Introduction
Cardiovascular diseases are major global issues and have relatively high mortality
rates as compared with other diseases. Researchers, health care providers, and health
care organizations are continuously struggling to improve coronary artery disease
(CAD) prevention and care and provide better and on-time treatment to patients to
reduce its disease burden. In Germany,[1] the rate of CAD continuously declined during the last few years, from 8.8% in 2017
to a still high rate of 8.1% of the population in 2024.[2] A further decline is predicted in Germany for CAD, except for the 70 to 79 age group.[3]
[4]
[5]
For the invasive treatment of CAD, two therapeutic options are available, percutaneous
coronary intervention (PCI) and coronary artery bypass grafting (CABG).[6] Generally, PCI is primarily recommended for patients with coronary one- or two-vessel
disease, while CABG can be preferred in patients with coronary three-vessel (multi-vessel)
disease and in case of left main stenosis.[7] Both the treatments are effective especially combined with a positive change in
lifestyle. According to the results of studies focusing on quality of life and clinical
results, there is no need for an invasive treatment in asymptomatic patients. The
latest ESC 2024 and ACC/AHA 2025 guidelines provide updated recommendations for coronary
revascularization, with general agreement on PCI as first-line in one- or two-vessel
disease (synergy between PCI with Taxus and cardiac surgery [SYNTAX] of ≤22) and CABG
preferred for left main/three-vessel disease, particularly in patients with diabetes
or reduced left ventricular ejection fraction (LVEF). Key differences emerge in the
approach to complex cases: ESC 2024 maintains a more conservative stance, favoring
culprit-only PCI in STEMI and restricting left main PCI to low SYNTAX scores, while
ACC/AHA 2025 permits single-session multivessel PCI and is more permissive of left
main stenting. Both the guidelines emphasize radial access, complete revascularization,
and recommend patient informed consent based on heart team evaluations, but ACC/AHA
offers greater flexibility in PCI for complex anatomy when surgical risk is high.[8]
[9] The ESC 2023 guideline for acute coronary syndromes updates provided foundational
evidence for these current recommendations, particularly in establishing CABG superiority
for multivessel disease with diabetes.[10]
Since re-hospitalization is common for the patients having acute myocardial infarction
(MI), and almost half of the patients die within 5 years with a recurrence,[11] it is crucial to ensure high-quality care for CAD patients. The indication for invasive
treatment decision-making must adhere to clinical guidelines, considering anatomical
complexity, comorbidities, and patient preferences.[12] Individual risk of patients should be evaluated using validated scoring systems
to support informed decision-making.
However, current legal QA programs in Germany remain narrowly focused on procedures
rather than adopting a comprehensive, patient-centered, and diagnosis-driven approach.
Therefore, this study proposes a transformative shift toward a diagnosis-based QA
framework designed to optimize individualized patient treatment selection, accurately
assess mid- and long-term outcomes, and establish transparent quality benchmarks.
By integrating established guideline recommendations with real-world evidence from
extended patients' pathway analyses, this approach aims to optimize indications, standardize
patient care, bridge gaps in existing QA models, and conclusively improve patient
outcome.
Study Design
The study's mixed-methods approach combines (step 1) qualitative analyses of treatment pathways, guidelines, and expert interviews based
on (step 2) secondary data analyses of two different data sources within the framework of a
retrospective cohort study: First, data from the German mandatory external (comparative)
quality assurance, collected by the Institute for Quality Assurance and Transparency
in Health Care (IQTIG) and in which each approved health care provider performing
PCI and/or CABG has to transmit standardized data from patients with statutory health
insurance. Second, routine data from the largest German statutory health insurance,
the AOK-die Gesundheitskasse, which insures around one-third of the German population.
Data Sources for Step 1
Guidelines for CAD treatment and CAD-related quality indicators will be retrieved
by a systematic literature and quality indicator data basis analysis and expert surveys.[6]
[13]
Searching for and Consenting on Quality Indicators
The development of quality indicators (QI) for treatment of CAD follows a structural,
multi-step process involving both clinical experts and patient representatives. It
starts with a process analysis workshop to describe common diagnostic and therapeutic
pathways and patient journeys for CAD. Next, a systematic literature and guideline
review identifies existing QIs, which are compiled into a preliminary list after removing
duplicates. This is supplemented by an analysis of current QA procedures to compare
indicators, data sources, and evaluation methods. Using these findings, a draft QI
set will be created, prioritizing indication quality, outcomes, and patient experience.
Then the indicators will be prioritized through an expert Delphi consensus process,
requiring 75% agreement for inclusion. Finally, suitable data sources are identified
for implementation, and potential limitations, such as validity or statistical concerns,
and patient preferences are addressed to ensure meaningful interpretation of the QI
results. The process emphasizes evidence-based, interdisciplinary collaboration to
produce actionable, patient-centered quality measures.[14]
[15]
[16]
The results of step 1 and step 2 will be used for a modified Delphi process involving
inter- and cross-disciplinary expertise to reach a consensus on the final quality
indicator set that will form the basis of the redesigned QA system.
Data Sources for Step 2
AOK: The AOK Research Institute (WIdO) provides access to its database on all reimbursement-related
data of the AOK insurees. The database generally contains demographic variables (age,
sex, region of residence, insurance status), “International Statistical Classification
of Diseases and Related Health Problems” (ICD) and “operation and procedure codes”
(OPS) of inpatient (hospitalization/rehospitalization) and outpatient visits, and
medications (ATC codes).
IQTIG: The Institute for Quality Assurance and Transparency in Health Care (IQTIG) operates
the external QA in Germany. In relation to CAD the IQTIG runs two different QA schemes:
one for coronary angiography and PCI and another for isolated CABG. Health care providers
are obliged to submit pseudonymized clinical and in-hospital outcome data for each
patient with statutory health insurance (for CABG Germany-wide to IQTIG, for PCI at
first to Federal State–related QA institutions and then to IQTIG). The IQTIG analyses,
among others, measured values of predefined quality indicators and other basic information.
The results are published annually. Further analyses that are not included in the
publicly available reports can be performed upon request as part of so-called secondary
data use (according to § 137a, paragraph 10, sentence 4, Social Code Book V).
Patient Selection
On the one hand, our study includes all AOK-insured patients with CAD who underwent
an invasive coronary procedure (isolated/consecutive PCI/CABG) in the years 2019 to
2023. On the other hand, those cases are included for which the IQTIG had proper QA
documentation from the PCI or CABG service areas for the years 2021 to 2023.
Inclusion and Exclusion Criteria
AOK data: Key elements of inclusion criteria are: age ≥18 years with a diagnosis of
CAD (ICD-10-GM Version 2019–23: I20-I25, I42-I43, I46 and I50), who underwent an invasive
coronary procedure (isolated/consecutive PCI/CABG) during the years 2019 to 2023 (procedure
codes: OPS 5–360*, 5–361*, 5–362*, 5–363*, 5–364*, 8–837*; EBM 34291, 34292), and
were insured at AOK at least 1 year before and after treatment or until death.
IQTIG data: All cases that were registered within the IQTIG service area “KCHK-KC”
or “PCI” for the years 2021 to 2023.
Patients with missing key data such as diagnosis code or treatment record are excluded.
Study Variables
Several distinct variables associated with CAD are considered for our study. This
will capitalize on a comprehensive set of variables, extracted from AOK database and
IQTIG quality reports on demographic, clinical, treatment, and outcome measures. Age
(≥18), sex/gender, and region (urban versus rural based on postal code/AOK exclusive)
are generally considered as demographics.
Clinical data include cardiac and non-cardiac comorbidities such as hypertension (ICD-10
code I10-I15), diabetes mellitus (E10-E14), acute or chronic kidney disease (N17,
N18), heart failure (I50), prior myocardial infarction (I21-I22), and other heart,
vascular, or neurological diseases. Additionally, the disease categorization and associated
information related to CAD from IQTIG/OPS codes, e.g., multi-vessel disease, left
ventricular ejection fraction, and ACS type (STEMI (I21.0-I21.3) versus NSTEMI (I21.4)
versus unstable angina (I20.0) are recorded.
Treatment variables encompass revascularization types, PCI (OPS 8–837*; EBM 34291 + 34292)
or CABG (OPS 5–36*), and pharmacotherapy adherence (Proportion of Days Covered, PDC)
≥80% for antiplatelets (ATC code B01AC), statins (ATC code C10AA), β-blockers (ATC
code C07AB), and ACE inhibitors (ATC code C09AA)/ARBs (ATC code C09CA), date of hospitalization,
date of treatment, and transfer between hospitals.[17]
[18]
In our study, survival, 30-day and 1-year all-cause mortality of CAD patients will
be considered as primary outcomes.
Major adverse cardiovascular and cerebrovascular events (MACCE,[19] e.g., AMI, stroke, and cardiovascular death), cardiovascular related rehospitalization
within 12 months, and a deterioration in long-term care dependence will be considered
as secondary outcomes.
Data Cleaning
Incomplete, irrelevant, and duplicate data will be removed. Considering our large
dataset, we will not use imputation, as it can be misleading or can be considered
as data manipulation. Besides that, normalization will also be performed to organize
the attributes of database and to overcome data redundancy.
Secondary Data Analysis
Our data analysis is based on two purposes in connection with IQTIG and AOK patient's
secondary data.
Regarding IQTIG secondary data, PCI and CABG cases from the PCI and KCHK-KC QA schemes
for the years 2021 to 2023 will be compared regarding the indication and procedure-related
data and—if available—the 1-year follow-up outcomes. For this purpose, the most comparable
elective cases will be selected using propensity score matching (variables taken into
account: isolated/consecutive cases, age, gender, and other recorded risk factors),
and the achieved treatment/indicator results will be compared using inferential statistics.[20] These analyses demonstrate the extent to which comparable cases have different short-term
and 1-year outcomes depending on the invasive procedure type and how often the indication
criteria were met, e.g., depending on gender.
Based on secondary AOK data of CAD patients who underwent an invasive coronary therapy
(isolated/consecutive PCI/CABG) between 2019 and 2023, their treatment courses and
short- and long-term outcomes (between 2014 and 2023) will be analyzed retrospectively.
The aim of the analyses is to identify typical treatment courses (patient journeys,
e.g., various, possibly consecutive, multiple procedures) and to evaluate the procedure
related as well as the potential determinants of these outcomes.
The isolation of typical treatment procedures and their related outcomes can reveal
the most relevant criteria for determining the indication for PCI or CABG and also
identify strata of patients for whom a meaningful comparison of outcome indicators
is possible. This enables fair comparison of service provision related to diagnosis-based
rather than procedure-based QA, which can lead to long-term improvements in the quality
of care.
After an initial review of the various treatment courses, propensity score matching
will be used to create comparable groups (age and gender, pre-existing conditions,
and degree of severity/indications, for example, acute myocardial infarction and/or
three-vessel disease). The associated outcomes (survival/recurrence) will be compared
using inferential statistics (survival analysis; logistic regression) depending on
the last selected intervention (PCI/CABG). Based on the findings on gender-specific
differences in the conservative and invasive treatments of CAD, the analyses are also
performed on a gender-specific basis.[21]
[22]
Statistical Methods
By evaluating the outcomes of comparable patients who have undergone either PCI or
CABG,[23] the aim is to determine whether the choice of a particularly suitable invasive procedure
can achieve outcome benefits for selected patient groups.
Once data are properly cleaned, normality test can be performed to check either our
data are normally distributed or skewed. If all the assumptions of normal distribution
are fulfilled, parametric tests will be used for statistical analysis, otherwise non-parametric
tests are recommended. Not normal or skewed data can be transformed using log transformation
and then we can apply the parametric test to the transformed data and interpret the
results after anti-transformation. We will check the power using actual sample size,
under the assumptions of two-sided tail and 0.05 level of significance.[24]
Descriptive analysis will summarize baseline characteristics. Generally, continuous
variables will be summarized using means and standard deviations and will be compared
between groups using the unpaired student's t-test. Categorical variables measured on nominal scale will be analyzed using the
Fisher's exact test, while ordinal categorical variables will be assessed using non-parametric
methods, including the Wilcoxon rank sum test or Mann-Whitney U test, as appropriate.
Data visualization will be used for temporal trends and distributions.
Discrete outcomes will be reported as counts and percentages. Comparative group analyses
will involve calculating relative risks along with 95% confidence intervals.
Event-free survival will be estimated using the Kaplan-Meier method, differences between
groups will be assessed using the log rank test, and cox regression analyses will
determine the influence of different variables on survival time. Multivariable logistic
regression models will be developed to identify the independent predictors of the
primary outcomes—event-free survival/all-cause mortality at 1 year. The models will
include baseline clinical and angiographic characteristics, as well as procedural
variables such as the revascularization strategy (CABG versus PCI).[24]
[25]
[26]
Strengths and Limitations
Overall, we have a relatively big and comprehensive dataset based on nationwide IQTIG
procedural data and CAD patients registered with AOK.
Due to robust up to 5-year follow-up (2019–2023), long-term insights will enable to
capture the critical endpoints, which is somehow not possible in short-term quality
metrics (QM).
In some cases, patients switch their insurance from AOK to another statutory health
or private health insurance company. In these cases, follow-up will be impossible,
and it may cause attrition bias, especially if registration cancellation relates to
unmeasured health status. If we try to cover this attrition bias as a censoring event
by using competing risk regression, still there can be traces of residual bias for
long-term outcomes.
As we use secondary data, some treatment indications may be challenging to verify,
if the necessary clinical data are missing, since it was not collected as part of
our study. In addition, the interpretation of results may be challenging due to the
possibility of coding errors or incomplete information when using partially unverified
data.
In terms of generalizability, although the AOK covers almost 33% of the German population,
this health insurance only covers those with statutory health insurance, so our conclusions
and decisions regarding this data obviously cannot be applied one-to-one to non-statutory
health insured patients (approximately 11% of the German population).
Finally, observational studies[27] are always confronted with the problem that unmeasured confounding factors such
as patient preferences or surgical skills influence the observed results and causal
interpretations are not possible.
Expected Outcomes and Impact
Step 1 of the study will provide a set of quality indicators as basis of a new combined
QA scheme for CAD patients who have undergone PCI or CABG.
In step 2 of the study, we will assess the association between treatment courses and
clinical outcomes of PCI and CABG patients by analyzing real-world data, thereby providing
evidence-based insights into how the procedures influence patients' health. Based
on these findings, the study will generate targeted policy recommendations to redesign
QA for CAD patients undergoing isolated or consecutive CABG and/or PCI.
Ultimately, it would be possible to evaluate the indication quality and outcome quality
of CAD patients undergoing PCI and/or CABG in a more detailed and comparative manner.
Software
PostgreSQL (Postgres) or other more efficient databases, e.g., DuckDB, will be used
for data management. Regarding data analyses, preferably we will use Python for data
analyses and Draw.io to draw flow diagram/charts and other relationship model as per
requirement. Multiple python libraries such as Pandas, Numpy, Matplotlib, Lifeline,
and Plotly will be used as per settings and specifications.[28] Other statistical software like SPSS, R, and Stata can also be helpful during our
data analyses process.
Ethical and Legal Aspects
Person-identifying data are removed by the 11 regional AOKs through encryption before
the data are delivered to the WIdO. The WIdO carries out a second pseudonymization,
so that the subsequent routine data analyses are only performed with double-pseudonymized
and thus anonymized data. This ensures that no social security data pursuant to Section
67 of the German Social Code Book X (SGB X) are used in this research project. Only
aggregated data will be passed on to external consortium partners in compliance with
the rules. The processing of research datasets within the WIdO is safe from a data
protection perspective.
Ethical and scientific standards will be adhered. In particular, the general principles
of scientific work are considered. The data registration, acquisition, and evaluation
are according to the guidelines and recommendations for good epidemiological and clinical
practice and Helsinki declaration for human subject and ethical principles in medical
research are particularly relevant to the project's research questions. Secondary
data will be evaluated using appropriate methods. Standardized and established procedures
to ensure data protection will be used in the planning and implementation of secondary
data analyses. The data will be analyzed on-site at the WIdO by staff from the IGVE,
Marburg.
Regarding the expert survey data and consensus conferences, the purpose of the data
collection is transparently explained to the participants in the consent processes,
and their consent to participate is obtained in advance.
Further, this study has been approved by the Ethics Committee of the Faculty of Medicine
at Marburg University, Germany (no. 25–193-BO).
