Gastrointestinal Bleeding in Hospitalized covid‐19 Patients: a propensity Score Matched Cohort Study

Accepted Article

---

This article is protected by copyright. All rights reserved

score matching procedures. Since it could not be included in the propensity score model, it was 

controlled for in analyses using Firth’s penalized method.

GI bleed characteristics were first summarized descriptively for all patients with GI bleed. Next, the 

variable balance for propensity score matching was assessed. Frequencies and percentages were 

presented for categorical variables and means, medians, standard deviations, and interquartile ranges 

were presented for continuous variables. Variable balance was assessed using the absolute 

standardized difference, with an absolute difference smaller than 0.25 considered acceptable[23].

To assess the association between hospital medication use with GI bleed during hospitalization, 

hospital medication use was summarized descriptively across GI bleed groups (frequencies, 

percentages, means, medians). Conditional logistic regression was then used to account for the 

propensity score matched data. Hospital medication use for each medication type was considered as a 

potential risk factor. For the medications antiplatelets, anticoagulants, steroids, and NSAIDs, hospital 

medication use was considered to occur for patients with GI bleed only if the medication was ordered 

prior to the GI bleed. Antiplatelet agents searched for included: aspirin, dipyridamole, cilostazol, 

thienopyridines (clopidogrel, prasugrel, ticlodipine, ticagrelor), GPllb/llla inhibitors (tirofiban, 

abcixab, eptifibatide), PAR-1 inhibitor (vorapaxar). Anticoagulants searched for included: warfarin, 

unfractionated heparin, low molecular weight heparin, dalteparin, fondaparinux, rivaroxaban, 

apixaban, edoxaban, dabigatran, and desirudin.

For the medications proton pump inhibitors (PPIs) and histamine receptor blocker (H2R blockers), 

hospital medication use for all patients was considered to occur for patients if prescribed 

prophylactically within the first 24 hours of admission, and additionally only for patients with GI 

bleed if the medication was ordered prior to the GI bleed. 

 Interaction between home and hospital 

medication use was assessed for each medication type in the multivariable models but none 

were found to be significant. All models were assessed for multicollinearity. Timing and 

adherence to home medications could not be verified and thus we did not look at medication 

risk factors for GI bleed on admission.