Hospital Procedure Volume Affects Outcomes of Emergency Transcatheter Aortic Valve Implantations in Germany
Since 2005, data on all hospitalizations in Germany have been available for scientific purposes via the Diagnosis-Related Group (DRG) statistics collected by the Research Data Center of the Federal Statistical Office (DESTATIS). This hospitalization data, including diagnoses and procedures, is a valuable source of nationally representative data on the treatment of patients in the hospital. This database represents an almost complete collection of all hospitalizations in German hospitals that are reimbursed according to the diagnosis-related group system. From this database, we extracted data on all isolated transfemoral expandable or self-expandable TAVI procedures performed in 2018. As previously described, patients with an initial diagnosis of pure aortic regurgitation (primary or secondary diagnosis other than I35. 0, I35.2, I06.0, I06.2) and those who had concurrent cardiac surgery or percutaneous coronary intervention were not included in this analysis2. A complete list of procedure codes can be found in a previous article17. In addition, patients admitted urgently were extracted. Emergency admission was a predefined code which was made available for scientific analysis by DESTATIS after transmission by the hospitals.
The analysis focused on eight different endpoints: in-hospital mortality, bleeding events, stroke, acute kidney injury, postoperative delirium, permanent pacemaker implantation, mechanical ventilation greater than 48 h, length of hospital stay, and repayment. Stroke and acute kidney injury were defined using the ICD, tenth revision (ICD-10) codes (secondary diagnosis I63* or I64 and N17*, respectively). Bleeding was defined as requiring a transfusion of >5 units of red blood cells and identified using the German classification of operations and procedures (OPS) codes (8–800.c1 to 8–800.cr). In-hospital mortality, duration of mechanical ventilation and length of hospital stay were among the main set of DESTATIS variables. For all other comorbidities, existing distinctive anamnestic or acute codes were used (we have discussed OPS and ICD codes in detail previously2).
For the calculation of the estimated logistic EuroSCORE (European Cardiac Operative Risk Assessment System), we were able to fill in all the fields except the critical preoperative state and the left ventricular function. In these, we assumed a discrete state (i.e. no critical preoperative state and no left ventricular dysfunction) and therefore calculated a best case scenario.
In a previous study, Reinöhl et al.2 identified 20 baseline patient characteristics to describe risk profiles between procedural groups. As the patients were not randomized, potential confounding factors were taken into account by the propensity score method. In detail, the propensity score was used for the adjustment. The propensity score was estimated using a multivariate linear regression model, with the number of emergency TAVI cases as the dependent variable and all baseline characteristics listed in Table 1 as the independent variables. Then, a propensity score adjustment was applied with the number of emergency TAVI cases and the propensity score as continuous covariates. Thus, logistic or linear regression models were used as appropriate. To account for the correlation of error terms of patients treated at the same hospital, a random intercept was included at the center level. Based on these eight risk-adjusted logistic or linear regression analyses, the predicted probabilities (or means) were calculated using marginal standardization and visualized in certain volume categories18.
No imputation for missing values could be performed due to missing codes indicating missing data. If the patient’s electronic health record did not include information about a clinical feature, it was assumed that this feature was not present. Additionally, no adjustment for multiple testing was made. Thus, p-values may not be interpreted as confirmatory but are descriptive in nature and inferences drawn from 95% confidence intervals may not be reproducible.
All analyzes were performed with Stata 16 (StataCorp, College Station, Texas, USA).
Ethical approval and informed consent
Our study did not imply direct access by the investigators to individual patient data but only access to the summary results provided by DESTATIS. Therefore, approval by an ethics committee and informed consent were deemed not required, according to German law. All summary results have been anonymized by DESTATIS. In practice, this means that any information allowing conclusions to be drawn about a single patient or a specific hospital has been redacted by DESTATIS to ensure data protection. Furthermore, in order to avoid the possibility of drawing conclusions on a single hospital, the data is checked and censored by situation by DESTATIS in these cases. All methods were performed in accordance with current guidelines and regulations.