AI-MET: AI Differentiates MIS-C from Endemic Typhus with High Accuracy

Overview

Artificial intelligence was used to develop AI-MET, a two-phase clinical decision support system that distinguishes multisystem inflammatory syndrome in children (MIS-C) from endemic typhus with near-perfect accuracy. Using clinical and laboratory features available within 6 hours of presentation, AI-MET achieved 100% accuracy in the initial cohort and 99% accuracy in a validation cohort.

Background

MIS-C is a serious post-COVID-19 inflammatory condition in children that shares clinical and laboratory features with endemic typhus, a flea-borne rickettsial infection common in tropical and subtropical regions. Both diseases present with fever, rash, gastrointestinal symptoms, and laboratory abnormalities such as thrombocytopenia and elevated inflammatory markers, complicating diagnosis. Early and accurate differentiation is critical because treatments differ significantly: immunomodulation for MIS-C and tetracycline antibiotics for typhus. Traditional serologic testing for typhus has delayed turnaround times, limiting its utility at presentation.

Data Highlights

Key Findings

• AI-MET uses a two-phase approach: Phase 1 with 17 features (MET-17) and Phase 2 adding 13 more features (MET-30) if confidence is low.
• Although 24 of 30 features differed statistically between MIS-C and typhus, individual features overlapped too much to be clinically useful alone.
• AI-MET achieved 100% accuracy in distinguishing MIS-C from typhus in the initial cohort of 220 patients.
• In a separate validation cohort of 111 MIS-C patients, AI-MET maintained 99% classification accuracy.
• The model relies on clinical and laboratory data rapidly available within 6 hours of presentation, facilitating timely decision-making.
• Early differentiation enables appropriate treatment initiation: immunomodulation for MIS-C and tetracycline antibiotics for typhus.

Clinical Implications

AI-MET provides frontline clinicians in endemic regions with a rapid, reliable tool to differentiate MIS-C from typhus, conditions with overlapping presentations but divergent treatments. Utilizing readily available clinical and laboratory data, AI-MET can guide early therapeutic decisions, potentially improving outcomes. This approach may reduce diagnostic uncertainty and reliance on delayed serologic testing.

Conclusion

The AI-MET decision support system demonstrates exceptional accuracy in distinguishing MIS-C from endemic typhus using early clinical data, addressing a critical diagnostic challenge in pediatric febrile illnesses. Its implementation could enhance timely, appropriate management in endemic areas.

References

1. Centers for Disease Control and Prevention 2020 -- MIS-C Case Definition
2. World Health Organization 2020 -- Multisystem Inflammatory Syndrome in Children and Adolescents Temporally Related to COVID-19
3. Original Article Authors 2024 -- Utilizing Artificial Intelligence to Differentiate Between Multisystem Inflammatory Syndrome in Children and Endemic Typhus: A Comparative Study of MIS-C and Typhus (AI-MET)