Abstract
Aim
Kingella kingae seems to be the most common cause of osteoarticular infections (OAI) in children under 48 months of age (1). Recent studies had shown that K. kingae is poorly susceptible to anti-staphylococcal penicillin and some isolates produce beta-lactamase (2). This led to the need for new treatment guidelines for OAI in populations in which K. kingae is frequent. Our study aimed to design a model which could predict K. kingae OAI in order to initiate appropriate empirical treatment on hospital admission.
Method
We performed a retrospective cohort study in children from 1 month to 15 years old diagnosed with OAI, hospitalized between 2006 and 2018. Mann-Whitney test and Fisher's exact test were used for data analysis. The model predicting K. kingae OAI was designed using logistic regression.
Results
247 children were included in the study, 126 (51%) had osteomyelitis (OM), 83 (33.6%) septic arthritis (SA) and 38 (15.4%) combined OM and SA. The median age was 52 (IQR 20–122) months, male-to-female ratio was 1.57:1. Pathogens were isolated in 101 (40.9%) cases with the following frequency: Staphylococcus aureus (n=59), K. kingae (n=13), Streptococcus pyogenes (n=11), Streptococcus pneumoniae (n=8).
Patients with K. kingae OAI had lower CRP levels compared to other pathogens (p<0.05). WBC was higher compared to S. aureus OAI (p=0.011), children with K. kingae OAI were younger than children infected with S. aureus (p<0.001) and S. pyogenes (p=0.003).
Based on this information we designed a predictive model using previous parameters as predictors of outcomes. The model had a 92.3% sensitivity and a 77.5% specificity.
Then, we tried to test the model's predictive power based on the treatment failure of empirical anti-staphylococcal antibiotics in the group of children with OAI without the known pathogen. In the subgroup for which the model predicted K. kingae OAI, antibiotic treatment had to be changed in 6/59 cases. It had to be changed in only 1/83 cases in the non-K. kingae group (p=0.021).
Conclusions
Despite poor specificity of the model, we found it to be more important to include all K. kingae OAI, that can be then properly treated. Additionally, with good specificity we acquire good negative predictive value, which means that children, for whom the model did not predict K. kingae OAI, can be safely treated with anti-staphylococcal penicillin.
