BackgroundIn many studies, it is of interest to identify population subgroups that are relatively homogeneous with respect to an outcome. The nature of these subgroups can provide insight into effect mechanisms and suggest targets for tailored interventions. However, identifying relevant subgroups can be challenging with standard statistical methods.Main textWe review the literature on decision trees, a family of techniques for partitioning the population, on the basis of covariates, into distinct subgroups who share similar values of an outcome variable. We compare two decision tree methods, the popular Classification and Regression tree (CART) technique and the newer Conditional Inference tree (CTree) technique, assessing their performance in a simulation study and using data from the Box Lunch Study, a randomized controlled trial of a portion size intervention. Both CART and CTree identify homogeneous population subgroups and offer improved prediction accuracy relative to regression-based approaches when subgroups are truly present in the data. An important distinction between CART and CTree is that the latter uses a formal statistical hypothesis testing framework in building decision trees, which simplifies the process of identifying and interpreting the final tree model. We also introduce a novel way to visualize the subgroups defined by decision trees. Our novel graphical visualization provides a more scientifically meaningful characterization of the subgroups identified by decision trees.ConclusionsDecision trees are a useful tool for identifying homogeneous subgroups defined by combinations of individual characteristics. While all decision tree techniques generate subgroups, we advocate the use of the newer CTree technique due to its simplicity and ease of interpretation.Electronic supplementary materialThe online version of this article (doi:10.1186/s12982-017-0064-4) contains supplementary material, which is available to authorized users.

Objectives. To evaluate a multicomponent obesity prevention intervention among diverse, low-income preschoolers. Methods. Parent–child dyads (n = 534) were randomized to the Now Everybody Together for Amazing and Healthful Kids (NET-Works) intervention or usual care in Minneapolis, MN (2012–2017). The intervention consisted of home visits, parenting classes, and telephone check-ins. The primary outcomes were adjusted 24- and 36-month body mass index (BMI). Results. Compared with usual care, the NET-Works intervention showed no significant difference in BMI change at 24 (–0.12 kg/m2; 95% confidence interval [CI] = −0.44, 0.19) or 36 months (–0.19 kg/m2; 95% CI = −0.64, 0.26). Energy intake was significantly lower in the NET-Works group at 24 (–90 kcal/day; 95% CI = −164, −16) and 36 months (–101 kcal/day; 95% CI = −164, −37). Television viewing was significantly lower in the NET-Works group at 24 (rate ratio = 0.84; 95% CI = 0.75, 0.93) and 36 months (rate ratio = 0.88; 95% CI = 0.78, 0.99). Children with baseline overweight or obesity had lower BMI in the NET-Works group than those in usual care at 36 months (–0.71 kg/m2; 95% CI = −1.30, −0.12). Hispanic children had lower BMI in the NET-Works group than those in usual care at 36 months (–0.59 kg/m2; 95% CI = −1.14, −0.04). Conclusions. In secondary analyses, NET-Works significantly reduced BMI over 3 years among Hispanic children and children with baseline overweight or obesity. Trial Registration: ClinicalTrials.gov Identifier: NCT01606891.

Introduction In the U.S., children from low-income families are more likely to be obese. The impact of parent modeling of physical activity (PA) and sedentary behaviors in low-income American ethnic minorities is unclear, and studies examining objective measures of preschooler and parent PA are sparse. Methods This cross-sectional study examined 1,003 parent–child pairs who were of low income, largely Latino and African American, and living in one of two geographically disparate metropolitan areas in the U.S. Parents and children wore GT3X/GT3X+ accelerometers for an average of >12 hours/day (7:00am–9:00pm) for 1 week (September 2012 to May 2014). Analysis occurred in 2015–2016. Results About 75% of children were Latino and >10% were African American. Mean child age was 3.9 years. The majority of children (60%) were normal weight (BMI ≥50th and <85th percentiles), with more than a third were overweight/obese. Children's total PA was 6.03 hours/day, with 1.5 hours spent in moderate to vigorous PA (MVPA). Covariate-adjusted models showed a monotonic, positive association between parent and child minutes of sedentary behavior (β=0.10, 95% CI=0.06, 0.15) and light PA (β=0.06; 95% CI=0.03, 0.09). Child and parent MVPA were positively associated up to 40 minutes/day of parent MVPA, but an inverse association was observed when parental MVPA was beyond 40 minutes/day (p=0.002). Conclusions Increasing parental PA and reducing sedentary behavior correlate with increased PA-related behaviors in children. However, more work is needed to understand the impact of high levels of parental MVPA on the MVPA levels of their children.

Behavioral interventions for pediatric obesity are promising, but detailed information on treatment fidelity (i.e., design, training, delivery, receipt, and enactment) is needed to optimize the implementation of more effective interventions. Little is known about current practices for reporting treatment fidelity in pediatric obesity studies. This systematic review, in accordance with PRISMA guidelines, describes the methods used to report treatment fidelity in randomized controlled trials. Treatment fidelity was double-coded using the NIH Fidelity Framework checklist. Three hundred articles (N=193 studies) were included. Mean inter-coder reliability across items was 0.83 (SD=0.09). Reporting of treatment design elements within the field was high (e.g., 77% of studies reported designed length of treatment session), but reporting of other domains was low (e.g., only 7% of studies reported length of treatment sessions delivered). Few reported gold standard methods to evaluate treatment fidelity (e.g., coding treatment content delivered). General study quality was associated with reporting of treatment fidelity (p<0.01) as was the number of articles published for a given study (p<0.01). The frequency of reporting treatment fidelity components has not improved over time (p=0.26). Specific recommendations are made to support pediatric obesity researchers in leading health behavior disciplines toward more rigorous measurement and reporting of treatment fidelity.