Assessing the Validity of Childhood Blood Pressure Screening: Unbiased Estimates of Sensitivity, Specificity, and Predictive Values

Abstract

Blood pressure measurement in childhood can be considered as a screening test for future blood pressure levels. Evaluating this test involves calculating sensitivities, specificities, and predictive values for a blood pressure measurement at an initial time for predicting underlying true blood pressure at a subsequent time. We demonstrate the use of childhood blood pressure tracking correlations that are corrected for within-person variability to calculate unbiased estimates of these test characteristics. In a cohort of 333 schoolchildren, we measured blood pressure on multiple visits in each of 4 successive years. Using these data for within-person variances and corrected tracking correlations, and U.S. population data for means and total variances, we tabulated positive predictive values, sensitivities, and specificities for the case of predicting a 9-year-old male's true systolic blood pressure 3 years later. For example, if a 9-year-old's average blood pressure from 4 visits is 115 mmHg, the probability of his true blood pressure being >116 mmHg (90th percentile) at age 12 is 0.50. With longer follow-up, the use of predictive values, sensitivities, and specificities that incorporate corrected correlations should allow determination of the accuracy of predicting adult blood pressure from childhood values, and therefore the usefulness of screening children for those at high risk of developing hypertension. (Epidemiology 1992;3:40-46)