Development and Validation of a Nomogram-Based Prognostic Model to Predict High Blood Pressure in Children and Adolescents-Findings From 342,736 Individuals in China

Abstract

Objectives: Predicting the potential risk factors of high blood pressure (HBP) among children and adolescents is still a knowledge gap. Our study aimed to establish and validate a nomogram-based model for identifying youths at risk of developing HBP.

Methods: HBP was defined as systolic blood pressure or diastolic blood pressure above the 95th percentile, using age, gender, and height-specific cut-off points. Penalized regression with Lasso was used to identify the strongest predictors of HBP. Internal validation was conducted by a 5-fold cross-validation and bootstrapping approach. The predictive variables and the advanced nomogram plot were identified by conducting univariate and multivariate logistic regression analyses. A nomogram was constructed by a training group comprised of 239,546 (69.9%) participants and subsequently validated by an external group with 103,190 (30.1%) participants.

Results: Of 342,736 children and adolescents, 55,480 (16.2%) youths were identified with HBP with mean age 11.51 ± 1.45 years and 183,487 were boys (53.5%). Nine significant relevant predictors were identified including: age, gender, weight status, birth weight, breastfeeding, gestational hypertension, family history of obesity and hypertension, and physical activity. Acceptable discrimination [area under the receiver operating characteristic curve (AUC): 0.742 (development group), 0.740 (validation group)] and good calibration (Hosmer and Lemeshow statistics, P > 0.05) were observed in our models. An available web-based nomogram was built online on https://hbpnomogram.shinyapps.io/Dyn_Nomo_HBP/.

Conclusions: This model composed of age, gender, early life factors, family history of the disease, and lifestyle factors may predict the risk of HBP among youths, which has developed a promising nomogram that may aid in more accurately identifying HBP among youths in primary care.

Keywords: children and adolescents; cross-sectional study; high blood pressure; nomogram; risk classification.

Figure 1

Predictor selection using the Lasso binary logistic regression model. A Lasso coefficient of the total 14 predictors. (A) Lasso coefficient profiles of all predictors, a coefficient profile plot was provided against the log (Lambda) sequence. (B) Predictors selection by Lasso via minimum criteria, predictor selection in the Lasso model used 10-fold cross-validation via minimum criteria. Red-dotted vertical lines were drawn at the optimal values by using the minimum criteria (minimizing the mean-squared error), the value 9 represents that those 14 predictors were reduced to 9 non-zero features by Lasso.

Figure 2

Clinical nomogram for predicting probability of developing HBP among children and adolescents, and its predictive performance. To use the nomogram, an individual HBP contact's values are located on each variable axis, and a line is drawn downward to the risk of HBP axes to detect the hypertension probability. As an example of how this nomogram can be calculated, we can take an 18-year-old obese boy who was bottle-fed in early life, with gestational hypertension, family history of hypertension and obesity, and <1-h physical activity expenditure. By drawing a line up toward the points for each of the variables this student will have 100 points (age), 32 points (gender), 74 points (BMI status), 42 points (feeding mode), 20 points (gestational hypertension), 21 points (family obesity), 59 points (family hypertension), and 26 points (physical activity), giving a total of 374 points (at the bottom of the figure), and a probability of HBP of 80%.

Figure 3

Receiver operating characteristic (ROC) curves for the prediction of high blood pressure in the training group and validation group. (A) ROC curves of the factors and nomogram in the development group; (B) ROC curves of the factors and nomogram in the training group; (C) calibration plot of nomogram prediction in the development group; (D) calibration plot of nomogram prediction in the validation group. ROC curves from the prediction model and other predictive strategies (Model 1: Age, gender, gestational hypertension, weight status, family history of hypertension, family history of obesity, and average outdoor physical activity time; Model 2: Age, gender, gestational hypertension, weight status, family history of hypertension, family history of obesity, average outdoor physical activity time, birth weight, and feeding mode; and Model 3: Age, gender, gestational hypertension, weight status, family history of hypertension, family history of obesity, average outdoor physical activity time, birth weight, feeding mode, parental smoking status, parental education level, household monthly income, average screen-based time, and fried food intake) for comparison. The calibration curve represents the calibration of the nomogram, which shows the consistency between the predicted probability of conversion and actual conversion probability of HBP patients. The x-axis is the predicted probability by the nomogram and the y-axis is the actual conversion rate of HBP patients. The gray line represents a perfect prediction by an ideal model, and the black-dotted line shows the performance of the nomogram, of which a closer fit to the gray line means a better prediction.