Algorithms to Define Abnormal Growth in Children: External Validation and Head-To-Head Comparison

Background: Growth monitoring of apparently healthy children aims at early detection of serious conditions by use of both clinical expertise and algorithms that define abnormal growth. The seven existing algorithms provide contradictory definitions of growth abnormality and have a low level of validation. Objective: An external validation study with head-to-head comparison of the seven algorithms combined with study of the impact of use of the World Health Organization (WHO) vs national growth charts on algorithm performance. Design: With a case-referent approach, we retrospectively applied all algorithms to growth data for children with Turner syndrome, GH deficiency, or celiac disease (n = 341) as well as apparently healthy children (n = 3406). Sensitivity, specificity, and theoretical reduction in time to diagnosis for each algorithm were calculated for each condition by using the WHO or national growth charts. Results: Among the two algorithms with high specificity (>98%), the Grote clinical decision rule had higher sensitivity than the Coventry consensus (4.6% to 54% vs 0% to 8.9%, P < 0.05) and offered better theoretical reduction in time to diagnosis (median: 0.0 to 0.9 years vs 0 years, P< 0.05). Sensitivity values were significantly higher with the WHO than national growth charts at the expense of specificity. Conclusion: The Grote clinical decision rule had the best performance for early detection of the three studied diseases, but its limited potential for reducing time to diagnosis suggests the need for better-performing algorithms based on appropriate growth charts.