key: cord-0958681-f1kf9ohb
authors: Rainer, Lukas; Bachner, Florian; Eglau, Karin; Ostermann, Herwig; Siebert, Uwe; Zuba, Martin
title: Comorbidities and COVID-19 hospitalization, ICU admission and hospital mortality in Austria: A retrospective cohort study
date: 2022-05-24
journal: Wien Klin Wochenschr
DOI: 10.1007/s00508-022-02036-9
sha: 00c5377b10e5ad50a4e462d0d3142ba51fbb0ee8
doc_id: 958681
cord_uid: f1kf9ohb

BACKGROUND: The protection of vulnerable populations is a central task in managing the Coronavirus disease 2019 (COVID-19) pandemic to avoid severe courses of COVID-19 and the risk of healthcare system capacity being exceeded. To identify factors of vulnerability in Austria, we assessed the impact of comorbidities on COVID-19 hospitalization, intensive care unit (ICU) admission, and hospital mortality. METHODS: A retrospective cohort study was performed including all patients with COVID-19 in the period February 2020 to December 2021 who had a previous inpatient stay in the period 2015–2019 in Austria. All patients with COVID-19 were matched to population controls on age, sex, and healthcare region. Multiple logistic regression was used to estimate adjusted odds ratios (OR) of included factors with 95% confidence intervals (CI). RESULTS: Hemiplegia or paraplegia constitutes the highest risk factor for hospitalization (OR 1.61, 95% CI 1.44–1.79), followed by COPD (OR 1.48, 95% CI 1.43–1.53) and diabetes without complications (OR 1.41, 95% CI 1.37–1.46). The highest risk factors for ICU admission are renal diseases (OR 1.76, 95% CI 1.61–1.92), diabetes without complications (OR 1.57, 95% CI 1.46–1.69) and COPD (OR 1.53, 95% CI 1.41–1.66). Hemiplegia or paraplegia, renal disease and COPD constitute the highest risk factors for hospital mortality, with ORs of 1.5. Diabetes without complications constitutes a significantly higher risk factor for women with respect to all three endpoints. CONCLUSION: We contribute to the literature by identifying sex-specific risk factors. In general, our results are consistent with the literature, particularly regarding diabetes as a risk factor for severe courses of COVID-19. Due to the observational nature of our data, caution is warranted regarding causal interpretation. Our results contribute to the protection of vulnerable populations and may be used for targeting further pharmaceutical interventions.

Results Hemiplegia or paraplegia constitutes the highest risk factor for hospitalization (OR 1.61, 95% CI 1.44-1.79), followed by COPD (OR 1.48, 95% CI 1.43-1.53) and diabetes without complications (OR 1.41, 95% CI 1.37-1.46). The highest risk factors for ICU admission are renal diseases (OR 1.76, 95% CI 1.61-1.92), diabetes without complications (OR 1.57, 95% CI 1.46-1.69) and COPD (OR 1.53, 95% CI 1.41-1.66). Hemiplegia or paraplegia, renal disease and COPD constitute the highest risk factors for hospital mortality, with ORs of 1.5. Diabetes without complications constitutes a significantly higher risk factor for women with respect to all three endpoints. Conclusion We contribute to the literature by identifying sex-specific risk factors. In general, our results are consistent with the literature, particularly regarding diabetes as a risk factor for severe courses of COVID-19. Due to the observational nature of our data, caution is warranted regarding causal interpretation. Our results contribute to the protection of vulnerable populations and may be used for targeting further pharmaceutical interventions.

Keywords Epidemiology · Risk factors · Underlying conditions · Sex differences · Charlson comorbidity index Background At the end of February 2020 the first cases of the ongoing severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic were reported in Austria associated with one of the first European superspreading events. After a first peak in case numbers and hospitalizations in March and April 2020, the highest case numbers and hospitalizations have been ob-original article served during a second peak in November with 706 intensive care unit (ICU) beds occupied by patients with Coronavirus disease 2019 (COVID- 19) accounting for 34% of the total number of ICU beds for adults in Austria. In a third peak in April 2021 and a fourth peak in December 2021, ICU occupancy exceeded 600 patients with COVID-19. Due to regional variation in COVID-19 incidence and ICU capacities, COVID-19 associated occupancy rates reached almost 70% in some regions leading to substantial pressure on the Austrian healthcare system, characterized by high ICU bed density compared to other countries [1] .

To protect vulnerable populations, numerous protective measures were implemented including special working conditions or duty leave. The Austrian COVID-19 risk group regulation defines the high-risk group based on medical conditions, such as chronic lung diseases, chronic heart diseases or cancer drawing from available evidence by May 2020.

Recent literature identifies similar conditions as risk factors for severe courses of COVID-19, although ranking and magnitude differs across available studies. Based on a national surveillance study in Ireland, Bennett [2] . In a cohort study capturing the first wave of the pandemic in Scotland, McGurnaghan et al. identified diabetes as a risk factor for fatal or critical care unit-treated COVID-19 with an OR of 1.40 (95% CI 1.30-1.49) [3] . A surveillance study in Spain showed significant associations between chronic renal disease (OR 1.47, 95% CI 1.29-1.68) and diabetes (OR 1.23, 95% CI 1.14-1.33) and mortality among chronic conditions in hospital cases [4] . Ahlström [7] .

However, only few surveillance studies representative of the general population have been reported and observational studies on COVID-19 disease risk and severity may suffer from selection bias (collider bias) when samples rely on hospitalization with COVID-19 or voluntary participation [8, 9] . In addition, study results on the impact of major chronic conditions such as cancer on COVID-19 outcomes remain heterogeneous and inconclusive.

Furthermore, hardly any cohort studies on sex differences of comorbidities as risk factors for COVID-19 outcomes exists. In a literature review Kautzky-Willer analyzed sex differences with respect to risk factors for severe courses of COVID-19 [10] . Kautzky-Willer hypothesized that diabetes mellitus constitutes a higher risk factor for women compared to men due to biological factors; however, systematic empirical literature is missing for evaluating this hypothesis. In order to fill these gaps, we estimate the effect of comorbidities on COVID-19 hospitalization, ICU admission, and mortality, based on a matched cohort study using nationwide hospital billing data from Austria.

We performed a retrospective cohort study including all patients with COVID-19 in the period February 2020 to December 2021 who had an inpatient stay during the period 2015-2019 in Austria (N = 46,740). For each COVID-19 patient and outcome, we randomly drew five controls of the same age group, sex and healthcare region from the general inpatient population 2015-2019 without COVID-19 hospitalization (N = 3,558,072). We employed exact matching as an enrichment design in order to have a control group balanced at baseline. This is particularly important for young age-groups, which would be underrepresented in the control group without matching. Including healthcare region as a matching variable allows to control for the spatial variation in COVID-19 incidence and associated pressure on health system capacities across Austria. The choice of an appropriate baseline and follow-up period involves a trade-off between comprehensiveness and attrition. Choosing the 5-year period before the spread of COVID-19 in Austria (2015-2019) allows for measuring major baseline comorbidities, which is in line with the Swedish cohort study of Ahlström et al. [5] . In order to reduce attrition bias, we only included patients with a contact in the inpatient, outpatient or ambulatory care setting in 2019. Additionally, we excluded all patients discharged dead during 2019. However, we are not able to control for attrition with respect to other places of death due to lack of linked data, which is a caveat for analyzing severe comorbidities such as metastatic solid tumors. Due to the limited availability of data on comorbidities for other healthcare settings in Aus-Comorbidities and COVID-19 hospitalization, ICU admission and hospital mortality in Austria K original article tria, we focused on the hospital inpatient sector as our study population. This approach limits generalizability, as the Austrian inpatient population accounts for only 40% of the general population. Aiming at identifying vulnerable population groups, we consider this limitation as acceptable as we can assume that most of the vulnerable population has an inpatient stay within a 5-year period.

Aiming at estimating causal effects using observational data, we used the target trial framework to organize our study design around our causal question for the example of the effect of the "exposure" diabetes on the risk of COVID-19 ICU admission [11, 12] . Ideally, we would like to measure the causal effect of diabetes on COVID-19 associated ICU admission based on a randomized control trial, where patients in the treatment group are randomly assigned to the exposure of diabetes and the per protocol effect is calculated as the relative cumulative risk of an ICU admission with COVID-19 as principal or additional diagnosis among individuals assigned to each treatment strategy (see Table 7 ).

The main data source for this cohort study are hospital billing data related to the Austrian Diagnosis Related Groups(DRG)-like system (Leistungsorientierte Krankenhausfinanzierung) administrated by the Federal Ministry of Social Affairs, Health, Care and Consumer Protection. Besides demographic information on sex and age group, these data include patient level information on principal and additional diagnoses (based on the International Statistical Classification of Diseases and Related Health Problems 10 th Revision, ICD10). Data for the year 2021 are preliminary and may be subject to revision. COVID-19 is considered both as main and additional diagnosis because patients with COVID-19 as additional diagnosis often have main diagnoses associated with COVID-19 from a clinical perspective. Patients without valid patient IDs (i.e., 9% of all acute care admissions in 2019) had to be excluded as they cannot be matched over time. Moreover, patients who were not discharged by 31 December 2021 are not included.

For measuring comorbidities at baseline we consider all principal and additional diagnosis in the 5-year pre-COVID-19 period (2015-2019) and use categories of medical conditions according to the Charlson comorbidity index (CCI) [13] following the strategy of the comorbidity analysis of Cho et al. [7] . The advantage of clustering diagnosis to patient collectives according to the CCI is to obtain an easily comparable number of 19 comorbidities. We used the R package comorbidity [14] for mapping principal and secondary ICD-10 diagnosis codes with categories of medical conditions and computing CCI. For instance, type 1 diabetes mellitus without complications (ICD-10 E10.9) is categorized as diabetes without complication in the CCI. We additionally stratify for sex to account for potential sex differences in the impact of comorbidities on COVID-19 outcomes. While including sex as an interaction effect would increase statistical power due to larger sample size, we decided for stratification because of simpler interpretation of the adjusted odds ratios in the logistic regression model.

As robustness check, we use more granulated diagnoses categories based on the following aggregation strategy. As a starting point we aggregated the ICD-10 diagnoses at category level (e.g., E66: obesity). In order to avoid very low numbers of diagnoses we further aggregated diagnosis below a cut-off value of 0.5% to groups (e.g., E70-E90 metabolic disorders). In case of low frequencies of groups (below a cut-off of 1%) we further aggregated them to chapters (e.g., E00-E90: endocrine, nutritional and metabolic diseases) in order to analyze frequent categories (with a frequency above 0.5% of all documented diagnoses) individually while controlling for less frequent diagnosis as potential confounders. We considered chapter VII (diseases of the eye and adnexa) and VIII (diseases of the ear and mastoid process) as irrelevant conditions from a medical perspective and excluded them from the analysis. The aggregation strategy led to a list of 67 comorbidities.

For the primary analysis, we employ multivariable logistic regression to estimate adjusted odds ratios (OR) with 95 confidence intervals (95% CI) of comorbidities on the outcomes COVID-19 hospitalization, ICU admission, and hospital mortality using the 'glm' function of the 'stats' package and the 'summ' function of the 'jtools' package in R [15] .

Additionally, we calculate unadjusted risk ratios for each comorbidity without controlling for other comorbidities based on age and sex-standardized incidence rates of patients with a COVID-19 ICU admission compared to the control group without a COVID-19 hospitalization. We compute p-values based on χ 2tests, where the family-wise error rate is controlled for with the Holm method to reduce the likelihood of type 1 errors (false positives). Unless otherwise stated, we choose an alpha-level of 0.05 for statistical significance.

In the period February 2020 to December 2021, 1.26 million COVID-19 cases were identified in Austria. This represents a prevalence of identified cases of 14.2% of the total population. Thereof 13,360 died (case fatality rate, CFR: 1.06%), 68,183 were hospitalized (5.41%), and 11,793 were admitted to ICUs (0.94%) (see Table 1 ). Men show higher risks with respect to all three endpoints. This is particularly the case for ICU admissions, where rates for men (1.21%) are 83% higher compared to women (0.66%). Out of the 13,360 COVID-19 associated deaths 11,683 (87%) were registered in hospitals. Authors' calculation based on [16, 17] Pop refers to total population, k refers to 1000, and proportions refer to detected cases. Patients without valid patient IDs and patients who were not discharged by 31 December 2021 are not included ICU intensive care unit

Comorbidities and COVID-19 hospitalization, ICU admission and hospital mortality in Austria K original article Authors' calculation based on [16, 17] Pop refers to total population, k refers to 1000, and proportions refer to detected cases. Patients without valid patient-IDs and patients who were not discharged by 31 Dec 2021 are not included ICU intensive care unit Table 2 shows summary statistics for the study population (N = 3,604,812) representing 40.4% of the Austrian population, 68.6% of COVID-19 hospitalizations, 66.5% of ICU admissions and 80.3% of COVID-19 patients discharged dead. A proportion of 1.30% of the study population had a COVID-19 associated hospitalization in the study period, which is 69% higher compared to the total population (0.766%).

The following results refer to the effects of comorbidities on COVID-19 hospitalization, ICU admission and hospital mortality, based on logistic regression analysis, expressed as adjusted ORs and 95% CIs. For our primary analysis we use CCI categories for clustering comorbidities. Our results are robust with respect to other diagnosis clustering strategies (see Table 8 for results at category, group and chapter levels).

For the endpoint of COVID-19 hospitalization, hemiplegia or paraplegia (OR 1.61, 95% CI 1.44-1.79), COPD (OR 1.48, 95% CI 1.43-1.53) and diabetes without complications (OR 1.41, 95% CI 1.37-1.46) constitute the comorbidities with the highest risk (see Table 3 and Fig. 1a) . We observe significant sex differences for diabetes without complications, which is a higher risk factor for women. In general, the results are robust with respect to different clustering of diagnosis. Considering diagnoses at category level obesity (OR 1.35, 95% CI 1.29-1.40) and type 2 diabetes mellitus (OR 1.34, 1.30-1.38) constitute similar risk factors as summarized under the CCI category of diabetes with and without complications (see Table 8 ). Table 8 ).

For the endpoint of COVID-19, hospital mortality Hemiplegia or paraplegia (OR 1.51, 95% CI 1.21-1.89), renal disease (OR 1.51, 95% CI 1.42-1.61) and COPD (OR 1.51, 95% CI 1.41-1.61) constitutes the highest risk factors. As for the other endpoints we observe significant sex differences for diabetes mellitus, which is a higher risk factor for women (see Table 5 and Fig. 1c) . Table 6 shows age-standardized and sex-standardized frequencies of comorbidities for the control group and the COVID-19 group admitted to ICU ordered by relative risk (RR) not adjusted for other comorbidities. Hemiplegia or paraplegia represents Authors' calculation based on [16, 17] *, **, *** refer to significance at the p < 0.10, p < 0.05, and p < 0.01 level, respectively. Results refer to odds ratios (OR) and 95% confidence intervals (95% CI) obtained from logistic regression; analyses were adjusted for age group, sex and health care region

Adjusted effect sizes (OR) with 95% CIs of risk factors for COVID-19 hospitalisation, ICU admission and hospital mortality (authors' calculation based on [16, 17] Authors' calculation based on [16, 17] *, **, *** refer to significance at the p < 0.10, p < 0.05, and p < 0.01 level, respectively. Results refer to odds ratios (OR) and 95% confidence intervals (95% CI) obtained from logistic regression; analyses were adjusted for age group, sex and health care region Authors' calculation based on [16, 17] *, **, *** refer to significance at the p < 0.10, p < 0.05, and p < 0.01 level, respectively. Results refer to odds ratios (OR) and 95% confidence intervals (95% CI) obtained from logistic regression; analyses were adjusted for age group, sex and health care region the comorbidity with the highest RR (5.23) as 2.6% of COVID ICU patients had an inpatient stay with this condition compared to 0.5% of the control group. Diabetes with complications (RR 3.22) and renal disease (RR 3.00) constitute the further most relevant risk factors. The most frequent comorbidity of COVID ICU patients was diabetes without complications (13.2%). The higher burden of comorbidities is also visible referring to the CCI. While 5.3% of COVID ICU patients had a CCI larger or equal to 5, this was only the case for 2.4% of the control group. 60% of COVID ICU [16, 17] *, **, *** refer to significance at the p < 0.10, p < 0.05, and p < 0.01 level, respectively. Age-standardized and sex-standardized shares and risk ratios based on the Austrian population 2020; p-values refer to χ2-tests, where the family-wise error rate is controlled for with the Holm method CCI Charlson comorbidity index, RR relative risk, ICU intensive care unit patients had a CCI of zero, compared to 79% of the control group.

We performed a retrospective cohort study aiming at estimating the effect of comorbidities on COVID-19 hospitalization, ICU admission, and mortality, using nationwide hospital billing data from Austria. Our analysis revealed several comorbidities associated with an elevated risk of severe courses of COVID-19. Hemiplegia or paraplegia, COPD and diabetes without complications constitute the highest risk factor for COVID-19 hospitalization. The highest risk factors for ICU admission are renal disease, diabetes without complications and COPD. For the endpoint of COVID-19, hospital mortality hemiplegia or paraplegia, renal disease and COPD constitute the highest risk factors. The point estimates of the adjusted ORs range from 1.4 to 1.8. Diabetes without complications is a significantly higher risk factor for women, particularly with respect to ICU admission. We contribute to the literature on risk factors for severe courses of COVID-19 by analyzing sex differences. This is, to the best of our knowledge, one of the first large-scale cohort studies providing sex-stratified results. Our results on diabetes as a higher risk factor for women supports the hypothesis of Kautzky-Willer, that women with type 2 diabetes seem to lose their biological female advantage with respect to severe courses of COVID-19 [10] . Our results are partly in line with other published cohort studies. Our results for diabetes as a risk factor for hospitalization and ICU admission are very similar to the adjusted results from Rawshani [7] . In contrast to [2, 7] we hardly find any association of cancer and COVID-19 endpoints. We only find an association of cancer and COVID-19 hospitalization and mortality for women with a low effect size. We do not find an association of cancer and COVID-19 ICU admission, which is in line with the results of the Swedish cohort study conducted by Ahlström et al. [5] . Differences in the results might be partly explained by different periods where comorbidities are considered. By focusing on the 5-year period before COVID (2015-2019), we aim at capturing the most relevant period in terms of predicting severe courses of COVID-19.

Comorbidities and COVID-19 hospitalization, ICU admission and hospital mortality in Austria K original article

As other observational studies our cohort study has several limitations. Firstly, causal inference is limited due to the observational nature of our data. Analyzing the effect of underlying risk factors such as diabetes rather than interventions exceeds the concept of treatment strategies and assignment procedures because such risk factors cannot be assigned in practice. Thus, our feasible strategy is to observe comorbidities from hospital billing data at baseline Since the causal effect of observed comorbidities may be confounded by unobserved factors such as socioeconomic status, our results should be interpreted with caution, particularly with respect to a causal interpretation of our comorbidity factors. For instance, the results for obesity or diabetes mellitus may be confounded by socioeconomic status. The results for dementia, and hemiplegia or paraplegia may be associated with living in long-term care institutions, which have been particularly exposed to COVID-19 clusters in Austria during the second wave in autumn 2020. By the end of December 2021, 30% of all COVID-19 associated deaths were reported for residents of long-term care facilities [18] ; however, this effect does not distort the causal link of the conditions and COVID-19 outcomes as the conditions typically cause the admission to long-term care institutions.

Secondly, the comorbidity analysis relies on the coding quality in Austrian publicly funded hospitals. Since the data are primarily collected for accounting purposes within the Austrian DRG system, issues such as upcoding or incomplete diagnoses coding with respect to additional diagnoses exists. For instance, the distinction between diabetes with and without complications in diagnosis coding may get neglected in clinical practice due to time pressure. Due to the lack of linked data, we cannot control for the vaccination status of patients with COVID-19 which hinders the analysis of risk factors with respect to different vaccination status.

Thirdly, the analysis only contains patients with an inpatient stay in the period 2015-2019 due to limited data availability of comorbidities in other settings in Austria. Diagnoses coding in the ambulatory care sector is not yet fully implemented in Austria and information on medications is only available with substantial time lag. Thus, the cohort study is not representative for the general population as the inpatient population is characterized by higher burden of disease compared to the general population. The higher severity level of the inpatient population is, for instance, reflected by an 69% higher incidence rate of COVID-19 hospitalization compared to the general population. Aiming at identifying vulnerable population groups, we consider this limitation as acceptable as we can assume that most of the vulnerable population has an inpatient stay within a 5-year period.

Fourthly, individual behavior such as risk aversion may impact upon the results as patients are likely to choose their COVID risk behavior based on their med-ical record. Thus, patients with diagnoses listed on top of the Austrian COVID risk group regulation, such as chronic pulmonary disease, chronic heart diseases or cancer may have behaved more risk averse compared to patients with obesity or diabetes mellitus ranked number 7 and 8 on the list containing 9 groups of diagnosis in total, which may have partially led to a selfdefeating prophecy. Of note, this would not be confounding but rather it means that the biological effect of such a comorbidity is superimposed by reactive protective behavior that is also by the comorbidity. Isolating the biological effect would require a mediation analysis, which was not the goal of this study [19] .

Fifthly, we do not analyze how comorbidities affect each other over time, which may lead to time-varying confounding. Controlling for this phenomenon would require more sophisticated statistical models such as Robins' generalized methods (g methods) [20] .

Sixthly, analyzing hospital mortality, we consider endpoints at time of discharge rather than 30-day mortality due to significant time lag in data availability of linked data.

Seventhly, some comorbidities such as metastatic solid tumurs may be so severe that they fall under palliative care, reducing the likelihood of hospitalization and mortality in a hospital. Therefore, the estimated effect size of such comorbidities may be true but does not reflect the overall mortality risk of such patients.

Our results may be used for sharpening the risk group definition, which is essential for the protection of vulnerable populations by pharmaceutical and nonpharmaceutical interventions. In particular our study may contribute to raise awareness of large population groups such as diabetics by communicating the risk of severe courses of Coronavirus disease 2019 (COVID-19) and thus communicating the benefits of vaccination or new antiviral therapies.

Further research should include the joint influence of socioeconomic status and comorbidities because major risk factors such as diabetes are likely confounded by socioeconomic status. Furthermore, the effect of the immunization status should be included in order to analyze the risk of comorbidities with respect to different immunization status. This could support policy makers in the implementation of more target group-oriented pharmaceutical interventions such as the further vaccination program or the use of antiviral therapies. original article study design, data collection and analysis, decision to publish, or preparation of the manuscript.

The study was designed by all authors. LR wrote the first draft of the manuscript. LR and MZ performed all statistical analyses. All of the authors participated in analysis and interpretation. All authors vouch for the accuracy and completeness of the data and analyses and made the decision to submit the manuscript for publication. All authors meet the International Committee of Medical Journal Editors criteria for authorship for this article, take responsibility for Table 7 Summary of the protocol of a target trial to estimate the effect of diabetes without complications on COVID-19 ICU admission

At least one inpatient stay in the baseline period 2015-2019 in a public hospital in Austria and no presence or history of diabetes

Refrain from assigning diabetes without complications. Assigning diabetes without complications in the baseline and during the follow-up

Random assignment of participants to either strategy at baseline. Participants will be aware of the strategy to which they have been assigned

Follow-up period Starts in January 2020 before the onset of COVID-19 in Austria and ends at COVID-19 associated ICU admission, death, loss to follow-up, or 24 months after baseline

Outcome ICU admission with polymerase chain reaction (PCR) confirmed COVID-19 as principal or additional diagnosis

Per-protocol effect estimated via comparison of 16-month risk of COVID-19 associated ICU admission among individuals assigned to each treatment strategy

Authors' illustration based on [12] 

Underlying conditions and risk of hospitalisation, ICU admission and mortality among those with COVID-19 in Ireland: A national surveillance study

Risks of and risk factors for COVID-19 disease in people with diabetes: a cohort study of the total population of Scotland

Working group for the surveillance control of COVID-19 in Spain. The first wave of the COVID-19 pandemic in Spain: characterisation of cases and risk factors for severe outcomes

The swedish covid-19 intensive care cohort: risk factors of ICU admission and ICU mortality

Severe COVID-19 in people with type 1 and type 2 diabetes in Sweden: a nationwide retrospective cohort study

Impact of comorbidity burden on mortality in patients with COVID-19 using the Korean health insurance database

Collider bias undermines our understanding of COVID-19 disease risk and severity

A structural approach to selection bias

Does diabetes mellitus mitigate the gender gap in COVID-19 mortality?

The C-word: scientific euphemisms do not improve causal inference from observational data

Using big data to emulate a target trial when a randomized trial is not available

A new method of classifying prognostic comorbidity in longitudinal studies: development and validation

Comorbidity: an R package for computing comorbidity scores

R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing

Vienna: Federal Ministry of Social Affairs, Health, Care and Consumer Protection

Vienna: Federal Ministry of Social Affairs, Health, Care and Consumer Protection

Dateneinmeldung der Bundesländer an BMI und BMSGPK. Vienna: Federal Ministry of the Interior, Federal Ministry of Social Affairs, Health, Care and Consumer Protection

Mediation analysis: a practitioner's guide

An introduction to g methods

Acknowledgements We thank our colleagues from the Austrian COVID Forecast Consortium for fruitful discussions and methodological support.

Comorbidities and COVID-19 hospitalization, ICU admission and hospital mortality in Austria