Study design and cohorts

This retrospective, nation-wide study complied with the Declaration of Helsinki and was approved by the ethics committee of Lower Austria (“Ethikkommission Niederösterreich”: GS1-EK-4/747-2021). Patient informed consent was waived by the ethics committee of lower Austria due to the study design of a retrospective registry-based study. The data were available from the Austrian Health Insurance Funds. Approximately 98% of the Austrian population is registered in the public health insurance system. Health care in Austria is a national system with good access to care, few malpractice lawsuits, and little tendency toward overuse of medical resources. Access to individual services is regulated by social insurance law²⁴.

Patients ≤ 18 years of age, hospitalized in Austria due to the main diagnosis COVID-19 (ICD-10 Codes U071, U072, U049) from 1 January 2020 to 31 December 2021 (i.e., including patients from the first four pandemic waves) and patients ≤ 18 hospitalized in Austria due to the main diagnosis Influenza (ICD-10 codes: J09, J100, J101, J108, J110, J111, J118, J10) from 30 December 2015 to 31 December 2021 were included in this study. Patient characteristics (age, sex, region, and medication profiles) were obtained from 1 year before index-hospitalization until study cut-off. Two age-, sex-, and region-matched control groups (approximately 10 controls for each patient) were randomly chosen from the population registered in the Austrian Health Insurance Fund. One consisting of individuals not hospitalized due to COVID-19 in the years 2020 and 2021 and the second consisting of individuals not hospitalized due to Influenza in the years 2016–2021. Data on the control groups were available from 1 year before the first patient was hospitalized until study cut-off. Death dates were available at study cut-off.

Study outcomes

The primary outcome of the study was time from hospital admission to hospital discharge. Secondary outcomes were readmission after hospital discharge due to any reason (conditional on hospital survival), in-hospital death and all-cause death. Readmission was defined based on billing information (MEL codes) available from the Austrian Health Insurance Funds. For patients, we used the first readmission after the index hospital stay. Readmissions within 2 weeks after the index admission were assumed to belong to the index admission and were not counted as readmission. For controls, in-hospital death and time to death was evaluated from the index hospital admission date and re-admission was calculated from the index hospital discharge of the matched patient (Table S1).

Statistical analysis

For each patient, age, sex, region (Vienna, Lower Austria, Upper Austria, Burgenland, Carinthia, Salzburg, Tyrol, Styria, or Vorarlberg), and ATC codes describing prescribed medications were available from 1 year before the index hospitalization. ATC codes for medications before hospitalization were summarized in medication groups (Table S2). For each medication group, a binary variable was created, assigned a value of 1 if at least one prescription with the corresponding ATC code was recorded within 1 year prior to the index hospitalization, and 0 otherwise. Available ATC codes were categorized into medications groups based on their therapeutic use and/or chemical characteristics according to their relationships to the underlying disease processes they treat or prevent (MG 1: Anticoagulants, MG 2: Antibiotics, Antivirals, Antiprotozoals or Anthelmintics, MG 3: Insulins and other Antidiabetics, MG 4: heart drugs, MG 5: Antihypertensives incl. Diuretics and Renin–angiotensin–aldosterone system inhibitors, MG 6: Beta Blockers, MG 7: Statins, Fibrates incl. Proprotein convertase subtilisin/kexin type 9 inhibitors and Inclisiran, MG 8: Immunosuppressants and Immunomodulators, MG 9: Systemic Steroids, MG 10: Chemotherapy, MG 11: Iron supplements, Erythropoietic stimulating agents, Vitamin B12, folic acid, MG 12: Antacids incl. Antihistamines, MG 13: Vitamin D and other Vitamin supplements, MG 14: Caplacizumab, MG 15: Systemic Hemostatics, MG 16: Hereditary angioedema Therapeutics, MG 17: Peripheral Vasodilators, MG 18: Hormonal contraceptives and similar hormone preparations, MG 19: Immunoglobulins, MG 20: Interferons and CSF, MG 21: NSAR and other anti-inflammatory drugs, MG 22: Gout medications, MG 23: Antiepileptics, MG 24: Antipsychotics, MG 25: Rhinological and throat antiseptics, MG 26: inhaled anti-obstructive drugs, MG 27: inhaled steroids, MG 28: other COPD drugs, MG 29: Cold and Cough preparations, MG 30: Systemic Antihistamines). Medication groups prescribed to fewer than 5 patients in either the COVID-19 or Influenza group were excluded from statistical modeling due to underrepresentation. For controls, a similar medication profile was generated using prescribed medication 1 year before the index hospital stay of the matched patient. All medication groups, except Rhinological and throat antiseptics (MG 25), Cold and Cough preparations (MG 29) and hormonal contraceptives and similar hormone preparations (MG18) were summed up for the number of medication groups, categorized in 0, 1, ≥ 2, the category ≥ 2 indicating “polypharmacy”. These three medication groups were excluded from the summation because they were not assumed to be associated with severe underlying diseases potentially influencing outcome in children and adolescents.

Numbers and percentages were used to summarize categorical variables, medians or interquartile ranges for continuous variables. Medication groups were compared between groups using Fisher-Exact tests. To evaluate the association between medication groups, polypharmacy and time to hospital discharge, first, for each medication group simple Cox regression models were calculated accounting for sex, age (categorized in 0–5, 6–10, 11–15 and 16–18), polypharmacy, pandemic wave and the respective medication group with clustering variable region. Due to the small number of deaths, patients dying during the in-hospital follow-up were censored at the time of death. A multivariable Cox-Regression model was performed including sex, age, polypharmacy, wave and all medication groups with a p value smaller than 0.1 in the simple models. These analyses were performed separately for the COVID-19 and the Influenza Group.

To compare time to discharge between COVID-19 and Influenza hospitalized patients, first, propensity score matching was performed to account for potential bias. Propensity score matching was based on a logistic regression including age, sex, region and all medication groups with a p < 0.05 in the comparison between cohorts using Fishers exact test. To evaluate the association between disease groups and time to hospital discharge, simple and multivariable Cox-Regression models were performed, the multivariable model including group, sex, age, polypharmacy, wave and all medication groups with a p value smaller than 0.1 in the simple models. A hazard ratio larger than one is indicating a smaller probability for hospital discharge in the reference group, suggesting a shorter hospital stay. Time to readmission was evaluated using similar analyses. For time to readmission, a hazard ratio smaller than one is indicating a larger probability for readmission in the reference group. For a detailed definition of medication groups and confounder, see supplemental Tables S1 and S2. Due to the small number of deaths, statistical modelling or PSM could not be conducted. In-hospital mortality was exploratory compared with Fisher-Exact test and all-cause mortality using log-rank test. Schönfeld residuals were used to evaluate the proportional hazard assumption and variance inflation factors to evaluate multicollinearity. A p value smaller than 0.05 was considered significant. All analyses were performed using R, release 4.2.2.²⁵.