Effect of antimicrobial de-escalation strategy on 14-day mortality among intensive care unit patients: a retrospective propensity score-matched cohort study with inverse probability-of-treatment weighting

PURPOSE: This study aimed to investigate the prevalence of antimicrobial de-escalation (ADE) strategy and assess its effect on 14-day mortality among intensive care unit patients. METHODS: A single-center retrospective cohort study was conducted on patients admitted to the intensive care unit (ICU) with infectious diseases between January 2018 and December 2020. Patients were stratified into three groups based on the initial treatment regimen within 5 days of antimicrobial administration: ADE, No Change, and Other Change. Confounders between groups were screened using one-way ANOVA and Chi-square analysis. Univariate and multivariate analyses were performed to identify risk factors for 14-day mortality. Potential confounders were balanced using propensity score inverse probability of treatment weighting (IPTW), followed by multivariate logistic regression analysis to evaluate the effect of ADE strategy on 14-day mortality. RESULTS: A total of 473 patients met the inclusion criteria, with 53 (11.2%) in the ADE group, 173 (36.6%) in the No Change group, and 247 (52.2%) in the Other Change group. The 14-day mortality rates in the three groups were 9.4%, 11.6%, and 21.9%, respectively. After IPTW, the adjusted odds ratio for 14-day mortality comparing No Change with ADE was 1.557 (95% CI 1.078–2.247, P = 0.0181) while comparing Other Change with ADE was 1.282(95% CI 0.884–1.873, P = 0.1874). CONCLUSION: The prevalence of ADE strategy was low among intensive care unit patients. The ADE strategy demonstrated a protective effect or no adverse effect on 14-day mortality compared to the No Change or Other Change strategies, respectively. These findings provide evidence supporting the implementation of the ADE strategy in ICU patients. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s12879-023-08491-7.