Community Partnership to Co-Develop an Intervention to Promote Equitable Uptake of the COVID-19 Vaccine Among Pediatric Populations.

Objective: To describe the process of engaging community, caregiver, and youth partners in codeveloping an intervention to promote equitable uptake of the COVID-19 vaccine in non-Hispanic Black (Black) and Hispanic youth who experience higher rates of COVID-19 transmission, morbidity, and mortality but were less likely to receive the COVID-19 vaccine. Methods: A team of 11 Black and Hispanic community partners was assembled to codevelop intervention strategies with our interdisciplinary research team. We used a mixed-methods crowdsourcing approach with Black and Hispanic youth (n=15) and caregivers of Black and Hispanic youth (n=20) who had not yet been vaccinated against COVID-19, recruited from primary care clinics, to elicit perspectives on the acceptability of these intervention strategies. Results: We codeveloped five strategies: (1) community-tailored handouts and posters, (2) videos featuring local youth, (3) family-centered language to offer vaccines in the primary care clinic, (4) communication-skills training for primary care providers, and (5) use of community health workers to counsel families about the vaccine. The majority (56-96.9%) of youth and caregivers rated each of these strategies as acceptable, especially because they addressed common concerns and facilitated shared decision-making. Conclusions: Engaging community and family partners led to the co-development of culturally- and locally-tailored strategies to promote dialogue and shared decision-making about the COVID-19 vaccine. This process can be used to codevelop interventions to address other forms of public health disparities. Policy Implications: Intervention strategies that promote dialogues with trusted healthcare providers and support shared decision-making are acceptable strategies to promote COVID-19 vaccine uptake among youth from historically underserved communities. Stakeholder-engaged methods may also help in the development of interventions to address other forms of health disparities.

Rural Family Satisfaction With Telehealth Delivery of an Intervention for Pediatric Obesity and Associated Family Characteristics.

Objective: To describe satisfaction with the telehealth aspect of a pediatric obesity intervention among families from multiple rural communities and assess differences in satisfaction based on sociodemographic factors. Methods: This is a secondary analysis of data from a pilot randomized controlled trial of a 6-month intensive lifestyle intervention (iAmHealthy) delivered through telehealth to children 6-11 years old with BMI ≥85th%ile and their parents from rural communities. Parents completed a sociodemographic survey and a validated survey to assess satisfaction with the telehealth intervention across four domains (technical functioning, comfort of patient and provider with technology and perceived privacy, timely and geographic access to care, and global satisfaction) on a 5-point Likert scale. Kruskal-Wallis nonparametric rank test were used to compare mean satisfaction scores based on parent sociodemographics. Results: Forty-two out of 52 parents (67% White, 29% Black, 5% multiracial, and 50% with household income <$40,000) completed the survey. Mean satisfaction scores ranged from 4.16 to 4.54 (standard deviation 0.44-0.61). Parents without a college degree reported higher satisfaction across all domains compared with parents with a college degree, including global satisfaction (mean 4.64 vs. 4.31, p = 0.03). Parents reporting a household income <$40,000 (mean 4.70) reported higher scores in the comfort with technology and perceived privacy domain compared with parents with higher incomes (mean 4.30-4.45, p = 0.04). Discussion: Parents from rural communities, especially those from lower socioeconomic backgrounds, were highly satisfied with the iAmHealthy telehealth intervention. These findings can be used to inform future telehealth interventions among larger more diverse populations. ClinicalTrials.gov Identifier: NCT04142034.

Patient-Reported Outcomes to Describe Global Health and Family Relationships in Pediatric Weight Management.

Background: Patient-reported outcomes (PROs) can assess chronic health. The study aims were to pilot a survey through the PEDSnet Healthy Weight Network (HWN), collecting PROs in tertiary care pediatric weight management programs (PWMP) in the United States, and demonstrate that a 50% enrollment rate was feasible; describe PROs in this population; and explore the relationship between child/family characteristics and PROs. Methods: Participants included 12- to 18-year-old patients and parents of 5- to 18-year-olds receiving care at PWMP in eight HWN sites. Patient-Reported Outcomes Measurement Information System (PROMIS®) measures assessed global health (GH), fatigue, stress, and family relationships (FR). T-score cut points defined poor GH or FR or severe fatigue or stress. Generalized estimating equations explored relationships between patient/family characteristics and PROMIS measures. Results: Overall, 63% of eligible parents and 52% of eligible children enrolled. Seven sites achieved the goal enrollment for parents and four for children. Participants included 1447 children. By self-report, 44.6% reported poor GH, 8.6% poor FR, 9.3% severe fatigue, and 7.6% severe stress. Multiple-parent household was associated with lower odds of poor GH by parent proxy report [adjusted odds ratio (aOR) 0.69, 95% confidence interval (CI) 0.55-0.88] and poor FR by self-report (aOR 0.36, 95% CI 0.17-0.74). Parents were significantly more likely to report that the child had poor GH and poor FR when a child had multiple households. Conclusions: PROs were feasibly assessed across the HWN, although implementation varied by site. Nearly half of the children seeking care in PWMP reported poor GH, and family context may play a role. Future work may build on this pilot to show how PROs can inform clinical care in PWMP.

The Psychosocial Impact of the COVID-19 Pandemic on Families of Youth of Color: A Prospective Cohort Study.

OBJECTIVE: Prospectively examine racial and ethnic disparities in exposure to COVID-19-related stressors and their impact on families. METHODS: A racially, ethnically, and socioeconomically diverse cohort of caregivers of youth (n = 1,581) representative of the population served by a pediatric healthcare system completed the COVID-19 Exposure and Family Impact Scales in Oct/Nov 2020 and March/April 2021. Linear mixed-effects models were used to examine exposure to COVID-19-related events (Exposure), impact of the pandemic on family functioning and well-being (Impact), and child and parent distress (Distress) across time and as a function of race and ethnicity, adjusting for other sociodemographic variables. RESULTS: Exposure and Distress increased over time for all participants. After adjusting for sociodemographic factors, caregivers of Black and Hispanic youth reported greater Exposure than caregivers of White youth and caregivers of Black youth had a greater increase in Exposure over time than caregivers of White youth. Caregivers of White youth reported greater Impact than caregivers of Black and Other race youth. CONCLUSIONS: Exposure to and impact of the COVID-19 pandemic on family psychosocial functioning varied by race and ethnicity. Although exposure to COVID-19-related events was greater among Hispanic and non-Hispanic Black families, those of marginalized races reported less family impact than non-Hispanic White families, suggesting resiliency to the pandemic. Research should examine such responses to public health crises in communities of color, with a focus on understanding protective factors. These findings suggest the importance of culturally tailored interventions and policies that support universal psychosocial screenings during times of public health crises.

Associations of longitudinal BMI percentile classification patterns in early childhood with neighborhood-level social determinants of health.

Background: Understanding social determinants of health (SDOH) that may be risk factors for childhood obesity is important to developing targeted interventions to prevent obesity. Prior studies have examined these risk factors, mostly examining obesity as a static outcome variable. Objectives: This study aimed to identify distinct subpopulations based on BMI percentile classification or changes in BMI percentile classifications over time and explore these longitudinal associations with neighborhood-level SDOH factors in children from 0 to 7 years of age. Methods: Using Latent Class Growth (Mixture) Modelling (LCGMM) we identify distinct BMI% classification groups in children from 0 to 7 years of age. We used multinomial logistic regression to study associations between SDOH factors with each BMI% classification group. Results: From the study cohort of 36,910 children, five distinct BMI% classification groups emerged: always having obesity (n=429; 1.16%), overweight most of the time (n=15,006; 40.65%), increasing BMI% (n=9,060; 24.54%), decreasing BMI% (n=5,058; 13.70%), and always normal weight (n=7,357; 19.89%). Compared to children in the decreasing BMI% and always normal weight groups, children in the other three groups were more likely to live in neighborhoods with higher rates of poverty, unemployment, crowded households, and single-parent households, and lower rates of preschool enrollment. Conclusions: Neighborhood-level SDOH factors have significant associations with children's BMI% classification and changes in classification over time. This highlights the need to develop tailored obesity interventions for different groups to address the barriers faced by communities that can impact the weight and health of the children living within them.

Persistent Disparities in Pediatric Health Care Engagement During the COVID-19 Pandemic.

OBJECTIVE: The COVID-19 pandemic has disrupted traditional health care, including pediatric health care. We described the impact of the pandemic on disparities in pediatric health care engagement. METHODS: Using a population-based cross-sectional time-series design, we compared monthly ambulatory care visit volume and completion rates (completed vs no-show and cancelled visits) among pediatric patients aged 0-21 years in 4 states in the mid-Atlantic United States during the first year of the COVID-19 pandemic (March 2020-February 2021) with the same period before the pandemic (March 2019-February 2020). We used unadjusted odds ratios, stratified by visit type (telehealth or in-person) and sociodemographic characteristics (child race and ethnicity, caregiver primary language, geocoded Child Opportunity Index, and rurality). RESULTS: We examined 1 556 548 scheduled ambulatory care visits for a diverse pediatric patient population. Visit volume and completion rates (mean, 70.1%) decreased during the first months of the pandemic but returned to prepandemic levels by June 2020. Disparities in in-person visit completion rates among non-Hispanic Black versus non-Hispanic White patients (64.9% vs 74.3%), patients from socioeconomically disadvantaged versus advantaged communities as measured by Child Opportunity Index (65.8% vs 76.4%), and patients in rural versus urban neighborhoods (66.0% vs 70.8%) were the same during the remainder of the first year of the pandemic as compared with the previous year. Concurrent with large increases in telehealth (0.5% prepandemic, 19.0% during the pandemic), telehealth completion rates increased. CONCLUSIONS: Disparities in pediatric visit completion rates that existed before the pandemic persisted during the pandemic. These findings underscore the need for culturally tailored practices to reduce disparities in pediatric health care engagement.

Active vs Traditional Methods of Recruiting Children for a Clinical Trial in Rural Primary Care Clinics: A Cluster-Randomized Clinical Trial.

Importance: To our knowledge, there are no published randomized clinical trials of recruitment strategies. Rigorously evaluated successful recruitment strategies for children are needed. Objective: To evaluate the feasibility of 2 recruitment methods for enrolling rural children through primary care clinics to assess whether either or both methods are sufficiently effective for enrolling participants into a clinical trial of a behavioral telehealth intervention for children with overweight or obesity. Design, Setting, and Participants: This cluster-randomized clinical trial of 2 recruitment methods was conducted at 4 primary care clinics in 4 separate states. Each clinic used both recruitment methods in random order. Clinic eligibility criteria included at least 40% pediatric patients with Medicaid coverage and at least 100 potential participants. Eligibility criteria for children included a rural home address, age 6 to 11 years, and body mass index at or above the 85th percentile. Recruitment began February 3, 2020, and randomization of participants occurred on August 17, 2020. Data were analyzed from October 3, 2021, to April 21, 2022. Interventions: Two recruitment methods were assessed: the active method, for which a list of potential participants seen within the past year at each clinic was generated through the electronic health record and consecutively approached by research staff based on visit date to the clinic, and the traditional method, for which recruitment included posters, flyers, social media, and press release. Clinics were randomized to the order in which the 2 methods were implemented in 4-week periods, followed by a 4-week catch-up period using the method found most effective in previous periods. Main Outcomes and Measures: For each recruitment method, the number and proportion of randomized children among those who were approached was calculated. Results: A total of 104 participants were randomized (58 girls [55.8%]; mean age, 9.3 [95% CI, 9.0-9.6] years). Using the active method, 535 child-parent dyads were approached and 99 (18.5% [95% CI, 15.3%-22.1%]) were randomized. Using the traditional method, 23 caregivers expressed interest, and 5 (21.7% [95% CI, 7.5%-43.7%]) were randomized. All sites reached full enrollment using the active method and no sites achieved full enrollment using the traditional method. Mean time to full enrollment was 26.3 (range, 21.0-31.0) days. Conclusions and Relevance: This study supports the use of the active approach with local primary care clinics to recruit children with overweight and obesity from rural communities into clinical trials. Trial Registration: ClinicalTrials.gov Identifier: NCT04142034.

Conducting a pediatric randomized clinical trial during a pandemic: A shift to virtual procedures.

Background/Objective: Prior to the COVID-19 pandemic, our research group initiated a pediatric practice-based randomized trial for the treatment of childhood obesity in rural communities. Approximately 6 weeks into the originally planned 10-week enrollment period, the trial was forced to pause all study activity due to the COVID-19 pandemic. This pause necessitated a substantial revision in recruitment, enrollment, and other study methods in order to complete the trial using virtual procedures. This descriptive paper outlines methods used to recruit, enroll, and manage clinical trial participants with technology to obtain informed consent, obtain height and weight measurements by video, and maintain participant engagement throughout the duration of the trial. Methods: The study team reviewed the IRB records, protocol team meeting minutes and records, and surveyed the site teams to document the impact of the COVID-19 shift to virtual procedures on the study. The IRB approved study changes allowed for flexibility between clinical sites given variations in site resources, which was key to success of the implementation. Results: All study sites faced a variety of logistical challenges unique to their location yet successfully recruited the required number of patients for the trial. Ultimately, virtual procedures enhanced our ability to establish relationships with participants who were previously beyond our reach, but presented several challenges and required additional resources. Conclusion: Lessons learned from this study can assist other study groups in navigating challenges, especially when recruiting and implementing studies with rural and underserved populations or during challenging events like the pandemic.

Flexible-Window Predictions on Electronic Health Records.

Various types of machine learning techniques are available for analyzing electronic health records (EHRs). For predictive tasks, most existing methods either explicitly or implicitly divide these time-series datasets into predetermined observation and prediction windows. Patients have different lengths of medical history and the desired predictions (for purposes such as diagnosis or treatment) are required at different times in the future. In this paper, we propose a method that uses a sequence-to-sequence generator model to transfer an input sequence of EHR data to a sequence of user-defined target labels, providing the end-users with "flexible" observation and prediction windows to define. We use adversarial and semi-supervised approaches in our design, where the sequence-to-sequence model acts as a generator and a discriminator distinguishes between the actual (observed) and generated labels. We evaluate our models through an extensive series of experiments using two large EHR datasets from adult and pediatric populations. In an obesity predicting case study, we show that our model can achieve superior results in flexible-window prediction tasks, after being trained once and even with large missing rates on the input EHR data. Moreover, using a number of attention analysis experiments, we show that the proposed model can effectively learn more relevant features in different prediction tasks.

Associations Between Weight and Lower Respiratory Tract Disease Outcomes in Hospitalized Children.

OBJECTIVE: To identify associations between weight status and clinical outcomes in children with lower respiratory tract infection (LRTI) or asthma requiring hospitalization. METHODS: We performed a retrospective cohort study of 2 to 17 year old children hospitalized for LRTI and/or asthma from 2009 to 2019 using electronic health record data from the PEDSnet clinical research network. Children <2 years, those with medical complexity, and those without a calculable BMI were excluded. Children were classified as having underweight, normal weight, overweight, or class 1, 2, or 3 obesity based on Body Mass Index percentile for age and sex. Primary outcomes were need for positive pressure respiratory support and ICU admission. Subgroup analyses were performed for children with a primary diagnosis of asthma. Outcomes were modeled with mixed-effects multivariable logistic regression incorporating age, sex, and payer as fixed effects. RESULTS: We identified 65 132 hospitalizations; 6.7% with underweight, 57.8% normal weight, 14.6% overweight, 13.2% class 1 obesity, 5.0% class 2 obesity, and 2.8% class 3 obesity. Overweight and obesity were associated with positive pressure respiratory support (class 3 obesity versus normal weight odds ratio [OR] 1.62 [1.38-1.89]) and ICU admission (class 3 obesity versus normal weight OR 1.26 [1.12-1.42]), with significant associations for all categories of overweight and obesity. Underweight was also associated with positive pressure respiratory support (OR 1.39 [1.24-1.56]) and ICU admission (1.40 [1.30-1.52]). CONCLUSIONS: Both underweight and overweight or obesity are associated with increased severity of LRTI or asthma in hospitalized children.

Obesity Prediction with EHR Data: A deep learning approach with interpretable elements.

Childhood obesity is a major public health challenge. Early prediction and identification of the children at an elevated risk of developing childhood obesity may help in engaging earlier and more effective interventions to prevent and manage obesity. Most existing predictive tools for childhood obesity primarily rely on traditional regression-type methods using only a few hand-picked features and without exploiting longitudinal patterns of children's data. Deep learning methods allow the use of high-dimensional longitudinal datasets. In this paper, we present a deep learning model designed for predicting future obesity patterns from generally available items on children's medical history. To do this, we use a large unaugmented electronic health records dataset from a large pediatric health system in the US. We adopt a general LSTM network architecture and train our proposed model using both static and dynamic EHR data. To add interpretability, we have additionally included an attention layer to calculate the attention scores for the timestamps and rank features of each timestamp. Our model is used to predict obesity for ages between 3-20 years using the data from 1-3 years in advance. We compare the performance of our LSTM model with a series of existing studies in the literature and show it outperforms their performance in most age ranges.

COVID-19 Exposure and Family Impact Scales for Adolescents and Young Adults.

OBJECTIVE: To understand the impact of the coronavirus disease 2019 (COVID-19) pandemic on adolescents and young adults (AYAs), we adapted the COVID-19 Exposure and Family Impact Scales (CEFIS; Kazak et al., 2021) for AYAs. Here, we report on the development, structure, and psychometric properties of the CEFIS-AYA. METHODS: The CEFIS-AYA was developed by a multidisciplinary, multi-institutional team using a rapid iterative process. Data from 3,912 AYAs from 21 programs at 16 institutions across the United States were collected from May 2020 to April 2021. We examined the underlying structure of the CEFIS-AYA using principal component analysis (PCA), calculated internal consistencies, and explored differences in scores by gender and age. RESULTS: Participants reported exposure to a range of COVID-19-related events (M = 9.08 events, of 28). On the bidirectional 4-point Impact scale, mean item scores were mostly above the midpoint, indicating a slightly negative impact. Kuder-Richardson 20/Cronbach's Alpha was good for Exposure (α = .76) and excellent for Impact (α = .93). PCA identified seven factors for Exposure (Severe COVID-19, Loss of Income, Limited Access to Essentials, COVID-19 Exposure, Disruptions to Activities, Disruptions to Living Conditions, and Designation as an Essential Worker) and five for Impact (Self and Family Relationships, Physical Well-Being, Emotional Well-Being, Social Well-Being, and Distress). Gender and age differences in CEFIS-AYA scores were identified. DISCUSSION: Initial reliability data are strong and support use of the CEFIS-AYA for measuring the effect of the COVID-19 pandemic on AYAs in research and clinical care.

Medical factors associated with caregiver intention to vaccinate their children against COVID-19.

OBJECTIVE: To describe medical factors that are associated with caregiver intention to vaccinate their children against COVID-19. METHODS: We conducted a cross-sectional study of families receiving primary care in a mid-Atlantic pediatric healthcare system, linking caregiver-reported data from a survey completed March 19 to April 16, 2021 to comprehensive data from the child's EHR. RESULTS: 513 families were included (28% Black, 16% Hispanic, 44% public insurance, 21% rural, child age range 0-21 years). 44% of caregivers intended to vaccinate their children against COVID-19, while 41% were not sure and 15% would not. After adjusting for socio-demographics, the only medical factors that were associated with caregiver COVID-19 vaccine hesitancy were caregiver COVID-19 vaccination status at the time of the survey (aOR 3.0 if the caregiver did not receive the vaccine compared to those who did, 95% CI 1.7-5.3) and child seasonal influenza immunization history (aOR 3.3 if the child had not received the influenza vaccine in the 2020-2021 season compared to those who did, 95% CI 2.0-5.4). Other medical factors, including family medical experiences with COVID-19, other child immunization history, child health conditions like obesity and asthma, and family engagement with the healthcare system were not associated with caregiver intention to vaccinate their children against COVID-19. CONCLUSIONS: This study highlights important factors, such as general attitudes towards vaccines and understanding of COVID-19 morbidity risk factors, that healthcare providers should address when having conversations with families about the COVID-19 vaccine.

Measuring BMI change among children and adolescents.

BACKGROUND: Weight control programs for children monitor BMI changes using BMI z-scores that adjust BMI for the sex and age of the child. It is, however, uncertain if BMIz is the best metric for assessing BMI change. OBJECTIVE: To identify which of 6 BMI metrics is optimal for assessing change. We considered a metric to be optimal if its short-term variability was consistent across the entire BMI distribution. SUBJECTS: 285 643 2- to 17-year-olds with BMI measured 3 times over a 10- to 14-month period. METHODS: We summarized each metric's variability using the within-child standard deviation. RESULTS: Most metrics' initial or mean value correlated with short-term variability (|r| ~ 0.3 to 0.5). The metric for which the within-child variability was largely independent (r = 0.13) of the metric's initial or mean value was the percentage of the 50th expressed on a log scale. However, changes in this metric between the first and last visits were highly (r ≥ 0.97) correlated with changes in %95th and %50th. CONCLUSIONS: Log %50 was the metric for which the short-term variability was largely independent of a child's BMI. Changes in log %50th, %95th, and %50th are strongly correlated.

Interrelationships among age at adiposity rebound, BMI during childhood, and BMI after age 14 years in an electronic health record database.

OBJECTIVE: This study compared the importance of age at adiposity rebound versus childhood BMI to subsequent BMI levels in a longitudinal analysis. METHODS: From the electronic health records of 4.35 million children, a total of 12,228 children were selected who were examined at least once each year between ages 2 and 7 years and reexamined after age 14 years. The minimum number of examinations per child was six. Each child's rebound age was estimated using locally weighted regression (lowess), a smoothing technique. RESULTS: Children who had a rebound age < 3 years were, on average, 7 kg/m2 heavier after age 14 years than were children with a rebound age ≥ 7 years. However, BMI after age 14 years was more strongly associated with BMI at the rebound than with rebound age (r = 0.57 vs. -0.44). Furthermore, a child's BMI at age 3 years provided more information on BMI after age 14 years than did rebound age. In addition, rebound age provided no information on subsequent BMI if a child's BMI at age 6 years was known. CONCLUSIONS: Although rebound age is related to BMI after age 14 years, a child's BMI at age 3 years provides more information and is easier to obtain.

Validation of the COVID-19 Exposure and Family Impact Scales.

OBJECTIVE: The COVID-19 Exposure and Family Impact Scales (CEFIS) were developed in Spring 2020 to assess effects of the COVID-19 pandemic on families and caregivers. Initial psychometric properties were promising. The current study examined the factor structure and evaluated convergent and criterion validity of the CEFIS in a new sample. METHODS: In October and November 2020, caregivers (N = 2,531) of youth (0-21 years) scheduled for an ambulatory care visit at Nemours Children's Hospital, Delaware completed the CEFIS and measures of convergent (PROMIS Global Mental Health Scale, Family Assessment Device) and criterion validity (PTSD Checklist-Civilian). Confirmatory factor analysis was used to examine the factor structure of the CEFIS. Bivariate correlations and logistic regression were used to examine convergent and criterion validity. RESULTS: Factor analysis supported the original six- and three-factor structures for the Exposure and Impact scales, respectively. Second-order factor analyses supported the use of Exposure, Impact, and Distress total scores. Higher scores on the CEFIS Exposure, Impact, and Distress scales were associated with increased mental health concerns and poorer family functioning. Higher scores on all CEFIS scales were also associated with greater odds of having clinically significant posttraumatic stress symptoms. CONCLUSIONS: The CEFIS is a psychometrically sound measure of the impact of the COVID-19 pandemic on family and caregiver functioning and may also be useful in identifying families who would benefit from psychological supports.

Predicting Attrition Patterns from Pediatric Weight Management Programs.

Obesity is a major public health concern. Multidisciplinary pediatric weight management programs are considered standard treatment for children with obesity who are not able to be successfully managed in the primary care setting. Despite their great potential, high dropout rates (referred to as attrition) are a major hurdle in delivering successful interventions. Predicting attrition patterns can help providers reduce the alarmingly high rates of attrition (up to 80%) by engaging in earlier and more personalized interventions. Previous work has mainly focused on finding static predictors of attrition on smaller datasets and has achieved limited success in effective prediction. In this study, we have collected a five-year comprehensive dataset of 4,550 children from diverse backgrounds receiving treatment at four pediatric weight management programs in the US. We then developed a machine learning pipeline to predict (a) the likelihood of attrition, and (b) the change in body-mass index (BMI) percentile of children, at different time points after joining the weight management program. Our pipeline is greatly customized for this problem using advanced machine learning techniques to process longitudinal data, smaller-size data, and interrelated prediction tasks. The proposed method showed strong prediction performance as measured by AUROC scores (average AUROC of 0.77 for predicting attrition, and 0.78 for predicting weight outcomes).

Weight Status and Risk of Inpatient Admission for Children With Lower Respiratory Tract Disease.

OBJECTIVES: To identify associations between weight category and hospital admission for lower respiratory tract disease (LRTD), defined as asthma, community-acquired pneumonia, viral pneumonia, or bronchiolitis, among children evaluated in pediatric emergency departments (PEDs). METHODS: We performed a retrospective cohort study of children 2 to <18 years of age evaluated in the PED at 6 children's hospitals within the PEDSnet clinical research network from 2009 to 2019. BMI percentile of children was classified as underweight, healthy weight, overweight, and class 1, 2, or 3 obesity. Children with complex chronic conditions were excluded. Mixed-effects multivariable logistic regression was used to assess associations between BMI categories and hospitalization or 7- and 30-day PED revisits, adjusted for covariates (age, sex, race and ethnicity, and payer). RESULTS: Among 107 446 children with 218 180 PED evaluations for LRTD, 4.5% had underweight, 56.4% had healthy normal weight, 16.1% had overweight, 14.6% had class 1 obesity, 5.5% had class 2 obesity, and 3.0% had class 3 obesity. Underweight was associated with increased risk of hospital admission compared with normal weight (odds ratio [OR] 1.76; 95% confidence interval [CI] 1.69-1.84). Overweight (OR 0.87; 95% CI 0.85-0.90), class 1 obesity (OR 0.88; 95% CI 0.85-0.91), and class 2 obesity (OR 0.91; 95% CI 0.87-0.96) had negative associations with hospital admission. Class 1 and class 2, but not class 3, obesity had small positive associations with 7- and 30-day PED revisits. CONCLUSIONS: We found an inverse relationship between patient weight category and risk for hospital admission in children evaluated in the PED for LRTD.

Concurrent Imputation and Prediction on EHR data using Bi-Directional GANs: Bi-GANs for EHR imputation and prediction.

Working with electronic health records (EHRs) is known to be challenging due to several reasons. These reasons include not having: 1) similar lengths (per visit), 2) the same number of observations (per patient), and 3) complete entries in the available records. These issues hinder the performance of the predictive models created using EHRs. In this paper, we approach these issues by presenting a model for the combined task of imputing and predicting values for the irregularly observed and varying length EHR data with missing entries. Our proposed model (dubbed as Bi-GAN) uses a bidirectional recurrent network in a generative adversarial setting. In this architecture, the generator is a bidirectional recurrent network that receives the EHR data and imputes the existing missing values. The discriminator attempts to discriminate between the actual and the imputed values generated by the generator. Using the input data in its entirety, Bi-GAN learns how to impute missing elements in-between (imputation) or outside of the input time steps (prediction). Our method has three advantages to the state-of-the-art methods in the field: (a) one single model performs both the imputation and prediction tasks; (b) the model can perform predictions using time-series of varying length with missing data; (c) it does not require to know the observation and prediction time window during training and can be used for the predictions with different observation and prediction window lengths, for short- and long-term predictions. We evaluate our model on two large EHR datasets to impute and predict body mass index (BMI) values and show its superior performance in both settings.