Development and validation of a risk score for predicting positivity of blood cultures and mortality in patients with bacteremia and fungemia

Time-Sensitive Antibiotic Adjustments in Gram-Negative Bacteremia: A Survival Perspective

BACKGROUND This retrospective study was designed to assess risk factors for increased 30-day mortality from bacteremia caused by Escherichia coli or Klebsiella pneumoniae and to calculate the optimal point of time for patients to move from inappropriate empirical antimicrobial therapy to appropriate treatment. MATERIAL AND METHODS This retrospective study analyzed data from 610 patients with a diagnosis of E. coli- and K. pneumoniae-induced bacteremia collected between 2020 and 2023, including population-based information, infection bacteria, comorbidities, and treatment duration. Patients were categorized into 2 groups based on whether they received appropriate empirical antibiotic therapy (AEAT) or inappropriate empirical antibiotic therapy (IEAT). Propensity score-matched (PSM) analysis was performed using 1: 1 nearest neighbor matching. Receiver operating characteristic (ROC) curve analysis determined the time point for patients in the IEAT group to transition to appropriate therapy. RESULTS The study found that 30-day mortality was higher in the IEAT group than in the AEAT group (P=0.043). Multifactorial Cox regression analysis after PSM indicated that Pitt score (P<0.001), age-adjusted Charlson comorbidity index score (P<0.001), and inappropriate treatment days (P=0.018) were independent risk factors for increased 30-day mortality. The area under the ROC curve value was 0.613, and the maximum Youden index corresponded to a time turning point of 8 days. CONCLUSIONS Pitt and aCCI scores and inappropriate treatment days are significant risk factors for increased 30-day mortality in patients with E. coli or K. pneumoniae. Timely transition from inappropriate antibiotic to appropriate antibiotic therapy within 8 days was found to improve survival.

Sepsis death risk factor score based on systemic inflammatory response syndrome, quick sequential organ failure assessment, and comorbidities

OBJECTIVE

In this study, we aimed to evaluate the death risk factors of patients included in the sepsis protocol bundle, using clinical data from qSOFA, SIRS, and comorbidities, as well as development of a mortality risk score.

DESIGN

This retrospective cohort study was conducted between 2016 and 2021.

SETTING

Two university hospitals in Brazil.

PARTICIPANTS

Patients with sepsis.

INTERVENTIONS

Several clinical and laboratory data were collected focused on SIRS, qSOFA, and comorbidities.

MAIN VARIABLE OF INTEREST

In-hospital mortality was the primary outcome variable. A mortality risk score was developed after logistic regression analysis.

RESULTS

A total of 1,808 patients were included with a death rate of 36%. Ten variables remained independent factors related to death in multivariate analysis: temperature ≥38 °C (odds ratio [OR] = 0.65), previous sepsis (OR = 1.42), qSOFA ≥ 2 (OR = 1.43), leukocytes >12,000 or <4,000 cells/mm3 (OR = 1.61), encephalic vascular accident (OR = 1.88), age >60 years (OR = 1.93), cancer (OR = 2.2), length of hospital stay before sepsis >7 days (OR = 2.22,), dialysis (OR = 2.51), and cirrhosis (OR = 3.97). Considering the equation of the binary regression logistic analysis, the score presented an area under curve of 0.668, is not a potential model for death prediction.

CONCLUSIONS

Several risk factors are independently associated with mortality, allowing the development of a prediction score based on qSOFA, SIRS, and comorbidities data, however, the performance of this score is low.

Sequences of Events from the Electronic Medical Record and the Onset of Infection

We present a novel model of time-series analysis to learn from electronic health record (EHR) data when infection occurred in the intensive care unit (ICU) by translating methods from proteomics and Bayesian statistics. Using 48,536 patients hospitalized in an ICU, we describe each hospital course as an 'alphabet' of 23 physician actions ('events') in temporal order. We analyze these as k-mers of length 3-12 events and apply a Bayesian model of (cumulative) relative risk (RR). The log2-transformed RR (median=0.248, mean=0.226) supported the conclusion that the events selected were individually associated with increased risk of infection. Selecting from all possible cutoffs of maximum gain (MG), MG>0.0244 predicts administration of antibiotics with PPV 82.0 %, NPV 44.4 %, and AUC 0.706. Our approach holds value for retrospective analysis of other clinical syndromes for which time-of-onset is critical to analysis but poorly marked in EHRs, including delirium and decompensation.