This study was performed in the ICUs of the University Hospital Basel, Switzerland, an academic tertiary medical care center treating more than 47,000 emergencies and with more than 35,000 admissions per year. Patients were derived from an ongoing prospective cohort of patients with BSI .
The study was approved by the local ethics committee (Ethikkommission Nordwest-und Zentralschweiz, Nr. 2014/165) according to the declaration of Helsinki, and patients’ consent was waived.
Demographics and clinical characteristics
From 2011 to 2014, all consecutive adult patients with BSI treated in the medical or surgical ICUs of an academic medical care center were identified. Patients without consecutive ICDSC scoring or with delirium prior to the diagnosis of BSI were excluded. We assessed pertinent clinical, laboratory, and microbiological data including principal diagnoses, the SAPS II scores, comorbidities, transient episodes of coma on the day of BSI diagnosis, and serum levels of acute-phase proteins as described below. Treatment characteristics of all patients were assessed including ICU and hospital stay, duration of mechanical ventilation, the number of catheters and drainages, the use of anesthetics and sedatives during the three days prior to BSI, and the administration of anesthetics and neuroleptic drugs during the three days following BSI. The Intensive Care Delirium Screening Checklist (ICDSC) and outcomes were assessed as mentioned below.
In all patients treated with sedatives, sedation was managed and titrated by using the Riker Sedation-Agitation Scale (SAS). In all patients treated with sedatives, daily interruption of sedation was an integral component of the sedation protocol. To establish sedation, short active sedatives such as disoprivan were preferred providing absence of hemodynamic instability. In patients with hemodynamic instability, sedation was achieved by administration of benzodiazepines given at the lowest possible dose and titrated according to the sedation level as determined by SAS.
Definition and detection of infections and BSIs
A protocol for monitoring infections was established for all patients in the ICUs during the entire study period. It included drawing cultures of blood and urine, cultures from tracheal aspirates or sputum, and performance of a chest X-ray in any patient with new onset of fever or hypothermia. If other foci were suspected, these were sampled accordingly. Infections were diagnosed based on the patients’ clinical examination, radiological exams, laboratory findings, and microbiological results according to the Centers for Disease Control and Prevention (CDC) criteria . The diagnosis of BSI was defined according to the current guidelines  and confirmed by two infectious diseases specialists in patients with at least one positive blood culture and concurrent presence of the systemic inflammatory response syndrome. Systemic inflammatory response was diagnosed if two or more of the following criteria were fulfilled: body temperature >38 or <36 °C; heart rate >90/min; respiratory rate >20/min; serum leukocyte count >12 or <4 G/L or the presence of >10% immature neutrophil granulocytes in the blood samples. Catheter-related BSI was defined according to the CDC criteria (www.cdc.gov/hai/bsi/bsi.html).
New definitions for sepsis and septic shock were published in 2016 , recommending to define sepsis as life-threatening organ dysfunction, captured by the Sequential [Sepsis-related] Organ Failure Assessment (SOFA) score, and defining septic shock as the need for vasopressors to maintain a mean arterial blood pressure of ≥65 mmHg and serum lactate level >2 mmol/L in the absence of hypovolemia. Therefore, we retrospectively collected the respective information to calculate the SOFA score and to identify septic shock accordingly.
Definition and detection of new-onset delirium
Among several screening methods to detect delirium in ICUs, the Confusion Assessment Method for the ICU (CAM-ICU)  and the Intensive Care Delirium Screening Checklist (ICDSC)  have been most frequently employed. Both have been recommended for the screening of delirium in ICUs by the Society of Critical Care Medicine based on high-quality evidence.  Direct comparisons of the diagnostic accuracy of the CAM-ICU and the ICDSC have been performed in recent studies with heterogenous ICU populations revealing a higher sensitivity and specificity of the ICDSC than the CAM-ICU [22–24].
Based on these data, in our institution, the ICDSC is daily assessed in all ICU patients. For the present study, the ICDSC scores on the day before, the day of, and the day following diagnosis of BSI were used. To reduce the interference of sedation in mechanically ventilated patients, ICDSC assessments were performed in mechanically ventilated patients after routine daily interruption of sedation at our institution.
According to the studies and guidelines mentioned above, an ICDSC ≥4 was defined as delirium [9, 21].
Measurements of acute-phase proteins
For the present study, serum levels of acute-phase proteins (including CRP [C-reactive protein] and albumin), measured daily in all consecutive patients during the study period, were included on the day before, the day of, and the day following the diagnosis of BSI. CRP concentrations were determined by an enzyme immunoassay with a detection limit of 0.5 mg/l (EMIT; Merck Diagnostica, Switzerland). Values <10 mg/l are considered as normal.
The primary outcome was the development of delirium ±24 h around the diagnosis of BSI. Secondary outcomes were death during the same hospital stay, and return to functional baseline and unfavorable outcome in survivors defined as a Glasgow Outcome Score (GOS) of 1–3 at discharge.
The Chi-square and Fisher exact test (where appropriate) were used for comparisons of proportions. For continuous variables, the Shapiro–Wilk test was used to distinguish between normal and abnormal distributions. Normally distributed variables were analyzed using the Student’s t test and non-normally distributed variables using the Mann–Whitney U test. To address possible confounding, all variables found to be significant in univariable analyses and known risk factors (i.e., SAPS II, dementia/leukoencephalopathy) were included in the multivariable regression models after calculating odds ratios by logistic regression. In addition, SAS and SOFA were included into the multivariable logistic regression model to correct for the effects of sedation and organ failure on ICDSC scores . Stepwise logistic regression using stepwise forward and backward selection, as well as Akaike information criterion (AIC), was applied to identify variables independently associated with ICDSC ≥4, and collinearity between covariates was assessed.
The Hosmer–Lemeshow goodness-of-fit test was applied to check the final models. Areas under the receiver operating characteristic (ROC) curve were calculated for the final models to evaluate discrimination.
Two-sided p values ≤0.05 were considered statistically significant. Statistical analyses were performed with STATA® version 13.0 (Stata Corp., College Station, TX).