Klebsiella pneumoniae is a pathogen commonly found in community-onset bacteremia. It causes an invasive syndrome that is frequently presented by metastatic infections and abscesses elsewhere and that is necessary for surgical or drainage intervention. To achieve a scoring algorithm to identify patients with community-onset K. pneumoniae bacteremia (CoKPB) who are at risk for abscess occurrences, a retrospective cohort study consisting of adults with CoKPB was conducted.A 6-year cohort study consisting of adults having community-onset monomicrobial K. pneumoniae bacteremia was conducted. In addition to clinical variables collected from medical records, the hypermucoviscosity (HMV)-related gene (rmpA and magA) and an HMV phenotype were integrated into the proposed scoring algorithm.Of the 258 eligible adults, only 79 (30.6%) had abscesses related to bacteremia. Besides the presence of magA (ie, capsular serotype K1) and the HMV-phenotype, five clinical predictors were significantly associated with abscesses in a multivariate analysis: male gender, comorbidities with diabetes mellitus or neurological disorders, recent chemotherapy, and high c-reactive protein levels. Together, these predictors were used to calculate the CoKPB abscess score. Based on the scoring algorithm, a cut-off value of +2 yielded the high sensitivity (93.7%) and the acceptable specificity (50.8%); the area under the ROC curve was 0.83.A simple scoring algorithm that has substantial sensitivity and satisfactory specificity was proposed and the importance of the HMV phenotype or capsular K1 serotype was emphasized. The proposed predictive model needs external validation, but this scoring algorithm might be convenient and useful for clinicians.