# The value of metagenomic next-generation sequencing in lower respiratory tract infections among critically ill patients in the ICU

## TL;DR

This study evaluates how metagenomic next-generation sequencing helps diagnose and treat lower respiratory tract infections in ICU patients.

## Contribution

The study demonstrates the clinical utility of mNGS in guiding treatment adjustments and predicting patient outcomes in ICU LRTI cases.

## Key findings

- mNGS detected pathogens with 80.1% sensitivity and 66.3% accuracy in ICU LRTI patients.
- Survivors had more diverse microbial profiles, with Corynebacterium striatum dominant compared to Pseudomonas aeruginosa in non-survivors.
- mNGS-guided treatment adjustments improved clinical symptoms in 58.6% of ICU patients.

## Abstract

Lower respiratory tract infections (LRTIs) frequently occur as a severe complication in intensive care unit (ICU) patients, substantially raising patient mortality rates and extending hospitalization periods. In this study, a retrospective cohort study of 261 suspected LRTI patients in the ICU of Nanjing Drum Tower Hospital between April 2021 and February 2024 was conducted. The results showed that metagenomic next-generation sequencing (mNGS) had a sensitivity of 80.1%, a specificity of 35%, and an accuracy of 66.3% across all samples. For pathogen detection, mNGS outperformed conventional microbiological testing (CMT) in detecting bacteria and DNA viruses, while CMT had a slight advantage in RNA virus detection, though the difference was not statistically significant (p = 0.305). When comparing microbial profiles between survival and death groups, survivors had a more diverse pathogen spectrum, particularly in bacteria and RNA viruses. There were 262 species detected in both groups, with Corynebacterium striatum being the dominant species in the survival group and Pseudomonas aeruginosa in the death group. In the 128 patients whose treatment plans were adjusted based on mNGS results, 59.4% underwent escalation, 25.8% had their medications changed, and 1.6% initiated new treatment regimens. Further follow-up revealed that mNGS - guided treatment adjustments were effective in improving clinical symptoms in 58.6% of ICU patients. A predictive model for patient outcomes was developed utilizing the random forest algorithm, achieving an area under the receiver operating characteristic curve (AUC) of 0.722.