A Smartphone-Based Predictor of Hypokalemia in Pediatric Acute Gastroenteritis: A Pilot Diagnostic Accuracy Study

Abstract. Acute gastroenteritis is a common cause of pediatric admission in low- and middle-income settings, where serum potassium results may not be available when initial treatment decisions are made. This pilot study evaluated the feasibility and preliminary diagnostic accuracy of the SE-PO Serum Potassium Predictor, an Android application that converts bedside clinical and electrocardiographic findings into a predicted serum potassium value. A cross-sectional diagnostic accuracy design was used among 530 children aged five months to 17 years admitted for acute gastroenteritis at a tertiary government hospital in the southern Philippines from February to September 2017. The bedside assessment was completed before intravenous fluids were started and before laboratory results were released. Venous serum potassium served as the reference standard, with hypokalemia defined as less than 3.5 mEq/L. Hypokalemia was confirmed in 267 children (50.4%). The application showed 92.1% sensitivity, 97.3% specificity, a positive likelihood ratio of 34.62, a negative likelihood ratio of 0.08, and an area under the receiver operating characteristic curve of 0.988. Bland-Altman analysis showed a mean underestimation of 0.18 mEq/L. Detection was 100% for moderate and severe hypokalemia but 76.1% for mild hypokalemia, while specificity decreased to 81.5% among children with hypovolemic shock. Bedside use was feasible within the admission workflow. These findings are promising but should be interpreted as single-center pilot estimates because the model was evaluated in the same institution in which it was developed. Laboratory measurement must remain the basis for potassium correction, and prospective external validation is required before clinical adoption. Keywords: diagnostic accuracy; hypokalemia; mHealth; pediatric acute gastroenteritis; point-of-care prediction