BACKGROUND:

Blood pressure measurements in childhood and young adulthood are less predictive of future levels than those taken in middle age. In part, this may be due to the fact that within-person variability appears to make up a larger proportion of total variability in childhood than adulthood. Previous work on an NHLBI supported grant indicated that repeated blood pressure measurements and visits led to higher childhood tracking correlations over a period of three years.

DESIGN NARRATIVE:

Follow-up data were used as well as multiple visits which reduced the large within-person variability of blood pressure measurements and improved the tracking correlations. In addition, 'true' or 'corrected' tracking correlations were provided by eliminating the effects of random measurement error. The effects were examined of time-varying covariates on both the observed and true tracking correlations. Besides computing tracking correlations, predictive values were computed for young adult blood pressure given childhood levels. This was the probability that a young adult's true blood pressure was above a specific cutpoint conditional on childhood blood pressure. These values were validated using data from the Fels Longitudinal Study, which included serial blood pressure measurements over the age range in the study. The prediction models were also derived including terms for covariates such as age, sex, height and weight. From these models nomograms were constructed which were useful to physicians for prognostic purposes. Thus, because of the unique multiple-visit approach used in these data, the effect of random measurement error was eliminated. These methodologic improvements strengthened the usefulness of blood pressure screening in childhood to detect those at high risk of developing hypertension.