How an Online Intervention to Prevent Excessive Gestational Weight Gain Is Used and by Whom: A Randomized Controlled Process Evaluation

[view this table]Table 1. Sample characteristics.

Data Collection

Five sources of data were used in this analysis: screening for eligibility, postpartum height and weight visit, medical chart audit, website activity, and survey. At baseline screening, which took place at less than 20 weeks gestation, the following self-reported variables were collected: race, ethnicity, date of birth, height, current weight, early pregnancy (13 weeks) or pre-pregnancy weight, and a measure of low-income using a participant’s insurance type to determine if a participant qualified for Women, Infants and Children Program (WIC)/ Medicaid/ Prenatal Care Assistance Program (PCAP). The self-reported ethnicity question asked, “Are you of Hispanic or Latino origin?” with the response categories of “yes” and “no”. The self-reported race item asked, “Which race best describes you?” with the response categories of “American Indian and Alaska Native”, “Asian”, “Black or African American”, “White”, “Native Hawaiian and other Pacific Islander”, and “Other race, please specify”.

To categorize women’s weight status, measured weight and height were used to calculate BMI for the vast majority of the sample. Height was collected from three data sources: (1) measured height from postpartum weight collection visits (1307/1689, 77.38% of sample), (2) measured height from the medical chart (358/1689, 21.20% of sample), or (3) self-reported height at screening (24/1689, 1.42% of sample). Pre-pregnancy or early pregnancy weight was collected from three data sources: (1) measured early pregnancy weight from the medical chart (1599/1689, 94.67%), (2) self-reported or measured pre-pregnancy weight from the medical chart (67/1689, 3.97%), or (3) self-reported pre-pregnancy or early pregnancy weight at screening (23/1689, 1.36% of sample).

Medical chart audit data were used to verify and correct the date of birth of the participant (33/1689 individuals with changed date of births, 1.95% of total study sample). Date of birth and date of consent were used to calculate age of the subject at time of entry into the study.

Each participant’s online activities were continuously collected throughout the study automatically by the website. Each website activity was time stamped and only activities from consent date to delivery date were included in this analysis. All activities associated with intervention features, rather than data collection activities such as surveys, were considered intervention use in this analysis.

All randomized participants were asked to complete a baseline questionnaire. The questionnaire was available online and via telephone from consent date and until the participant was greater than 28 weeks into pregnancy. A survey item that asked about home Internet use was included in this research.

Conceptualizing Measures of Engagement

Use of the following six intervention features were used to characterize engagement: health-related information (articles and FAQs), blogs, local resources, diet goal-setting tools, physical activity goal-setting tools, and a weight-gain tracker. Features were categorized based on expected use. Consistent use was expected for log-ins and entry of weights into the weight gain tracker. We expected women to track their weight in 30-day intervals but, to allow for difference in timing of doctor’s visits, we created 45-day intervals from time of enrollment to delivery. If a woman entered a weight during each of the 45-day intervals that she completed, she was categorized as a “consistent tracker”. If during at least of half of the intervals a woman entered a weight, she was categorized as an “almost consistent tracker”. If a woman had entered at least one weight but not during more than half of her intervals, she was categorized as an “inconsistent tracker”. Finally, if she never entered a weight during pregnancy, she was categorized as a “non-weight tracker”. The same procedure was used to categorize use for log-ins.

For all other features, consistent use was not expected. Use was expected on an “as needed” basis. Therefore, quantity of use defined engagement for the following features: health-related information, blogs, resources, diet goal-setting tools, and physical activity goal-setting tools. A woman’s engagement was categorized into three levels for each of these features: “high” (≥median among users), “low” (median among users), or “never” (0).

Demographic Subgroups

Since sociodemographic characteristics are correlated and most sociodemographic characteristics are measured categorically, several recent studies have employed latent class or subgroup analysis to group women with similar characteristics together [15,17,18,24]. Demographic/BMI subgroups were created in the analysis sample (n=1014) based on the following variables: race (white, black, or other), ethnicity (Hispanic or non-Hispanic), low-income status (185% poverty line or ≥185% poverty line), BMI category (normal (BMI 18-25), overweight (BMI 25-30), or obese class 1 (BMI 30-35), and age category (18-25 years, 25-30 years, or 30-35 years).

Home Internet Use

The data for home Internet usage came from the baseline questionnaire survey item: “How often do you access the Internet from your home?” and had the following response categories: never; less than once a week; a few times a week; most days of the week; every day (859/1014, 84.71% of analytic sample). For the purposes of this analysis, we used the following categorizations: “never/occasionally” (never to a few times a week), “most days of the week”, and “every day”.


Engagement Patterns

Latent class analysis (LCA) was used to identify patterns of feature use as a measure of overall intervention engagement [25]. Often the latent class variable is used to organize multiple dimensions of behavior, such that individuals in each latent class share common behavior patterns. In our case, this analysis was used to group individuals based on their similar patterns feature use of the intervention website.

LCA models are fit in a series of steps starting with a one-class model; the number of classes is subsequently increased until there is no further improvement in the model. Model selection in LCA involved both absolute fit of a particular model and relative fit of two or more competing models. A common measure of absolute model fit in categorical models is the G2 likelihood-ratio chi-square statistic, which in our case tests the null hypothesis that the specified LCA model fits the data [26]. Relative fit of models with different numbers of latent classes (eg, 4 vs 5 classes) was analyzed using a series of standard fit indices, including the Bayesian information criterion (BIC [27]) and Akaike information criteria (AIC [28]), with a lower value suggest a more optimal balance between model fit and parsimony. All analyses were conducted using a SAS procedure, PROC LCA [16].

Demographic/Body Mass Index Subgroups

LCA was used to identify demographic/BMI subgroups. Given the strong correlation between demographic and BMI characteristics in this sample, we decided to take a person-centered approach to categorizing women. To do this, we used LCA to identify subgroups within the population based on race, ethnicity, income, BMI, and age. The same LCA model selection criteria were used as with the engagement patterns outlined above.

Association Between Demographic/Body Mass Index Subgroups and Engagement

Chi-square analysis was used to first examine the relationship between individual feature use and demographic subgroup. Next, chi-square analysis was used to examine the relationship between demographic subgroups and patterns of engagement.

The data analysis for this paper was generated using SAS software, version 9.3.

Characterize How Pregnant Women Engaged With Online Intervention Features

The first objective of this study was to capture multiple measures of how women used the intervention website. Most women logged into the website during pregnancy (87.97%, 892/1014) and engaged with the intervention features. As described earlier, consistency of use or quantity of use was used to characterize dose for each intervention feature. Of the intervention features, the weight tracker was most commonly used with 25.05% (254/1014) of women who consistently used, 28.99% (294/1014) almost consistently used, 19.03% (193/1014) inconsistently used, and 26.04% (264/1014) never used (Table 2). Health-related information and blogs were engaged by over half of the sample, while the diet and physical activity goal-setting tools were utilized by only a third of the sample.

When all intervention features were considered together, six patterns of engagement emerged from the latent class analysis, as shown in the column headings in Table 3. The first class was characterized by high and consistent usage of all features and is labeled “super-users” (13.02%, 132/1014). “Medium-users” (14.00%, 142/1014) were characterized by almost consistent weight-tracker use and high use of both health-related information and blogs. The next three classes were characterized in the latent class analysis solely based on weight-tracker use: “consistent weight-tracker users” (14.99%, 152/1014); “almost consistent weight-tracker users” (21.99%, 223/1014), and “inconsistent weight-tracker users” (15.98%, 162/1014). The final class, “non-users” (20.02%, 203/1014) were categorized by never engaging with the intervention features.

Comments Off on How an Online Intervention to Prevent Excessive Gestational Weight Gain Is Used and by Whom: A Randomized Controlled Process Evaluation

Tags: ,