38 - EVALUATION OF THE 2017 AHA INCOMPLETE KD DIAGNOSTIC ALGORITHM AND A DATA-DRIVEN PROPOSAL OF CHANGES

Background: The 2017 AHA algorithm for suspected incomplete KD was based on expert opinion and has not been validated due to lack of gold standard test for KD. We tested the performance of the algorithm in a contemporary cohort of incomplete KD patients with coronary artery aneurysms (CAA).


Methods: Using CAA (maximum age-adjusted Z-score of left anterior descending or right coronary artery diameter ≥ 2.5) as the gold standard for KD, we reviewed clinical data from KD patients diagnosed within 10 days from fever onset with < 4 clinical features and CAA. We included 53 KD patients and 53 febrile controls (FC) diagnosed between 2001-2022 (Table 1), who were propensity-score matched based on age, sex, and ethnicity. Using the 2017 AHA algorithm that requires ≥ 5 days of fever, 13 KD and 13 FC patients diagnosed before day 5 of illness were excluded. The diagnostic performance of the whole algorithm and each individual step (Table 2) was measured by area under curve (AUC), sensitivity, specificity, and positive and negative predictive value (PPV, NPV). In addition, we determined the optimal number of fever days for patients with 2-3 KD-associated physical signs and for infants ≤ 6 months with unexplained fever, as well as the optimal cutoffs for the six lab tests in the algorithm by grid search that exhaustively evaluated the diagnostic performance of over 2 million pre-defined combinations of lab cut-offs.


Results: Overall, the 2017 AHA algorithm was not sensitive as a screening tool for incomplete KD with CAA (sensitivity 77.5%, specificity 92.5%, PPV 91.2%, NPV 80.4%, AUC 0.850). Combined CRP and ESR was the most sensitive (91.2%) single step (Table 2). The KD diagnosis of 19.3% (17/88) of patients would have been delayed by the current 5-day fever criterion. Shortening the days of fever requirement to 3 days for all infants and children improved the diagnostic performance (sensitivity 90.2%, specificity 96.1%, PPV 95.8%, NPV 90.7%, AUC 0.931) (Figure 1). Intriguingly, the expert-suggested lab cutoffs in the 2017 AHA algorithm were consistent with the optimal cutoffs from the computer-based grid search.


Conclusion: Based on these findings, we propose shortening the required days of fever to 3 days for evaluation of patients with 2-3 KD-related physical signs and for infants with unexplained fever. These findings support earlier diagnosis and treatment for this group of incomplete KD patients with the high risk of long-term morbidity and mortality.