Factors associated with positive blood cultures in children in nine African and Asian countries: the ACORN2 surveillance network
Ardura-Garcia C., Hopkins J., Lee SJ., Waithira N., Painter C., Ling CL., Roberts T., Miliya T., Obeng-Nkrumah N., Opintan JA., Abbeyquaye EP., Hamers RL., Saharman YR., Sinto R., Karyanti MR., Ibrahim RF., Akech SO., Duangnouvong A., Choumlivong K., Feasey NA., Kululanga D., Lissauer S., Karkey A., Kunwar N., Erakhaiwu JE., Okeke IN., Adebiyi I., Oduola AB., Ogunbosi BO., Tongo OO., Ude IA., Aboderin O., Adeyemo AT., Edward SS., Osagie U., Nguyen Thi H., Thach PN., Giang TV., Hoang Thi LH., Trinh HT., van Doorn HR., Ashley EA., Turner P.
Background Blood culture (BC) in children has relatively low diagnostic yield and high contamination rates, limiting cost-effectiveness. We aimed to determine readily available baseline characteristics to identify hospitalised children with a likelihood of higher diagnostic yield in low- and middle-income countries. Methods We used data from ACORN2, a prospective clinical surveillance network including 19 hospitals across Africa and Asia. We included participants <18 years, hospitalised for a suspected infection, prescribed parenteral antibiotics and with a BC sample. Sociodemographic and clinical data were recorded for each infection episode and linked to routine microbiology data. We described true pathogen (non-contaminant) BC positivity proportion and performed mixed-effects logistic regression, with study site and patient as the random effect, to identify factors associated with BC positivity. Results Of the 26 407 paediatric infection episodes, 17 815 (67%) had a BC sample and 15 384 were included in the analysis. BC results were: true pathogens in 689 (4.5%), contaminants in 1399 (9%) and uncertain pathogens in 143 (0.9%). In the multivariable model, factors associated with a positive BC were age (29 days–12-month-olds OR 1.33, 95% CI 1.06 to 1.66 and 5–18 year-olds OR 1.62, 95% CI 1.30 to 2.01 vs 1–4 year-olds), number of clinical severity signs (OR 1.29, 95% CI 1.18 to 1.40 per one sign) and hospital acquired infection (OR 3.05, 95% CI 2.30 to 4.06 vs community-acquired). Suspected diagnosis of sepsis (OR 2.09, 95% CI 1.67 to 2.61), gastrointestinal/abdominal (OR 2.36, 95% CI 1.78 to 3.13), skin and soft tissue or bone (OR 3.64, 95% CI 2.57 to 5.14) and genitourinary infection (OR 2.22, 95% CI 1.39 to 3.56) were more likely to have a positive BC, compared with respiratory infections. Conclusion We confirmed the low BC yield among hospitalised children. We identified groups for which diagnostic stewardship efforts to increase BC uptake should be prioritised and others in which it could be limited in times of financial or logistic constraints.