The following is a summary of “Development and derivation of bacteremia prediction model in patients with hepatobiliary infection,” published in the November 2023 issue of Emergency Medicine by Choi, et al.

People who go to the emergency room (ED) often have hepatobiliary infections, which have a high death rate. A good bacteremia forecast model would help find bacteremia quickly and help doctors treat hepatobiliary diseases correctly in the emergency department. For a study, researchers sought to make a bacteremia prediction model for hepatobiliary infections in the ED that could be validated inside and outside the hospital. Patients with hepatobiliary infections were taken from two major hospitals’ old group databases, one from January 2018 to December 2019 and the other from January 2016 to December 2019. Multivariable logistic regression was used to find independent risk factors in a developing sample. They gave predicted factors a weighted value and made a prediction model that was checked by people inside and outside the company. 

The area under the receiver area under the curve (AUC) was used to measure discrimination. Randomly, 1,568 patients from one hospital cohort were split into two groups: a developing group with 927 patients (60%) and an internal validation group with 641 patients (40%). For external validation, 736 patients from the other hospital cohort were used. Bacteremia rates were 20.5% in the developing cohort, 18.1% in the internal validation cohort, and 23.1% in the external validation cohort. Age, three vital signs, and five lab tests were among the nine important factors to identify bacteremia. 

They used their bacteremia prediction rule on the validation sample and found that 56.5% of the internal and 53.8% of the external validation groups had low-risk bacteremia (8.6% and 13.9%, respectively). It was 0.727 (95% CI: 0.686–0.767) for the developing cohort, 0.730 (95% CI: 0.679–0.781) for the internal validation cohort, and 0.715 (95% CI: 0.672–0.758) for the external validation cohort. The sensitivity was 73.2% for internal validation and 67.6% for external validation. The specificity was 63.0% for both. A bacteremia forecast model could be useful to figure out the chance of getting bacteremia from a hepatobiliary infection. Also, it might mean that low-risk people don’t need blood cultures as often.

Source: sciencedirect.com/science/article/abs/pii/S0735675723004333