Skip to main content

Table 5 Estimated HAP effects on year-specific and year-to-year mean changes in BMI percentile and z-score

From: Healthy apple program to support child care centers to alter nutrition and physical activity practices and improve child weight: a cluster randomized trial

  Baseline Implementation
year 1
Follow-up Implementation
year 2
Year 2011–2012 2012–2013 2013–2014 2014–2015
HAP exposure for CCHP + HAP No Yes No Yes
HAP exposure for CCHP + HAP Delayed No No No Yes
Hierarchical linear model estimates b (SE) P b (SE) P b (SE) P b (SE) P
Year-specific model predicting child BMI percentile, %a
 Time 1.0 (0.7) 0.14 −0.4 (0.7) 0.57 −0.5 (0.7) 0.48 −2.7 (0.8) 0.001
 Time x HAP 0.6 (0.9) 0.49 −1.7 (1.0) 0.08 −2.6 (0.9) 0.003 1.5 (1.0) 0.12
Year-specific model predicting child BMI z-score   
 Time 0.00 (0.02) 0.88 −0.03 (0.03) 0.32 −0.03 (0.03) 0.25 −0.11 (0.03) 0.001
 Time x HAP 0.04 (0.03) 0.18 −0.05 (0.04) 0.17 −0.08 (0.03) 0.007 0.06 (0.04) 0.11
Year-to-year model predicting child care center mean annual change in BMI percentile, %b
 Year Reference   −0.2 (1.8) 0.90 0.01 (1.5) 0.99 −5.6 (2.2) 0.01
 Year x HAP Reference   −4.3 (2.7) 0.11 −5.1 (2.3) 0.03 −2.1 (2.4) 0.37
Year-to-year model predicting child care center mean annual change in BMI z-score
 Year Reference   0.02 (0.06) 0.77 0.02 (0.05) 0.72 −0.2 (0.08) 0.02
 Year x HAP Reference   −0.2 (0.09) 0.06 −0.2 (0.08) 0.01 −0.1 (0.08) 0.18
  1. BMI Body mass index, BMI percentiles and z-scores were calculated relative to the CDC 2000 growth reference; Study time was defined according to academic calendar year. CCHP Child Care Health Program, HAP Healthy Apple Program; Child care centers were randomly allocated to the CCHP + HAP or CCHP + HAP Delayed group in Autumn 2012; b (SE): model coefficient and standard error. P: p-value associated with the model coefficient
  2. aYear-specific hierarchical linear models used child-level data to test for main effects of Time (change from Autumn to Spring during the school year) on BMI percentile and BMI z-score, as well as effect modification by invitation to the HAP (Time x HAP interaction). The year-specific models included child sex, age and overweight or obese status (yes/no) in the Autumn, and type of center (CCHP + HAP or CCHP + HAP Delayed) as predictors of the intercept and CCHP + HAP vs CCHP + HAP Delayed status as predictor of the slope, i.e. annual change from Autumn to Spring. Year-specific model random-effect parameters used an independent covariance structure
  3. bYear-to-year hierarchical linear models used child care center-level data to test if the child care center annual mean changes in BMI percentile and z-score in 2012–2013, 2013–2014 and 2014–2015 differed from the corresponding changes in 2011–2012 (main effects of Year) and if differences from year to year were modified by invitation to the HAP (Year x HAP interaction). Year-to-year models included the child care center mean child age, percent male, and prevalence of overweight or obesity in Autumn as predictors of the intercept, and HAP invitation as predictor of the slope. Year-to-year models specified an unstructured covariance structure