Introduction: Early recognition of Kawasaki disease (KD) relies on identifying hallmark clinical features that overlap with common paediatric febrile and infectious illnesses. In settings with a high burden of infectious disease and limited access to echocardiography, distinguishing KD from other causes of fever at presentation can be challenging. We aimed to identify clinical and laboratory features that differentiate KD from other febrile illnesses in children in a South African setting.
Methods: We conducted a cross-sectional analysis of children admitted to Red Cross Children’s Hospital, Cape Town, with fever and suspected KD between June 2020 and December 2025. Clinical and laboratory features on the day of admission were compared between children with confirmed KD and those with an alternative diagnosis, classified inflammatory controls (IC).
Results: Fifty-seven children were diagnosed with KD and 86 were classified as IC. Among ICs, 75/86 (87%) had infectious diagnoses: confirmed viral (n=15), bacterial (n=11), parasitic (n=1), and clinically diagnosed infections without microbiological confirmation (n=48; e.g. appendicitis, gastroenteritis). Eleven ICs had non-infectious diagnoses: MIS-C (n=4) and other inflammatory conditions (n=7; e.g. systemic juvenile idiopathic arthritis, rheumatic fever, scurvy).
As expected, AHA KD diagnostic criteria strongly predicted KD: rash (OR 6.7), conjunctivitis (OR 6.5), lymphadenopathy (OR 5.0), extremity changes (OR 6.7), and mucositis (OR 8.9) (all p<0.001). Arthritis or arthralgia was also associated with KD (OR 2.6, p=0.02). Vomiting was the only feature that predicted against KD (OR 0.37, p=0.02). Diarrhoea, abdominal pain, shock, and pulmonary or central nervous system involvement did not differentiate KD from IC.
Median admission CRP, sodium, albumin, and platelet counts did not differ between groups. KD patients had higher median white cell (15.2 vs 10.7 x 109/L, p=0.016) and neutrophil counts (9.7 vs 6.2 x109/L, p=0.011). Detection of any organism did not differentiate KD from ICs (OR 0.69, p=0.32).
Discussion: In this setting, detectable infection, gastrointestinal symptoms, shock, or lung or CNS involvement at presentation did not exclude KD. Established clinical criteria and elevated white cell and neutrophil counts increased the likelihood of KD, while vomiting favoured an alternative diagnosis. These findings highlight the need for continued vigilance for KD in high-infectious-burden settings.