Development and validation of a prediction model based on clinical and CT features for invasiveness of K. pneumoniae liver abscess

Klebsiella pneumoniae liver abscess (KPLA) complicated with extrahepatic migratory infection (EMI) is defined as invasive KPLA. The current study aimed to develop and validate a risk prediction model for the invasiveness of KPLA.From 2010 to 2020, KPLA patients from four institutes were selected retrospectively. In the development cohort, risk factors from a logistic regression analysis were utilized to develop the prediction model. External validation was performed using an independent cohort.A total of 382 KPLA patients comprised two separate cohorts: development cohort (institute 1, n = 286) and validation cohort (institute 2-4, n = 86). The overall incidence of EMI was 19.1% (development cohort, n = 55; validation cohort, n = 18, p > 0.05). In the development cohort, four risk factors (age ≤ 40 years, fasting blood glucose (FBG) > 7 mmol/L, no rim enhancement, and thrombophlebitis on CT), significantly associated with EMI, were incorporated into the scoring system. The area under curve (AUC) of the receiver operating characteristic curve (ROC) in the development and validation cohorts was 0.931 (95% confidence interval [CI]: 0.93-0.95) and 0.831 (95% CI: 0.86-0.91), respectively. The calibration curves fitted well. The incidence of EMI was 3.3% and 56.5% for the low- (total scores ≤ 4) and high-risk (total scores > 4) groups in the development cohort, and 3.2% and 66.7% in the validation cohort (all p < 0.001), respectively.Age ≤ 40 years, FBG > 7 mmol/L, no rim enhancement, and thrombophlebitis were independent risk factors for EMI. This validated prediction model may aid clinicians in identifying KPLA patients at increased risk for invasiveness.• Four risk factors are significantly associated with extrahepatic migratory infections (EMI): age ≤ 40 years, fasting blood glucose (FBG) > 7 mmol/L, no rim enhancement, and thrombophlebitis on CT. • Based on these risk factors, the current study developed and validated a prediction model for the invasiveness of Klebsiella pneumoniae liver abscess (KPLA). • This validated prediction model may in the help early identification of KPLA patients at increased risk for invasiveness.