines (400) (Table 13) and medical aids

(700) (Table 15).

The partial correlations between three of the four

variables considered in quality assurance are analysed

below. This can be viewed as three sides of a polyhe-

dron. The modulus value described in the methods

section indicates whether the correlations are larger

or smaller than the partial correlations. As described

in the methods section, this then applies to all permu-

tations of the three variables. For the interpretation

of the tetrahedron, it can be deduced that the modu-

lus values are greater than 1 on all three sides, which

means that the partial correlations are smaller than the

correlations in terms of amount.

In a comparison between the internal and overar-

ching tests, there are only moderate differences in the

correlations, partial correlations and the module value

for all combinations of criteria under consideration:

Table 16: The analysis of the correlation between the ﬁrst

and second of the four criteria shows that if the correlation

via the third criterion is excluded, a value of the partial cor-

relation close to zero can be assumed, which means that

the two criteria considered can be regarded as independent.

Since all values of the modulus value are greater than 1, all

correlations are greater than the associated partial correla-

tions.

Criteria int./ Partial Correlation Modul

1 2 3 ext. ρ

12,3

13,2

23,1

value

9 14 15 int 0.175 0.290 0.319 1.062

9 14 15 ext 0.135 0.336 0.276 1.048

9 14 16 int 0.115 0.223 0.556 1.081

9 14 16 ext 0.071 0.288 0.485 1.062

9 15 16 int 0.136 0.354 0.509 1.160

9 15 16 ext 0.115 0.386 0.440 1.120

14 15 16 int 0.244 0.421 0.203 1.091

14 15 16 ext 0.261 0.403 0.222 1.102

A comparison of the criteria considered in the in-

ternal evaluation results in the following comparison

for the individual cause groups:

According to the explanations in the methods sec-

tion, all partial correlations are smaller in amount than

the correlations. This says nothing about the sign, i.e.

the change from positive to negative correlation or the

reverse. In the present analysis, a change of sign oc-

curs only once, namely in the medical ﬁeld 200 (hos-

pital care) in the correlation without the criterion 14

(medical content).

4 CONCLUSIONS

The four quality criteria under examination interact

intensively, as evidenced by the fact that the partial

correlations are consistently smaller than the correla-

tions. It is frequently posited that the high correlation

observed between two variables is a consequence of

their dependence on a third variable. In this instance,

all three variables are of signiﬁcant content relevance

when the triangular conﬁguration of statistical corre-

Relationships Between Central Quality Assurance Criteria for the Assessment of Statutory Health Insurance Patients for Further

Development Using Medical Informatics Methods

717

Table 17: Analysis of the partial correlations considered in

table 16 and the module value separately for the individual

medical ﬁelds.

Criteria Med Partial Correlation Modul

1 2 3 ﬁeld ρ

12,3

13,2

23,1

value

9 14 15 100 0.129 0.115 0.477 1.029

9 14 15 200 0.202 0.214 0.087 1.011

9 14 15 400 0.336 0.326 0.358 1.204

9 14 15 500 0.215 0.272 0.400 1.102

9 14 15 700 0.185 0.417 0.260 1.087

9 14 16 100 0.079 0.175 0.542 1.040

9 14 16 200 0.068 0.200 0.573 1.050

9 14 16 400 0.314 0.313 0.415 1.223

9 14 16 500 0.191 0.230 0.505 1.113

9 14 16 700 0.107 0.265 0.584 1.113

9 15 16 100 0.274 0.371 0.405 1.232

9 15 16 200 -0.101 0.343 0.596 1.009

9 15 16 400 0.314 0.348 0.397 1.224

9 15 16 500 0.221 0.441 0.406 1.247

9 15 16 700 0.123 0.305 0.566 1.148

14 15 16 100 0.074 0.523 0.183 1.037

14 15 16 200 0.150 0.306 0.229 1.036

14 15 16 400 0.318 0.381 0.328 1.210

14 15 16 500 0.229 0.521 0.201 1.127

14 15 16 700 0.374 0.326 0.233 1.131

lations is taken into account. This remains the case

when three edges of each of the four variables are

considered as tetrahedrons. It is important to note

that correlations and partial correlations are statistical

variables that may not be able to explain a causal rela-

tionship. In the context under consideration, there are

numerous substantive discussions between physicians

with the participation of computer scientists in the

consensus conferences described in the introduction.

The extensive details of the discussion provided a ba-

sis for statistical examination. The result was that the

substantive arguments put forward by doctors in the

review of expert opinions were largely conﬁrmed sta-

tistically. This alternating suggestion between practi-

tioners and statisticians will contribute to the further

development of quality assurance in the Medical Ad-

visory Boards with high performance, cf. (N

uchtern

et al., 2015), (Newhouse, 2002).

A further point of discussion in the further devel-

opment of quality assurance in medical assessments

is the relationship between cost and beneﬁt. A re-

duction in the number of criteria to be assessed will

result in a reduction in the effort required. On the one

hand, the expert groups are engaged in a discussion

regarding the identiﬁcation of criteria that are simi-

lar in content and, therefore, amenable to a summary.

This is particularly pertinent to criteria that are em-

ployed solely within individual event groups. The ra-

tionale for these criteria was based on the assumption

that they could not be evaluated using the general cri-

teria. However, the testing experience from the dis-

cussions in the consensus conferences has shown that

this is certainly possible. This has resulted in a mod-

iﬁcation of the test criteria. This paper does not show

to what extent the audit has led to a signiﬁcant im-

provement in the assessment. While a modiﬁcation

and summarisation of the review criteria may appear

to result in a deterioration, it has, in fact, led to an

overall improvement.

In addition to the discussion of the proximity of

test criteria in terms of content, the question also

arises as to which criteria are viewed in the same way

by the physicians in each case without content-related

proximity. This concerns an association of judgement

errors. This can also motivate a reduction in the num-

ber of endpoints.

This paper focussed on the fact that there are crite-

ria whose further reduction would not make sense and

would be clearly counterproductive for quality assur-

ance overall.

REFERENCES

Altenstetter, C. and Busse, R. (2005). Health care reform in

germany: patchwork change within established gov-

ernance structures. Journal of health politics, policy

and law, 30(1-2):121–142.

Anja Dippmann, et al.. (2024). Qualtit

atssicherung der Be-

gutachtung f

ur die Gesetztliche Krankenversicherung

QSKV - Jahresbericht 2024 - Datenjahr 2023. Home-

page MD Bund.

Gaertner, T. and Gnatzy, W. (2011). Zum

sachverst

andigenstatus im medizinischen dienst

der krankenversicherung am beispiel des mdk hessen.

GuP, 5:166–173.

Gaertner, T. and van Essen, J. (2024). Der Medi-

zinische Dienst - eine sozialmedizinische Institution

der Qualit

atssicherung im Gesundheitssystem. Das

Gesundheitswesen.

Gemeinsamer Bundesausschuss (2022). Beschluss

des Gemeinsamen Bundesausschusses

uber die

Ver

offentlichung des Berichts des Medizinischen

Dienstes Bund gem

aß §16 Teil A MD-QK-RL:

Bericht

uber die im Jahr 2021 durchgef

uhrten

Qualit

atskontrollen. Homepage Gemeinsamer

Bundessausschuss.

Gemeinsamer Bundesausschuss (2023). Richtlinie des

Gemeinsamen Bundesausschusses nach §137 Absatz

3 SGB V zu Kontrolle des Medizinischen Dienstes

nach §275a SGB V. Homepage Gemeinsamer Bun-

dessausschuss.

Gerlach, F. M. (2001). Qualit

atsf

orderung in Praxis und

Klinik: eine Chance f

ur die Medizin; 24 Tabellen.

Thieme.

Jan, J. (2005). Medical image processing, reconstruction

and restoration: concepts and methods. Crc press.

Medizinischer Dienst Bund (2023). Regelm

aßige Berichter-

stattung der Medizinischen Dienste

uber ihre T

atigkeit

HEALTHINF 2025 - 18th International Conference on Health Informatics

718

und Personalausstattung. Richtlinie des Medizinischen

Dienstes Bund nach §283 Absatz 2 Satz 1 Nr. 8SGB V.

Newhouse, J. P. (2002). Why is there a quality chasm?

Health affairs, 21(4):13–25.

uchtern, E., Bahemann, A., Egdmann, W., van Essen, J.,

Gostomzyk, J., Hemmrich, K., Manegold, B., M

uller,

B., Robra, B., R

oder, M., et al. (2015). Soziale

Sicherheit braucht Sozialmedizin Selbstverst

andnis

von

arztinnen und

arzten in der sozialmedizinischen

Begutachtung und Beratung. Das Gesundheitswesen,

77(08/09):580–585.

Ries, V., Thiele, K.-P., Schuster, M., and Schuster, R.

(2020). It-structures and algorithms for quality assur-

ance in the health insurance medical advisory service

institutions in germany. In HEALTHINF, pages 353–

360.

Ries, V., Thiele, K.-P., van Treeck, B., Schroeer, S., Witt,

C., and Schuster, R. (2023). It-structures and algo-

rithms for quality assurance in the medical advisory

service institutions in germany. step 2: To err is hu-

man. consensus-conferences. In HEALTHINF, pages

271–278.

Schuster, R. (2022). Strukturelle Beratung der

Krankenkassen durch den Medizinischen Dienst

in der Pharmakotherapie. Die Verwendung epidemi-

ologischer und gesundheits

okonomischer Analysen

im Zeitalter von Digitalisierung und Bid Data. Agor,

K and Knieps, F and Hartweg, HR.

Zeng, G. L. (2010). Medical image reconstruction, volume

530. Springer.

Relationships Between Central Quality Assurance Criteria for the Assessment of Statutory Health Insurance Patients for Further

Development Using Medical Informatics Methods

719