Evaluation and prediction of individual growth trajectories

Abstract

Background

Conventional growth charts offer limited guidance to track individual growth.

Aim

To explore new approaches to improve the evaluation and prediction of individual growth trajectories.

Subjects and methods

We generalise the conditional SDS gain to multiple historical measurements, using the Cole correlation model to find correlations at exact ages, the sweep operator to find regression weights and a specified longitudinal reference. We explain the various steps of the methodology and validate and demonstrate the method using empirical data from the SMOCC study with 1985 children measured during ten visits at ages 0–2 years.

Results

The method performs according to statistical theory. We apply the method to estimate the referral rates for a given screening policy. We visualise the child’s trajectory as an adaptive growth chart featuring two new graphical elements: amplitude (for evaluation) and flag (for prediction). The relevant calculations take about 1 millisecond per child.

Conclusion

Longitudinal references capture the dynamic nature of child growth. The adaptive growth chart for individual monitoring works with exact ages, corrects for regression to the mean, has a known distribution at any pair of ages and is fast. We recommend the method for evaluating and predicting individual child growth.

Keywords:

• Conditional SDS gain
• correlation model
• longitudinal growth reference
• multiple testing
• adaptive growth chart

Acknowledgements

The author wishes to thank the collaborators from the Child Health Clinics of the SMOCC cohort study for their contributions to the data collection.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data used in this study are not public. The point of contact for the data is dr. Paul Verkerk (email: [email protected]).

Additional information

Funding

This study was financially supported by the Prevention, Work and Health research program, TNO, Leiden.