Neonatal Percentile Curves: A Multivariate Normal Probability Density Approach

Author(s): Michelle Montopoli M.A, EMT*, George Montopoli Ph.D, EMT, William ’Will’ Smith MD, EMT-P, Delia Montopoli CNM, NP

Background and Objectives: Percentile growth charts have been known as the standard to measure neonatal and child growth. However, there are several problems users face when using and analyzing percentile charts. Current univariate growth charts are generalized to global populations. To optimally identify highrisk neonates, we outline a multivariate approach. We first hypothesize that morphometrics will vary according to demographics such as location, gender, race, etc. Further, we propose creating percentile charts involving any number of growth parameters for identifying high-risk individuals for specific locations and populations based on the Multivariate Normal Probability Density.

Methods: We obtained data from neonates (38 to 42 weeks) involving four morphometrics (body length, chest circumference, cephalic perimeter and weight) from three locations: Phoenix and Yuma, AZ, and Jackson, WY. We investigated whether neonatal morphometrics differed significantly with respect to location, gender, and race; and that assumptions of the multivariate approach were met, justifying the procedure. We then applied the multivariate approach for different combinations of morphometrics for specific populations. Ten scenarios were designed to evaluate and compare percentile computations for different demographics and morphometrics.

Results: Neonatal morphometrics varied significantly for different genders, races and locations. Morphometric data presented no serious deviation from normality and assumptions of the multivariate approach were supported. The analysis of different combinations of the four morphometrics for Yuma Hispanics demonstrated the importance of our procedure in identifying high-risk neonates over the current univariate charts. MANOVAs, ANOVAs, and Independent- samples t-Tests generally demonstrated that morphometric data varied for different populations based on demographics: Location effect was significant on Body Length and Weight; Gender effect was significant on Body Length and Weight; Race effect was significant on Body Length and Weight. Location did not significantly affect Cephalic Perimeter, while Gender and Race did. Two-variable percentile curves were constructed and percentiles (for more than two variables) were calculated for various scenarios and compared to conventional charts.

Conclusions: Demographic differences demonstrate that the multivariate percentile approach may better identify high-risk individuals because percentile calculations involve more morphometric information and the multivariate procedure accounts for inter-correlations. Specific locations throughout the world could potentially utilize our approach for global validation for more reliable identification of highrisk neonates. Furthermore, this user-friendly approach could be used in a multitude of scenarios involving morphometrics for any given population. It can also be used to study and understand current national trends and compare how neonatal growth has changed, showing greater need for a new and more accurate percentile- curve model.