ΔCt-Informed, Calibrated Logistic Regression Accurately Attributes mecA in Staphylococcus aureus-Positive Wound Specimens.

Abstract

In wound specimens, co-detection of mecA and Staphylococcus aureus by PCR does not necessarily indicate MRSA because coagulase-negative staphylococci (CoNS) frequently harbor mecA. We evaluated a {Delta}Ct-informed, biologically gated, calibrated logistic regression to attribute mecA to S. aureus versus CoNS. Using paired culture/AST and multiplex real-time PCR Ct values (internal n=93; external n=47), we trained 5-fold cross-validated models in the culture-positive S. aureus subset (n=36) and applied an S. aureus PCR gate (no attribution when S. aureus PCR is negative). The primary model achieved sensitivity 90.9% and specificity 92.0% for MRSA attribution with AUC 0.931 (out-of-fold). Decision curve analysis showed positive net benefit across clinically relevant thresholds; at the prespecified 50% cutoff, the model achieved a net benefit of 0.222 compared with negative benefit for a treat-all strategy. In an external cohort, S. aureus detection by PCR versus culture showed 92.3% sensitivity and 97.1% specificity; within S. aureus PCR-positives (n=12), MRSA attribution reached 100% sensitivity and 87.5% specificity (accuracy = 91.7%). This framework improves mecA interpretability in polymicrobial specimens and can reduce unnecessary MRSA-directed antibiotics.