Reproducible Identification of Staphylococcus aureus Bacteremia Clinical Subphenotypes.

BACKGROUND

Clinical heterogeneity in Staphylococcus aureus bacteremia (SAB) complicates clinical management and research. We have previously identified 5 clinically distinct subphenotypes of SAB associated with differences in outcomes and response to adjunctive rifampicin. Here, we aimed to identify these subphenotypes in geographically diverse observational cohorts, including a higher prevalence of methicillin-resistant S. aureus (MRSA) bacteremia and the USA300 clone.

METHODS

We studied 3 cohorts of adults with SAB from observational studies: a UK retrospective study (Edinburgh cohort 2; n = 463); a Dutch prospective study (IDISA [Improved Diagnostic Strategies in Staphylococcus aureus bacteremia study]; n = 490); and a prospective US study (SABG-PCS [Staphylococcus aureus Bacteremia Group Prospective Cohort Study]; n = 755). Subphenotypes were identified from routinely available clinical data using latent class analysis.

RESULTS

Patients from the SABG-PCS cohort had greater multimorbidity and more MRSA bacteremia (40.2% [303 of 755]), including infection with the USA300 clone (14.7% [111 of 755]). Five distinct subphenotypes were identified in each cohort: (1) older age and cardiometabolic multimorbidity; (2) nosocomial acquisition and intravenous catheter portal of entry; (3) community acquisition and metastatic infection; (4) chronic kidney disease; and (5) younger age, injection drug use, and metastatic infection. Bacterial genotypes varied substantially between the Edinburgh 2 and SABG-PCS cohorts but did not differ between subphenotypes within each cohort. 90-day mortality was highest in subphenotype A, and persistent bacteremia in subphenotypes C and E.

CONCLUSIONS

We have reproducibly identified 5 clinical subphenotypes of SAB in observational cohorts including diverse bacterial genetic lineages and a cohort with a high prevalence of MRSA and USA300 bacteremia. These robustly reproducible clinical subphenotypes provide a framework to rationalize the heterogeneity intrinsic to SAB.