A quick and qualitative assessment of gross motor development in preschool children

Abstract

There is a need for a quick, qualitative, reliable, and easy tool to assess gross motor development for practitioners. The aim of this cross-sectional study is to present the Zurich Neuromotor Assessment-Q (ZNA-Q), which assesses static and dynamic balance in children between 3 and 6 years of age in less than 5 min. A total of 216 children (103 boys; 113 girls; median age 4 years, 4 months; interquartile range 1 year, 3 months) were enrolled from day-care centers, kindergartens, and schools, and were tested with 5 different gross motor tasks: standing on one leg, tandem stance, hopping on one leg, walking on a straight line, and jumping sideways. All ordinal measures (consisting of qualitative measures and scales) featured a marked developmental trend and substantial inter-individual variability. Test-retest reliability was assessed on 37 children. It varied from .17 for tandem stance to .43 for jumping sideways for the individual tasks, and it was .41 and .67 for the static and dynamic balance components, respectively. For the whole ZNA-Q, test-retest reliability was .7.

Conclusion: Ordinal scales enable practitioners to gather data on children’s gross motor development in a fast and uncomplicated way. It offers the practitioner with an instrument for the exploration of the current developmental motor status of the child.

What is Known: • Measurement of gross motor skills in the transitional period between motor mile stones and quantitative assessments is difficult. • Assessment of gross motor skills is relatively easy.

What is New: • Supplementary and quick gross motor test battery for children for practitioners. • Normative values of five gross motor skills measured with ordinal scales.

Statistical appendix

Multinomial regression model

A multinomial regression was used to model observed ordinal scores for each task separately. More specifically and for any particular task, the random variable Y_i describing the ordinal score of child i (whose observed score is y_i) was modeled as a function of linear age and sex, such that

$$ \log \left(\frac{p_i(k)}{1-{p}_i(k)}\right)={\beta}_{0k}+{\beta}_{1k}{\mathrm{age}}_i+{\beta}_{2k}{\mathrm{sex}}_i\kern5.25em k\in \left\{0,1,2,3,4\right\} $$

where

$$ {p}_i(k)=P\left({Y}_i\le k|{\mathrm{age}}_i,{\mathrm{sex}}_i\right) $$

(1)

refers to the (cumulative) probability to obtain a score smaller or equal to k given the age and sex of the child, and β_0k, β_1k, and β_2k are regression coefficients that depend on the score level k. Constrained versions of this model that contain fewer parameters were also fitted for comparative purposes. For example, a partial proportional odds model where the effect of sex is assumed constant across levels was fitted by letting β_2k = β₂ for all k. A full proportional odds model where the effect of both age and sex is constant across levels was also fitted by letting β_1k = β₁ and β_2k = β₂ for all k. Goodness-of-fit was assessed using a chi-square test by comparing observed and predicted frequencies in each ordinal category for different age groups (3–4, 4–5, 5–6 years) and sex.

Visualizing the developmental trend

The developmental trend was visualized by plotting the expected ordinal score $\widehat{S}\left(\mathrm{age},\mathrm{se}x\right)=\sum \limits_{k=0}^4k\cdot \widehat{P}\left(Y=k|\mathrm{age},\mathrm{se}\mathrm{x}\right)$ as a function of age and sex, where $\widehat{P}\left(Y=k|\mathrm{age},\mathrm{sex}\right)$ is the predicted probability (derived from the model) of obtaining a score k given age and sex. Additionally, the estimated cumulative probability $\widehat{P}\left(Y\le k|\mathrm{age},\mathrm{sex}\right)=\sum \limits_{j=0}^k\widehat{P}\left(Y=j|\mathrm{age},\mathrm{sex}\right)$ was also plotted as a function of age and sex.

Calculation of standard deviation scores (SDS)

For each particular task, an unbiased estimate of the percentile ${\widehat{p}}_i^{\ast }$ associated with the observed score y_i of child i was calculated as

$$ {\widehat{p}}_i^{\ast }=\Big\{{\displaystyle \begin{array}{c}1-\frac{{\widehat{p}}_i\left({y}_i\right)}{2}\kern9.25em {y}_i=0\\ {}1-\frac{{\widehat{p}}_i\left({y}_i-1\right)+{\widehat{p}}_i\left({y}_i\right)}{2}\kern4.25em 0<{y}_i\le 4\end{array}} $$

with ${\widehat{p}}_i(k)$ defined in Eq. (1). This percentile was then converted into a standard deviation score (SDS) ${\widehat{z}}_i={\varPhi}^{-1}\left({\widehat{p}}_i^{\ast}\right)$ with Φ⁻¹ denoting the standard normal cumulative distribution function.