Skip to main content

Table 3 Adjusted 6–18 months outcomes by randomized maintenance assignment in the SIPsmartER condition (n = 120)

From: Supporting maintenance of sugar-sweetened beverage reduction using automated versus live telephone support: findings from a randomized control trial

  Change 6 to 18 monthsa Relative 6 to 18 month effects between conditionsa,b
Outcome IVR behavior n = 47 Human-delivered behavior n = 46 IVR control n = 27 IVR behavior vs. IVR control Human-delivered behavior vs. IVR control IVR behavior vs. Human-delivered behavior
SSB, kcals −12.1 (− 65, 41) 15.6 (− 47, 78) 86 (− 19, 192) − 98.5 (− 196, − 0.55)* −70 (− 209, 64) 27.7 (− 69.6, 125.2)
BMI, kg/m2 0.11 (−0.28, 0.51) 0.07 (− 57, .70) − 0.07 (− 0.60, 0.46) 0.18 (− 0.37, 0.74) 0.13 (− 0.49, 0.76) −0.05 (− 0.80, 0.70)
Weight, kg 0.04 (− 0.97, 1.1) −0.17 (− 1.4, 1.13) 0.36 (− 1.7, 2.44) −0.32 (− 2.7, 2.1) −0.53 (− 2.9, 1.88) −0.21 (− 1.9, 1.47)
QOL, # of unhealthy days 0.6 (−1.7, 2.9) − 0.7 (− 3.7, 2.3) 2.4 (− 0.6, 5.4) −1.8 (− 5.1, 1.4) −3.1 (− 8.3, 2.1) −1.3 (− 5.7, 3.2)
  1. Within condition and between condition statistical significance indicated by asterisks: *p < 0.05
  2. SSB=Sugar-sweetened beverages, BMI=Body Mass Index, QOL = Quality of Life
  3. aValues reported as mean change (95% Confidence Interval). All Δ change scores represent values from the first (earliest) time point subtracted from later time point
  4. bCondition dummy variable coded as interactive voice response (IVR) behavior support vs. IVR control condition; human-delivered behavior support vs. IVR control condition; IVR behavior support vs. human-delivered behavior support. Models controlled for baseline covariates including age, gender, race/ethnicity, income, education level, health literacy level, employment status, number of children, smoking status, and BMI. The 95% confidence intervals are also adjusted to be cohort robust. Analytic procedures use intention-to-treat last observation carried forward imputations and empirical models are multi-level mixed effect models