Prediction of excess pregnancy weight gain using psychological, physical, and social predictors: A validated model in a prospective cohort study
Journal Articles
Overview
Research
Identity
Additional Document Info
View All
Overview
abstract
OBJECTIVE: To develop and validate a prediction model for excess pregnancy weight gain using early pregnancy factors. DESIGN: Prospective cohort study. SETTING: We recruited from 12 obstetrical, family medicine, and midwifery centers in Ontario, Canada. PARTICIPANTS: We recruited English-speaking women with singleton pregnancies between 8+0-20+6 weeks. Of 1296 women approached, 1050 were recruited (81%). Of those, 970 women had complete data (970/1050, 92%) and were recruited at a mean of 14.8 weeks. PRIMARY OUTCOME MEASURE: We collected data on psychological, physical, and social factors and used stepwise logistic regression analysis to develop a multivariable model predicting our primary outcome of excess pregnancy weight gain, with random selection of 2/3 of women for training data and 1/3 for testing data. RESULTS: Nine variables were included in the final model to predict excess pregnancy weight gain. These included nulliparity, being overweight, planning excessive gain, eating in front of a screen, low self-efficacy regarding pregnancy weight gain, thinking family or friends believe pregnant women should eat twice as much as before pregnancy, being agreeable, and having emotion control difficulties. Training and testing data yielded areas under the receiver operating characteristic curve of 0.76 (95% confidence interval, 0.72 to 0.80) and 0.62 (95% confidence interval 0.56 to 0.68), respectively. CONCLUSIONS: In this first validated prediction model in early pregnancy, we found that nine psychological, physical, and social factors moderately predicted excess pregnancy weight gain in the final model. This research highlights the importance of several predictors, including relatively easily modifiable ones such as appropriate weight gain plans and mindfulness during eating, and lays an important methodological foundation for other future prediction models.
