Endocarditis risk stratification with scores: what about reproducibility? The case of NOVA and DENOVA scores for...

NOVA and DENOVA scores were developed to guide endocarditis risk assessment in Enterococcus faecalis Bacteremia (EfB), but some of their criteria may be open to interpretation. We aimed to evaluate their inter-rater reliability and feasibility.

Thirty-two physicians from four specialties involved in the management of endocarditis independently evaluated eight EfB patient records using the NOVA and DENOVA scores. Each score was applied eight times per case. Inter-rater reliability was measured with Krippendorff's alpha, and agreement with Fleiss' Kappa. Completion time was also recorded.

No record received identical scores from all raters. NOVA showed low inter-rater reliability (α = 0.37), while DENOVA reached moderate levels (α = 0.49). High agreement was found for extreme score values, but agreement dropped markedly for intermediate values. Among score items, Auscultation of murmur (A) and Valve disease (V) had the highest reliability (α > 0.8), while Duration of symptoms (D) and Origin of infection (O) had the lowest (α < 0.2). Completion times were similar between NOVA and DENOVA but varied by specialty.

The reproducibility of these scores is limited, especially near critical thresholds, highlighting the need to complement scoring tools with clinical judgment in EfB.