Infective endocarditis: Do we have an effective risk score model? A systematic review

BackgroundInfective endocarditis (IE) is a rare, highly morbid condition with 17% in-hospital mortality. A total of 25–30% require surgery and there is ongoing debate with regard to markers predicting patient outcomes and guiding intervention. This systematic review aims to evaluate all IE risk scores currently available.MethodsStandard methodology (PRISMA guideline) was used. Papers with risk score analysis for IE patients were included, with attention to studies reporting area under the receiver-operating characteristic curve (AUC/ROC). Qualitative analysis was carried out, including assessment of validation processes and comparison of these results to original derivation cohorts where available. Risk-of-bias analysis illustrated according to PROBAST guidelines.ResultsOf 75 articles initially identified, 32 papers were analyzed for a total of 20 proposed scores (range 66–13,000 patients), 14 of which were specific for IE. The number of variables per score ranged from 3 to 14 with only 50% including microbiological variables and 15% including biomarkers. The following scores had good performance (AUC > 0.8) in studies proposing the score (often the derivation cohort); however fared poorly when applied to a new cohort: PALSUSE, DeFeo, ANCLA, RISK-E, EndoSCORE, MELD-XI, COSTA, and SHARPEN. DeFeo score demonstrated the largest discrepancy with initial AUC of 0.88, compared to 0.58 when applied to different cohorts. The inflammatory response in IE has been well documented and CRP has been found to be an independent predictor for worse outcomes. There is ongoing investigation on alternate inflammatory biomarkers which may assist in IE management. Of the scores identified in this review, only three have included a biomarker as a predictor.ConclusionDespite the variety of available scores, their development has been limited by small sample size, retrospective collection of data and short-term outcomes, with lack of external validation, limiting their transportability. Future population studies and large comprehensive registries are required to address this unmet clinical need.

Abstract licence: CC BY

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.