We aimed to develop prediction models for depression among U.S. adults with hypertension using various machine learning (ML) approaches. Moreover, we analyzed the mechanisms of the developed models. This cross-sectional study included 8,628 adults with hypertension (11.3% with depression) from the National Health and Nutrition Examination Survey (2011–2020). We selected several significant features using feature selection methods to build the models. Data imbalance was managed with random down-sampling. Six different ML classification methods implemented in the R packagecaret—artificial neural network, random forest, AdaBoost, stochastic gradient boosting, XGBoost, and support vector machine—were employed with 10-fold cross-validation for predictions. Model performance was assessed by examining the area under the receiver operating characteristic curve (AUC), accuracy, precision, sensitivity, specificity, and F1-score. For an interpretable algorithm, we used the variable importance evaluation function incaret. Of all classification models, artificial neural network trained with selected features (n = 30) achieved the highest AUC (0.813) and specificity (0.780) in predicting depression. Support vector machine predicted depression with the highest accuracy (0.771), precision (0.969), sensitivity (0.774), and F1-score (0.860). The most frequent and important features contributing to the models included the ratio of family income to poverty, triglyceride level, white blood cell count, age, sleep disorder status, the presence of arthritis, hemoglobin level, marital status, and education level. In conclusion, ML algorithms performed comparably in predicting depression among hypertensive populations. Furthermore, the developed models shed light on variables’ relative importance, paving the way for further clinical research.

Phosphorus is a unique, limited, and a critical element in agriculture and industrial applications. However, in the absence of proper control, wastewater from industries may contain excessive phosphate, which causes eutrophication of lakes and rivers. Current research on the uptake and release of phosphate from wastewater is time-consuming and may produce a harmful effect on the environment. Herein, we developed an ultra-efficient process for the selective uptake and release of phosphate in wastewater based on graphene oxide (GO)/magnesium manganese-layered double hydroxide (LDH) composites calcined at 300 °C (GO/MgMn-LDH-300). In the proposed continuous electrosorption-desorption system, the adsorbed phosphate can be quickly released by controlling the applied voltage. The intercalated GO within the layered structure increases the surface area and generates more mesopores for phosphate adsorption. In addition, the abundant oxygen-containing functional groups on GO interact with metal ions, enhance Mn oxidation, and thereby produce more active sites for phosphate adsorption. The tunable applied potential difference controls the electrostatic interactions between the LDH composites and phosphate; the LDH composites show ultrahigh efficiency, excellent recyclability, and selective phosphate removal, and the phosphate adsorption and desorption rates of 0.79 and 3.56 mg P g⁻¹ min⁻¹, respectively, have been achieved. Via incorporating a single-voltage control process, the GO/MgMn-LDH-300 composite electrodes performed excellently in phosphate removal from wastewater and exhibited an outstanding desorption capacity; this showed potential of this process for commercial wastewater treatment.

Background: The randomized clinical trial is generally considered the most rigorous study design for evaluating overall intervention effects. Because of patient heterogeneity, subgroup analysis is often used to identify differential intervention effects. In research of behavioral interventions, such subgroups often depend on a latent construct measured by multiple correlated observed variables. Objectives: The purpose of this article was to illustrate latent class analysis/latent profile analysis as a helpful tool to characterize latent subgroups, conduct exploratory subgroup analysis, and identify potential differential intervention effects using clinical trial data. Methods: After reviewing different approaches for subgroup analysis, latent class analysis/latent profile analysis was chosen to identify heterogeneous patient groups based on multiple correlated variables. This approach is superior in this specific scenario because of its ability to control Type I error, assess intersection of multiple moderators, and improve interpretability. We used a case study example to illustrate the process of identifying latent classes as potential moderators based on both clinical and perceived risk scores and then tested the differential effects of health coaching in improving health behavior for patients with elevated risk of developing coronary heart disease. Results: We identified three classes based on one clinical risk score and four perceived risk measures for individuals with high risk of developing coronary heart disease. Compared to other classes we assessed, individuals in the class with low clinical risk and low perceived risk benefit most from health coaching to improve their physical activity levels. Discussion: Latent class analysis/latent profile analysis offers a person-centered approach to identifying distinct patient profiles that can be used as moderators for subgroup analysis. This offers tremendous opportunity to identify differential intervention effects in behavioral research.

Evidence-based intervention and policy strategies to address the recent surge of race-motivated hate crimes and other forms of racism against Asian Americans are essential; however, such efforts have been impeded by a lack of empirical knowledge, e.g., about racism, specifically aimed at the Asian American population. Our qualitative descriptive study sought to fill this gap by using a data-mining approach to examine the contents of tweets having the hashtag #StopAsianHate. We collected tweets during a two-week time frame starting on 20 May 2021, when President Joe Biden signed the COVID-19 Hate Crimes Act. Screening of the 31,665 tweets collected revealed that a total of 904 tweets were eligible for thematic analysis. Our analysis revealed five themes: “Asian hate is not new”, “Address the harm of racism”, “Get involved in #StopAsianHate”, “Appreciate the Asian American and Pacific Islander (AAPI) community’s culture, history, and contributions” and “Increase the visibility of the AAPI community.” Lessons learned from our findings can serve as a foundation for evidence-based strategies to address racism against Asian Americans both locally and globally.

The findings to date indicate that adverse childhood experiences (ACEs) increase the risk of cardiovascular disease (CVD) in later life. We demonstrate how network analysis, a statistical method that estimates complex patterns of associations between variables, can be used to model ACEs and CVD. The main goal is to explore the differential impacts of ACE components on CVD outcomes, conditioned on other ACEs and important covariates using network analysis. We also sought to determine which ACEs are most synergistically correlated and subsequently cluster together to affect CVD risk. Our analysis was based on cross-sectional data from the 2020 Behavioral Risk Factor Surveillance System, which included 31,242 adults aged 55 or older (54.6% women, 79.8% whites, mean age of 68.7 ± 7.85 years). CVD outcomes included angina/coronary heart disease (CHD) and stroke prevalence. Mixed graphical models were estimated using the R-package mgm, including all variables simultaneously to elucidate their one-to-one inter-relationships. Next, we conducted Walktrap cluster detection on the estimated networks using the R-package igraph. All analyses were stratified by gender to examine group differences. In the network for men, the variable "household incarceration" was most strongly associated with stroke. For women, the strongest connection was between "physical abuse" and stroke, followed by "sexual abuse" and angina/CHD. For men, angina/CHD and stroke were clustered with several CVD risk factors, including depressive disorder, diabetes, obesity, physical activity, and smoking, and further clustered with components of household dysfunction (household substance abuse, household incarceration, and parental separation/divorce). No clusters emerged for women. Specific ACEs associated with CVDs across gender may be focal points for targeted interventions. Additionally, findings from the clustering method (especially for men) may provide researchers with valuable information on potential mechanisms linking ACEs with cardiovascular health, in which household dysfunction plays a critical role.