Predicting clinically actionable metagenomic next-generation sequencing in sepsis through dynamic inflammatory trajectories

Metagenomic next-generation sequencing (mNGS) improves pathogen detection in sepsis, but positive results do not always prompt antimicrobial modification. We investigated whether dynamic host-response trajectories during the mNGS peri-report interval could predict clinical actionability.

This retrospective single-center study included 109 adults with sepsis who underwent mNGS and conventional culture, and 72 patients with complete prespecified predictor data entered model development. Baseline and dynamic logistic models were assessed for discrimination, calibration, prediction error, and exploratory decision-curve performance. Final fixed models underwent 1,000-resample bootstrap optimism correction.

mNGS positivity was 94.8% in the actionable group and 84.3% in the non-actionable group, with broader descriptive pathogen detection than culture. In the complete-case cohort, the peri-report change in white blood cell count (dWBC) remained associated with actionability after adjustment for the included baseline variables (adjusted OR, 1.14; 95% CI, 1.02-1.28; P = 0.025). Apparent AUC increased from 0.744 to 0.816 after adding dWBC. Optimism-corrected AUCs were 0.697 and 0.768, and corrected Brier scores were 0.228 and 0.207. No baseline-WBC interaction was detected (P = 0.494).

A less pronounced decline or an increase in WBC from pre-mNGS sampling to post-report reassessment was associated with a greater likelihood of subsequent documented mNGS-related antimicrobial modification.