ABSTRACT
Sepsis is one of the leading causes of hospitalization and death among hemodialysis patients. Infections due to multidrug-resistant organisms (MDROs) are common among these patients, but empiric broad-spectrum coverage for every septic patient is associated with unfavorable outcomes. A retrospective case-control study was conducted at Shamir Medical Center, Israel (July 2016-April 2020), to determine predictors of MDRO infections among septic (per SEPSIS-3) ambulatory adult hemodialysis patients with permanent dialysis access (i.e., fistula, graft, or tunneled Perm-A-Cath). MDROs were determined according to established definitions. Least Absolute Shrinkage and Selection Operator (LASSO) regression was used to construct a prediction score and determine its performance. Of 509 patients, 225 (44%) had microbiologically confirmed infection, and 79 patients (35% of 225) had MDROs. The eventual independent predictors of MDRO infections were Perm-A-Cath access (vs. fistula or graft, aOR = 3, CI-95% = 2.1-4.2) and recent hospitalization in the previous three months (aOR = 2.3, CI-95% = 1.6-3.3). The score to predict MDRO sepsis with the highest performances contained seven parameters and displayed an area under the receiver operating characteristic curve (ROC AUC) of 0.74. This study could aid in defining a group of hemodialysis patients for which empiric broad-spectrum agents could be safely avoided.
ABSTRACT
OBJECTIVE: In the era of widespread resistance, there are 2 time points at which most empiric prescription errors occur among hospitalized adults: (1) upon admission (UA) when treating patients at risk of multidrug-resistant organisms (MDROs) and (2) during hospitalization, when treating patients at risk of extensively drug-resistant organisms (XDROs). These errors adversely influence patient outcomes and the hospital's ecology. DESIGN AND SETTING: Retrospective cohort study, Shamir Medical Center, Israel, 2016. PATIENTS: Adult patients (aged >18 years) hospitalized with sepsis. METHODS: Logistic regressions were used to develop predictive models for (1) MDRO UA and (2) nosocomial XDRO. Their performances on the derivation data sets, and on 7 other validation data sets, were assessed using the area under the receiver operating characteristic curve (ROC AUC). RESULTS: In total, 4,114 patients were included: 2,472 patients with sepsis UA and 1,642 with nosocomial sepsis. The MDRO UA score included 10 parameters, and with a cutoff of ≥22 points, it had an ROC AUC of 0.85. The nosocomial XDRO score included 7 parameters, and with a cutoff of ≥36 points, it had an ROC AUC of 0.87. The range of ROC AUCs for the validation data sets was 0.7-0.88 for the MDRO UA score and was 0.66-0.75 for nosocomial XDRO score. We created a free web calculator (https://assafharofe.azurewebsites.net). CONCLUSIONS: A simple electronic calculator could aid with empiric prescription during an encounter with a septic patient. Future implementation studies are needed to evaluate its utility in improving patient outcomes and in reducing overall resistances.
