The utility of three predictors of childhood myopia: a Bayesian analysis.

Any treatment to prevent the onset of juvenile myopia will require predictive tests in order to determine which children should receive treatment. Three risk factors for myopia were evaluated for their ability to predict myopia: (a) refraction at school entry; (b) refraction in infancy; and (c) parental history of myopia. Bayes' theorem was used to estimate these conditional probabilities. Refraction at school entry had twice the power to predict myopia (probability of juvenile myopia given the child is near emmetropia at school entry = 0.53) compared to either infant refraction (0.21-0.28) or parental myopia (0.20-0.25). While a history of any parent having myopia had the highest test sensitivity (probability of a positive family history of myopia given juvenile myopia in the child = 0.90) and refraction at school entry the highest test specificity (probability of more hyperopia than +0.50 D at school entry given no juvenile myopia = 0.91), none of these three factors had high values for both sensitivity and specificity. Further work is required to develop a battery of tests which could predict the onset of juvenile myopia with both adequate sensitivity and specificity.