Study design and setting
This study was a retrospective analysis of data collected according to national legislation by Austrian hospitals and compiled by the Austrian National Public Health Institute (Gesundheit Österreich GmbH, GÖG). Anonymous data were provided to the study group by GÖG following approval of an open public data request directed to https://datenplattform-covid.goeg.at/.
The data set was originally documented for purposes of quality assurance and reimbursement. Collected information include basic sociodemographic data (age, sex), administrative data (length of ICU, IMCU, and hospital stay), and outcome information (survival status at ICU and hospital discharge) in all patients admitted to Austrian hospitals. In addition, information on severity of illness at ICU admission (measured by Simplified Acute Physiology Score 3 (SAPS 3)21,22), and intensity of intensive care provided per day of ICU stay (measured by Simplified Therapeutic Intervention Scoring System (TISS-28)23) were provided in patients admitted to ICUs.
Patient and public involvement
The study was planned and conducted as a collaboration of researchers from all public medical universities in Austria. It sought to inform both the public, officials, and the scientific community about real-world implications of health care provision and effects of pandemics. It was conducted in public data thankfully provided by the Austrian National Public Health Institute and was meant to be distributed throughout the public by means of open access publication.
Ethical approval and consent to participation
The anonymous fashion of the dataset precluded the need for ethical approval; informed consent was thus neither mandatory nor possible. All used methods were carried out in accordance with relevant guidelines and regulations. The General Data Protection Regulation (GDPR) was not applicable due to the anonymous character of the dataset.
Patient population
Data on all patient admissions to Austrian acute care hospitals with ICD-10 diagnoses U04.9, U07.1, and U07.2 between January 1st, 2020 and December 31st, 2021 were included in the original dataset. For the conduction of this study, only data on the chronologically first COVID-19-related hospital admission per patient, i.e., any admission due to the aforementioned diagnoses and consecutive transfers and re-admissions, were used. Datasets with missing identifiers were excluded from analyses.
Intensive and intermediate care units
ICUs and IMCUs were identified within the original dataset according to the definition in the Austrian health structure Plan (Österreichischer Strukturplan Gesundheit, ÖSG): ICUs were units that provide care for patients who require “monitoring and restoration of vital functions, that are deranged in a life-threatening manner and need to be restored or upkept by specific intensive interventions”, IMCUs (including Respiratory Care Units (RCU) and Cardiac Care Units (CCU)) were units that allow for the “monitoring and treatment of patients, whose vital functions are at risk” and “provide the possibility of short-term (i.e., limited to a few days) intensive care (especially invasive mechanical ventilation: 48 h maximum).
Primary and secondary endpoints
The primary endpoint was in-hospital mortality. Secondary endpoints of interest were primary admission to ICU or IMCU, whichever occurred first. Parameters of interest with possible influence on these endpoints were patient age, sex, region of care, and temporal course of the pandemic (i.e., waves).
Measurements and data handling
Patient age at hospital admission was reported in categories of five years each in the original dataset to ensure anonymity. Age categories between 0 and 39 years were further condensed into two categories (0–19 years, 20–39 years, respectively) due to a low number of cases in these categories.
Information on the regional area of (primary) hospital admission was provided by assigning every dataset to one of 32 care regions prespecified in the Austrian health structure plan (Österreichischer Strukturplan Gesundheit, ÖSG). Based on this information, care regions were broadly grouped into four regions: North (Salzburg, Upper Austria), East (Vienna, Lower Austria, Burgenland), South (Styria, Carinthia), and West (Tyrol, Vorarlberg).
Information on hospital admission dates were provided in calendar weeks only in order to ensure anonymity of the dataset. To model the time course of the pandemic, half-years were defined from January 1st, 2020 to June 22nd, 2020, June 23rd, 2020 to January 1st, 2021, January 2nd, 2021 to June 21st, 2021, and June 22nd, 2021 to January 1st, 2022, respectively. Information on length of stay was available in days.
Numerical values of SAPS 3 were calculated according to the original publications21,22. Data were then categorised in quintiles to circumvent possible deficiencies in adjustment without the need for special customisation24 and to allow for the inclusion of datasets with missing SAPS 3 values.
Statistical analysis
Data description was performed using frequencies and percentages (%) or median and inter-quartile range (IQR), as appropriate.
For the endpoints of in-hospital mortality, admission to IMCU or ICU, and in-hospital mortality following ICU admission, we used a hierarchical logistic regression model with logit link function and fixed factors sex, age and half-year. Additionally, region was included as a fixed factor and care region within region as a random factor (assumed to be conditionally independent) to account for spatial correlation at different levels. Hospital mortality was analysed in the total population, in patients admitted to ICU not previously admitted to IMCU (i.e., primarily admitted to ICU), and in patients admitted to IMCU not previously admitted to an ICU (i.e., primarily admitted to IMCU). For analyses concerning in-hospital mortality following ICU admission adjusted for baseline mortality risk according to the SAPS3 score, the age category of 0–19 years was excluded because the SAPS 3 score was not developed for patients under the age of 18 years21,22. Model effects were presented as odds ratios (OR) with 95% confidence intervals (95% CI) compared to the respective reference categories.
For sensitivity analyses, an interaction term between age and sex was included into the multivariable logistic regression analysis models for all endpoints except for the ICU mortality model adjusted for SAPS 3. Further sensitivity analyses encompassed analysis for in-hospital mortality including data with missing identifiers and analyses for in-hospital mortality in all patients who were admitted to ICU or IMCU at any timepoint during their respective hospital stays.
All calculations were performed using R version 4.2.0 with packages lme4 and lmerTest.
Ethics approval
The anonymous fashion of the dataset precluded the need for ethical approval.