Clinical risk factors and prediction model for infective endocarditis in patients with gram-positive bacteremia

Abstract

Abstract Introduction Gram-positive bacteremia (GPB) is a common condition in clinical practice that potentially leads to infective endocarditis (IE), a high morbidity and mortality disease. Clinicians still encounter challenges in determining the likelihood of developing IE in GPB patients. Comprehensive and invasive cardiac investigations in every patient with GPB for IE could be unnecessary and potentially harmful while missing proper investigations might result in the delay or misdiagnosis of this fatal condition. Objective We aimed to determine the significant clinical risk factors and develop the prediction model for IE in patients with GPB. Methods Data were obtained from medical records of hospitalized adult patients with GPB in our hospital from January 2016 to December 2020. IE was defined by the ‘definite’ group according to modified Duke criteria. The cohort was divided into training (80%) and test (20%) sets. We identified the association between clinical risk factors and the development of IE by logistic regression analyses. Significantly associated clinical risk factors in multivariate analysis were used in the prediction model. The area under the receiver operating characteristic curve (AUROC) was utilized to evaluate model performances. Results A total of 794 patients with GPB were included in the study. IE was diagnosed in 89 patients (11.2%). Among 21 selected clinical risk factors from univariate analysis, 10 variables, including the unknown origin of bacteremia, mechanical and bioprosthetic heart valve, duration of symptoms more than 7 days, presence of Roth spot, neurological symptoms, murmur, heart failure, number of positive hemoculture more than 2 specimens, and Staphylococcus aureus septicemia, were significantly associated with IE in multivariate analysis (Figure 1) and were utilized in the prediction model. The AUROC of prediction models were 0.943 and 0.865 in training and test sets, respectively. Conclusions Our prediction model based on clinical risk factors can precisely predict IE in patients with GPB. Using this model allows clinicians to address patients at risk and guide decision-making in which cases need an urgent and comprehensive workup for IE or a wait-and-see strategy.Figure 1