Predicting Minimum Inhibitory Concentrations for WHO Highest Priority Critically Important Antimicrobials in Non-Typhoidal Salmonella: A Linear Regression Model Using Resistance Determinants

Ceftriaxone and ciprofloxacin are WHO Highest Priority Critically Important Antimicrobials (HPCIAs) for the treatment of severe and invasive infections caused by non-typhoidal Salmonella (NTS). This study aimed to develop linear regression models based on resistance determinants to predict the minimum inhibitory concentrations (MICs) of these two antibiotics. A total of 321 NTS isolates were included for ceftriaxone prediction, and 15.6% (50/321) were resistant; 320 were included for ciprofloxacin prediction, and 50.6% (162/320) were non-susceptible. blaCTX-M, blaCMY-2, and blaDHA-1 increased the ceftriaxone MIC, with the former two conferring resistance. qnr genes and gyr mutations elevated ciprofloxacin MIC and conferred a non-susceptible phenotype. Using leave-one-out cross-validation, the exact MIC prediction rates were 96.9% for ceftriaxone and 73.8% for ciprofloxacin, while the prediction accuracy within ±1 doubling dilution reached 98.8% and 93.4%, respectively. For binary classification, the ceftriaxone model achieved a sensitivity of 94.0% and specificity of 100.0% in distinguishing susceptible from resistant isolates, while the ciprofloxacin model achieved a sensitivity of 98.8% and specificity of 100.0% in distinguishing susceptible from non-susceptible isolates. This study showed that a multivariable linear regression model incorporating resistance determinants can accurately predict MICs of ceftriaxone and ciprofloxacin for NTS, thereby facilitating rapid infection control.