Stephen L. Rathbun

Background: The collection of dietary behavior data is crucial in childbearing populations. In addition to observed inequities in perinatal dietary intake and quality, burdensome assessment methods (eg, 24-h dietary recall) may limit research participation for some groups. Ecological momentary assessment (EMA) is associated with reduced recall bias and participant convenience, but there is a dearth of studies with diverse cohorts. Objective: Our aim is to describe participant completion of food intake items in EMA surveys, overall and across individual characteristics (eg, prepregnancy BMI). Methods: Using secondary EMA data from participants in a longitudinal study, we report average completion rates of survey items regarding dietary behavior (eg, number of meals eaten in a day) across individual demographic variables (eg, age) and combined strata (eg, race+age) during late pregnancy and throughout 12 months post partum. Results: In our analytic sample (N=310), the average completion rate was 52.4% (SD 27.8%) during pregnancy, rising to 59.1% (SD 22.0%) after giving birth. Participants who were older (>30 y), overweight before pregnancy, self-identified as White, working, or earning higher annual income (>US $50,000) had higher average completion rates than their counterparts. Examining combined strata, we found some variation in survey completion within racial groups. Black participants using a study phone had higher average completion rates during pregnancy and post partum, but this relationship was reversed for White participants. Conclusions: Our secondary analysis showed relatively stable engagement with EMA surveys in a childbearing cohort across 15 months. Increased completion rates among privileged groups (eg, White, higher income) may demonstrate the impact of socioeconomic advantages on individual health behaviors. Investigators should consider how intersections between race and other factors (eg, employment) may impact participation and data collection.

Racial inequities in postpartum weight have been documented with limited studies on the influences of racism and other forms of discrimination. In a prospective longitudinal study applying ecological momentary assessment (EMA) and ambulatory assessment of weight, we measured the association between discrimination, stress, and postpartum weight change. The Postpartum Mothers Mobile Study is a cohort of 313 pregnant and birthing individuals who were followed during their second and third trimesters through 1 year postpartum. They were recruited in clinical settings between 2017 and 2020 in a major city center in Pennsylvania. Measures of racism and gender-based discrimination were collected via random EMA surveys throughout pregnancy and postpartum, and weight was collected via Bluetooth-enabled scales at least weekly. Among Black participants, a 10% increase in the number of days a participant experienced racial discrimination in the past month was associated with retaining 0.3 more pounds, a 10% increase in EMA gender discrimination was associated with retaining 0.4 more pounds, and EMA stress reported in the past month was associated with decreased weight retention. Chronic experiences of racial and gender discrimination may contribute to weight retention immediately after pregnancy and beyond.

BACKGROUND Background: Collection of dietary behavior data is crucial in childbearing populations. In addition to observed inequities in perinatal dietary intake and quality, burdensome assessment methods used (e.g., 24-hour dietary recall) may limit research participation for some groups. Ecological momentary assessment (EMA) is associated with reduced recall bias and participant convenience, but there is a dearth of studies with diverse cohorts. OBJECTIVE Objective: Our aim is to describe participant completion of dietary intake items in EMA surveys, overall and across individual characteristics (e.g., pre-pregnancy body mass index). METHODS Methods: Using EMA data from participants in a longitudinal study, we report average completion rates of survey items regarding dietary behavior (e.g., number of meals eaten in a day) across demographic variables (e.g., age) and intersectional strata (e.g., race + age) during late pregnancy and throughout 12 months’ postpartum. RESULTS Results: In our analytic sample (N=310), average completion rates were 59.1% during pregnancy, dropping to 52.4% after giving birth. Participants who were older (>30 years), overweight before pregnancy, self-identified as white, working, or earning higher annual income (>$50,000) had higher average completion rates than their counterparts. Examining intersectional strata, we found some variation in survey completion within racial groups. Black participants using a study phone had higher average completion rates during pregnancy and postpartum, but this relationship was reversed for white participants. CONCLUSIONS Discussion: We showed relatively stable engagement with EMA surveys in a childbearing cohort across 15 months. Increased completion rates among privileged groups (e.g., white, higher income) may demonstrate the impact of socioeconomic advantages on individual health behaviors. Investigators should consider how intersections between race and other factors (e.g., employment) may impact participation and data collection.

The era of big data has witnessed an increasing availability of observational data from mobile and social networking, online advertising, web mining, healthcare, education, public policy, marketing campaigns, and so on, which facilitates the development of causal effect estimation. Although significant advances have been made to overcome the challenges in the academic area, such as missing counterfactual outcomes and selection bias, they only focus on source-specific and stationary observational data, which is unrealistic in most industrial applications. In this paper, we investigate a new industrial problem of causal effect estimation from incrementally available observational data and present three new evaluation criteria accordingly, including extensibility, adaptability, and accessibility. We propose a Continual Causal Effect Representation Learning method for estimating causal effects with observational data, which are incrementally available from non-stationary data distributions. Instead of having access to all seen observational data, our method only stores a limited subset of feature representations learned from previous data. Combining selective and balanced representation learning, feature representation distillation, and feature transformation, our method achieves the continual causal effect estimation for new data without compromising the estimation capability for original data. Extensive experiments demonstrate the significance of continual causal effect estimation and the effectiveness of our method.

Wastewater surveillance has proven to be an effective tool to monitor the transmission and emergence of infectious agents at a community scale. Workflows for wastewater surveillance generally rely on concentration steps to increase the probability of detection of low-abundance targets, but pre-concentration can substantially increase the time and cost of analyses while also introducing additional loss of target during processing. To address some of these issues, we conducted a longitudinal study implementing a simplified workflow for SARS-CoV-2 detection from wastewater, using a direct column-based extraction approach. Composite influent wastewater samples were collected weekly for one year between June 2020 and June 2021 in Athens-Clarke County, Georgia, USA. Bypassing any concentration step, low volumes (280 µL) of influent wastewater were extracted using a commercial kit, and immediately analyzed by RT-qPCR for the SARS-CoV-2 N1 and N2 gene targets. SARS-CoV-2 viral RNA was detected in 76% (193/254) of influent samples, and the recovery of the surrogate bovine coronavirus was 42% (IQR: 28%, 59%). N1 and N2 assay positivity, viral concentration, and flow-adjusted daily viral load correlated significantly with per-capita case reports of COVID-19 at the county-level (ρ = 0.69 to 0.82). To compensate for the method's high limit of detection (approximately 10⁶–10⁷ copies L⁻¹ in wastewater), we extracted multiple small-volume replicates of each wastewater sample. With this approach, we detected as few as five cases of COVID-19 per 100 000 individuals. These results indicate that a direct-extraction-based workflow for SARS-CoV-2 wastewater surveillance can provide informative and actionable results.

Physical activity (PA) estimates from the Fitbit Flex 2 were compared to those from the ActiGraph GT9X Link in 123 elementary school children. Steps and intensity-specific estimates of PA and 3-month PA change were calculated using two different ActiGraph cut-points (Evenson and Romanzini). Fitbit estimates were 35% higher for steps compared to the ActiGraph. Fitbit and ActiGraph intensity-specific estimates were closest for sedentary and light PA, while estimates of moderate and vigorous PA varied substantially depending upon the ActiGraph cut-points used. Spearman correlations between device estimates were higher for steps (r_s = .70) than for moderate (r_s =.54 to .55) or vigorous (r_s =.29 to .48) PA. There was low concordance between devices in assessing PA changes over time. Agreement between Fitbit Flex 2 and ActiGraph estimates may depend upon the cut-points used to classify PA intensity. However, there is fair to good agreement between devices in ranking children’s steps and MVPA.

The foremost challenge to causal inference with real-world data is to handle the imbalance in the covariates with respect to different treatment options, caused by treatment selection bias. To address this issue, recent literature has explored domain-invariant representation learning based on different domain divergence metrics (e.g., Wasserstein distance, maximum mean discrepancy, position-dependent metric, and domain overlap). In this paper, we reveal the weaknesses of these strategies, i.e., they lead to the loss of predictive information when enforcing the domain invariance; and the treatment effect estimation performance is unstable, which heavily relies on the characteristics of the domain distributions and the choice of domain divergence metrics. Motivated by information theory, we propose to learn the Infomax and Domain-Independent Representations to solve the above puzzles. Our method utilizes the mutual information between the global feature representations and individual feature representations, and the mutual information between feature representations and treatment assignment predictions, in order to maximally capture the common predictive information for both treatment and control groups. Moreover, our method filters out the influence of instrumental and irrelevant variables, and thus it effectively increases the predictive ability of potential outcomes. Experimental results on both the synthetic and real-world datasets show that our method achieves state-of-the-art performance on causal effect inference. Moreover, our method exhibits reliable prediction performances when facing data with different characteristics of data distributions, complicated variable types, and severe covariate imbalance.