Potential drug-drug interactions and 30-day mortality in ICU patients with bloodstream infection: a single-center retrospective study.

backgroundPatients in intensive care units (ICUs) with bloodstream infection (BSI) commonly receive multiple antimicrobials and supportive drugs, which increases the likelihood of potential drug-drug interactions (pDDIs). Evidence focused specifically on ICU patients with BSI remains limited.

objectivesTo describe the prevalence and severity of pDDIs in ICU patients with BSI and to examine whether pDDI exposure was associated with 30-day mortality.

methodsWe performed a single-center retrospective cohort study of 90 adult ICU patients with an index episode of BSI between January 2019 and December 2024. Time zero was defined as the sampling time of the first qualifying positive blood culture or a clinically accepted positive blood metagenomic next-generation sequencing result for the index episode. Medication administration records, rather than prescription orders alone, were used for pDDI ascertainment. pDDIs were screened with the Micromedex Drug Interactions database (Merative, web-based version updated daily; accessed January 15, 2025). Exposure was defined within a fixed 48-hour window after time zero; a pDDI required actual administration of both interacting agents within the same 24-hour period during this window. Severity was standardized as mild, moderate, or severe. The primary outcome was 30-day all-cause in-hospital mortality. Because only 18 deaths occurred, the primary multivariable model included any pDDI exposure and SOFA score.

resultsSeventy of 90 patients (77.8%) had at least one pDDI within the fixed exposure window. Based on the highest patient-level severity, 13/70 (18.6%) had mild, 22/70 (31.4%) moderate, and 35/70 (50.0%) severe pDDIs. The most frequent combinations were vancomycin plus amikacin (18/90, 20.0%) and piperacillin/tazobactam plus vancomycin (15/90, 16.7%). The clinical consequences listed for common pairs were reference-predicted interaction consequences rather than adjudicated observed toxicities. In the parsimonious multivariable model, any pDDI exposure was associated with higher observed 30-day mortality (adjusted OR 4.23, 95% CI 1.27-14.09; P = 0.02), and each 1-point increase in SOFA score was also associated with mortality (adjusted OR 1.32, 95% CI 1.07-1.64; P = 0.01).

conclusionspDDIs were common in this ICU BSI cohort and were associated with higher observed 30-day mortality. These findings should be interpreted cautiously given the retrospective single-center design, limited event count, residual confounding, and incomplete control of time-dependent exposure. Structured pDDI screening may still support medication safety in critically ill patients. CLINICAL TRIAL NUMBER: Not applicable.