Severe acute respiratory infections (SARI) are a leading cause of hospitalisation and mortality globally. Many SARI cases remain undiagnosed because kit-based PCR diagnostic panels are typically limited to one or a small number of known pathogens and may fail to identify low-abundance infections or novel, poorly characterised organisms. Here, we used metatranscriptomic sequencing to profile the total infectome of 300 PCR-negative SARI nasopharyngeal samples collected through sentinel hospital-based surveillance in New Zealand between 2014-2021. Our analysis revealed actively transcribing potential pathogens in 43% of SARI cases, comprising 10 RNA viruses, three DNA viruses, nine bacterial species and four fungal species. Notably, co-infections occurred in 26% of cases, revealing polymicrobial infections missed by routine diagnostics. Human rhinoviruses were the most frequently identified, despite not being detected by PCR, and multiple common-cold coronaviruses, human parechovirus A1 and parainfluenza virus type 4, were identified, although these were not included in the PCR screening panel. We also detected a range of bacterial and fungal species and uncovered highly expressed virulence and antimicrobial resistance genes. Infectome composition and diversity were shaped by key demographic and epidemiological factors, with strongest effects observed for age and year of sample collection, indicating that host characteristics and temporal dynamics influence both microbial richness and community structure. These findings highlight the limitations of current diagnostic strategies and the value of metatranscriptomics for comprehensive microbial identification. Integrating such genomic approaches into both clinical and public health frameworks could improve diagnostic accuracy