Statistical analysis
Descriptive statistics were used to compare the patients with and
without prior therapeutic anticoagulation and to estimate the prevalence
of the outcomes. Categorical parameters were presented as counts with
percentages, continuous parameters with medians and interquartile
ranges, based on their non-normal distribution tested with the
Shapiro-Wilk test. Comparisons were performed using Mann-Whitney U test
or Chi-Squared test as appropriate. In order to compare the outcomes
while adjusted for potential confounding resulting from the
non-randomized design of our observational study, we applied propensity
score-matching methods. In our database, 174 (15%) patients had missing
information for BMI. Therefore, we first imputed BMI values with single
imputation using predicted values from multivariable models including
age, sex, hypertension, diabetes mellitus, and the outcome mortality.
Propensity scores were generated using a multivariable logistic
regression model with prior therapeutic anticoagulation use as outcome
and the 17 variables previously stated as predictors. The patients with
therapeutic anticoagulation were subsequently matched in a 1:3 ratio
with patients without therapeutic anticoagulation prior to admission on
these propensity scores with Nearest Neighbor Matching techniques
without replacement and a caliper width of 0·1 of the standard deviation
of the logit of the propensity score. To evaluate the balance of
measured confounders between exposed and unexposed groups, we calculated
the standardized mean difference (SMD). An SMD <0.25 indicated
balance of matched cohorts.30
Associations between prior therapeutic use of anticoagulants and the
outcomes overall in hospital mortality, admission to ICU, occurrence of
PE, critical respiratory state, and the need for invasive mechanical
ventilation were estimated as risk ratios (RR) with 95% confidence
intervals (CI). Estimating the differences in length of hospital stay
between the two groups was performed using Mann-Whitney U test
stratified by mortality.
In our secondary analysis, we compared VKA versus no therapeutic
anticoagulation, DOAC versus no therapeutic anticoagulation and VKA
versus DOAC on the before mentioned outcome parameters. Matching was
performed using R-software/studio Version 1.3.1093 and statistical
analysis were performed using STATA/SE 16.0 (Stata Corp, TX, USA).