Artificial intelligence tools in sepsis prediction: a systematic review and meta-analysis

Authors/institutions/venue: Aadith Ashok மற்றும் co-authors; Alfred Health, Monash University மற்றும் collaborators; npj Digital Medicine; 31 August; peer reviewed.

Problem/contribution: 53 studies, ஏழு மில்லியனுக்கும் அதிக admissions. Authors pooled AUROC 0.88, sensitivity 77.2%, specificity 84.7%, PPV 34.2% என report செய்கின்றனர். Significance: promising discrimination; low PPV alert burden உருவாக்கலாம். Limitations: I² >95% heterogeneity, bias, patient-outcome improvement நிரூபிக்கப்படவில்லை; independent deployment validation இல்லை. Nature paper