Tidyverse

Statistical analyses were performed using RStudio version 4.2.1 and packages: finalfit; dplyr; survival; survminer; TableOne, epiDisplay, and tidyverse (Integrated Development for R; Rstudio, PBC, USA). Delirium was coded as a binary variable: 4AT score ≥ 4 suggestive of delirium.^{22 (link),26 (link)} Continuous variables were assessed for differences between groups using an independent-samples t-test (parametric) or Mann-Whitney U test (non-parametric), and categorical variables using a chi-squared test. Multivariable logistic regression was used to analyse associations of delirium with: 1) mortality at 30 and 180 days; 2) need for post-acute inpatient rehabilitation; and 3) for patients discharged from the acute stay to a home within the catchment area, likelihood of readmission to an acute hospital within 180 days. All multivariable models included covariates: age, sex, SIMD quintile, COVID-19 within 30 days, and ASA grade. A p-value of < 0.05 was deemed significant. Assumptions of logistic regression were tested and satisfied.

Data were analysed using RStudio (RRID:SCR_000432), ggplot2 (RRID:SCR_014601), dplyr (RRID:SCR_016708), tidyverse (RRID:SCR_019186), ggrepel (RRID:SCR_017393), openxlsx (RRID:SCR_019185), ggthemes, ggsignif, gridExtra and Origin (RRID:SCR_014212) software. All datasets were tested for normality using the Shapiro–Wilkinson test. When a normal distribution was confirmed, a One-Way ANOVA test with a post hoc Tukey's HSD test was used for statistical comparison between multiple groups. For datasets that did not show normality, a Kruskal–Wallis test with a post hoc Dunn's test was applied. The asterisks in the figures correspond to the following p-values: *p < 0.05, **p < 0.01 and ***p < 0.001. Experiments were performed as biological triplicates on different clones and data with error bars are represented as mean ± standard deviation .