Metadata of study samples.

Antimicrobial resistance (AMR) is a rising global health threat and estimated to cause 700,000 deaths annually. Although blood cultures (BCs) are the reference standard to diagnose bloodstream infections and inference of antimicrobial susceptibility testing (AST), the method could fail to differentiate bacteria with similar biochemical characteristics. This study compared phenotypic method (AST) and whole genome sequencing (WGS) in bacterial species identification and assessment of AMR. Blood samples came from children who had been prescribed antibiotics at admission at the county referral hospitals in western Kenya and around the Lake Victoria region. BCs and AST were performed on BD Bactec 9050 and Phoenix 100 respectively. Out of the 960 BCs, 17 isolates were useable and were evaluated by WGS on the Oxford Nanopore PromethION platform. BD Phoenix system identified the 17 bacteria isolates as: 4 Escherichia coli, 8 Salmonella enterica serovar Typhi, 1 unspeciated Salmonella, 3 Staphylococcus aureus and 1 Streptococcus pneumoniae. WGS results differed from BCs in identifying the 9 Salmonella species, with WGS identifying the species as Salmonella enterica serovar Typhimurium. Conversely, WGS detected AMR determinants in bacteria that AST had classified as susceptible. In conclusion, we caution that BCs may not be providing correct identity of Salmonella species. The observed discrepancies between phenotypic and genotypic markers of drug resistance highlight the challenges in interpreting and predicting the functional utility of AMR determinants.