Nonalcoholic fatty liver disease (NAFLD) detection and deep learning in a Chinese community-based population

OBJECTIVES

We aimed to develop and validate a deep learning system (DLS) by using an auxiliary section that extracts and outputs specific ultrasound diagnostic features to improve the explainable, clinical relevant utility of using DLS for detecting NAFLD.

METHODS

In a community-based study of 4144 participants with abdominal ultrasound scan in Hangzhou, China, we sampled 928 (617 [66.5%] females, mean age: 56 years ± 13 [standard deviation]) participants (2 images per participant) to develop and validate DLS, a two-section neural network (2S-NNet). Radiologists' consensus diagnosis classified hepatic steatosis as none steatosis, mild, moderate, and severe. We also explored the NAFLD detection performance of six one-section neural network models and five fatty liver indices on our data set. We further evaluated the influence of participants' characteristics on the correctness of 2S-NNet by logistic regression.

RESULTS

Area under the curve (AUROC) of 2S-NNet for hepatic steatosis was 0.90 for ≥ mild, 0.85 for ≥ moderate, and 0.93 for severe steatosis, and was 0.90 for NAFLD presence, 0.84 for moderate to severe NAFLD, and 0.93 for severe NAFLD. The AUROC of NAFLD severity was 0.88 for 2S-NNet, and 0.79-0.86 for one-section models. The AUROC of NAFLD presence was 0.90 for 2S-NNet, and 0.54-0.82 for fatty liver indices. Age, sex, body mass index, diabetes, fibrosis-4 index, android fat ratio, and skeletal muscle via dual-energy X-ray absorptiometry had no significant impact on the correctness of 2S-NNet (p > 0.05).

CONCLUSIONS

By using two-section design, 2S-NNet had improved the performance for detecting NAFLD with more explainable, clinical relevant utility than using one-section design.

KEY POINTS

• Based on the consensus review derived from radiologists, our DLS (2S-NNet) had an AUROC of 0.88 by using two-section design and yielded better performance for detecting NAFLD than using one-section design with more explainable, clinical relevant utility. • The 2S-NNet outperformed five fatty liver indices with the highest AUROCs (0.84-0.93 vs. 0.54-0.82) for different NAFLD severity screening, indicating screening utility of deep learning-based radiology may perform better than blood biomarker panels in epidemiology. • The correctness of 2S-NNet was not significantly influenced by individual's characteristics, including age, sex, body mass index, diabetes, fibrosis-4 index, android fat ratio, and skeletal muscle via dual-energy X-ray absorptiometry.

Development and Validation of Age-Specific Resilience Instruments for Early Childhood Assessment: A Taiwan Birth Cohort Study

BACKGROUND

We sought to develop and validate age-specific instruments for measuring early childhood resilience at ages 3, 5 and 8 in the Taiwan Birth Cohort Study, a national longitudinal study.

METHODS

Using data from 18,553 mother-infant pairs, we conducted exploratory factor analysis (EFA) on a simple random half of our sample. We then used the remaining half of these data for confirmatory factor analysis (CFA) to further assess the fit of 3 CFA models (ie, first-order, second-order, and bifactor). Psychometric properties, distributions, and inter-item and inter-factor correlations of each instrument were also evaluated.

RESULTS

EFA and CFA showed that the bifactor model of resilience (which included a general resilience factor and 5 specific factors) had the best fit for all 3 resilience scales, with 19 items at year 3, 18 items at year 5, and 19 items at year 8. All 3 resilience scales showed good psychometric properties, including construct validity, internal consistency, and normal distributions. For predictive validity, we found that in the face of adversity (measured by the High Risk Family Score), individuals with high resilience scores at age 3 had better general health scores at ages 3, 5, and 8 compared to those with low resilience scores.

CONCLUSIONS

We describe the development and validation of age-appropriate survey instruments to assess resilience in young children at the population level. These instruments can be used to better understand how resilience can impact child health over time, and to identify key factors that can foster resilience.

Prospective Pilot Study Evaluating SARS-CoV-2 Transmission-Limiting Measures in an On-Site School

OBJECTIVES

The purpose of our study is to evaluate the feasibility and reliability of a comprehensive set of preventive measures in limiting secondary transmission of COVID-19 in schools.

METHODS

A prospective cohort study was conducted to evaluate SARS-CoV-2 transmission in an independent K-8 school in San Mateo County, California. The research was conducted between September 14, 2020 through March 22, 2021 and consisted of: 1) demographic and epidemiological questionnaires; 2) daily symptom reporting; 3) weekly RT-PCR testing; and 4) periodic on-site qualitative observations.

RESULTS

One hundred eighty (79%) students and 63 (74%) on-site staff/contractors were enrolled. Participants reported symptoms in 144 (<1%) daily surveys of the 19,409 collected. Among those who reported symptoms and exposures, none tested positive during the 22-week study period. Of all participants, a total of 6 tested positive for SARS-CoV-2 at least once by RT-PCR; all were asymptomatic at time of testing. No in-school transmission occurred. Mask adherence was high among all grades, and incidents of improper mask use mostly occurred during noninstruction time. Physical distancing was well-enforced during class time and snack breaks, although adherence during noninstruction time waned as the school year progressed.

CONCLUSIONS

Our comprehensive, prospective study following COVID-19 transmission over 22 weeks in a K-8 school demonstrates that: 1) surveillance testing is important for detecting asymptomatic infections in schools; 2) monitoring symptoms may not be necessary and/or sufficient for COVID-19; and 3) younger children can adhere to key mitigation measures (eg, masking) which have the potential to limit transmission.

Feasibility of Specimen Self-collection in Young Children Undergoing SARS-CoV-2 Surveillance for In-Person Learning

IMPORTANCE

There is an urgent need to assess the feasibility of COVID-19 surveillance measures in educational settings.

OBJECTIVE

To assess whether young children can feasibly self-collect SARS-CoV-2 samples for surveillance testing over the course of an academic year.

DESIGN, SETTING, AND PARTICIPANTS

This prospective pilot cohort study was conducted from September 10, 2020, to June 10, 2021, at a K-8 school in San Mateo County, California. The research consisted of quantitative data collection efforts: (1) demographic data collected, (2) student sample self-collection error rates, and (3) student sample self-collection time durations. Students were enrolled in a hybrid learning model, a teaching model in which students were taught in person and online, with students having the option to attend virtually as needed. Data were collected under waiver of consent from students participating in weekly SARS-CoV-2 testing.

MAIN OUTCOMES AND MEASURES

Errors over time for self-collection of nasal swabs such as contaminated swabs and inadequate or shallow swabbing; time taken for sample collection.

RESULTS

Of 296 participants, 148 (50.0%) were boys and 148 (50.0%) were girls. A total of 87 participants (29.2%) identified as Asian; 2 (0.6%), Black or African American; 13 (4.4%), Hispanic/Latinx; 103 (34.6%), non-Hispanic White; 87 (29.2%), multiracial; and 6 (2.0%), other. The median school grade was fourth grade. From September 2020 to March 2021, a total of 4203 samples were obtained from 221 students on a weekly basis, while data on error rates were collected. Errors occurred in 2.7% (n = 107; 95% CI, 2.2%-3.2%) of student encounters, with the highest rate occurring on the first day of testing (20 [10.2%]). There was an overall decrease in error rates over time. From April to June 2021, a total of 2021 samples were obtained from 296 students on a weekly basis while data on encounter lengths were collected. Between April and June 2021, 193 encounters were timed. The mean duration of each encounter was 70 seconds (95% CI, 66.4-73.7 seconds).

CONCLUSIONS AND RELEVANCE

Mastery of self-collected lower nasal swabs is possible for children 5 years and older. Testing duration can be condensed once students gain proficiency in testing procedures. Scalability for larger schools is possible if consideration is given to the resource-intensive nature of the testing and the setting's weather patterns.

Establishing a Cancer Registry in a Resource-Constrained Region: Process Experience From Ghana

PURPOSE

In a review of cancer incidence across continents (GLOBOCAN 2012), data sources from Ghana were classified as Frequencies, the lowest classification for inclusion, signifying the worst data quality for inclusion in the analysis. Recognizing this deficiency, the establishment of a population-based cancer registry was proposed as part of a broader cancer control plan.

METHODS

The registry was examined under the following headings: policy, data source, and administrative structure; external support and training; and definition of geographic coverage.

RESULTS

The registry was set up based on the Ghana policy document on the strategy for cancer control. The paradigm shift ensured subscription to one data collection software (CanReg 5) in the country. The current approach consists of trained registrars based in the registry who conduct active data abstraction at the departments and units of the hospital and pathologic services. To ensure good governance, an administrative structure was created, including an advisory board, a technical committee, and registry staff. External support for the establishment of the Accra Cancer Registry has come mainly from Stanford University and the African Cancer Registry Network, in collaboration with the University of Ghana. Unlike previous attempts, this registry has a well-defined population made up of nine municipal districts.

CONCLUSION

The Accra Cancer Registry was established as a result of the lessons learned from failed previous attempts and aim to provide a model for setting up other cancer registries in Ghana. It will eventually be the focal point where all the national data can be collated.

PretermConnect: Leveraging mobile technology to mitigate social disadvantage in the NICU and beyond

Preterm birth (PTB) - delivery prior to 37-weeks gestation - disproportionately affects low-income and minority populations and leads to substantial infant morbidity and mortality. The time following a PTB represents an optimal window for targeted interventions that encourage mothers to prioritize their own health and that of their babies. Healthcare teams can leverage digital strategies to address maternal and infant needs in this postpartum period, both in the neonatal intensive care unit and beyond. We therefore developed PretermConnect, a mobile app designed to educate, engage, and empower women at risk for PTB. This article describes the participant-centered design approach of PretermConnect, with preliminary findings from focus groups and co-design sessions in different community settings and suggested future directions for mobile technologies in population health. Apps such as PretermConnect can mitigate social disadvantage by serving as remote monitoring tools, providing social support, preventing recurrent PTB and lowering infant mortality rates.

Influence of Health Beliefs on Adherence to COVID-19 Preventative Practices: International, Social Media-Based Survey Study

BACKGROUND

Health behavior is influenced by culture and social context. However, there are limited data evaluating the scope of these influences on COVID-19 response.

OBJECTIVE

This study aimed to compare handwashing and social distancing practices in different countries and evaluate practice predictors using the health belief model (HBM).

METHODS

From April 11 to May 1, 2020, we conducted an online, cross-sectional survey disseminated internationally via social media. Participants were adults aged 18 years or older from four different countries: the United States, Mexico, Hong Kong (China), and Taiwan. Primary outcomes were self-reported handwashing and social distancing practices during COVID-19. Predictors included constructs of the HBM: perceived susceptibility, perceived severity, perceived benefits, perceived barriers, self-efficacy, and cues to action. Associations of these constructs with behavioral outcomes were assessed by multivariable logistic regression.

RESULTS

We analyzed a total of 71,851 participants, with 3070 from the United States, 3946 from Mexico, 1201 from Hong Kong (China), and 63,634 from Taiwan. Of these countries, respondents from the United States adhered to the most social distancing practices (χ23=2169.7, P<.001), while respondents from Taiwan performed the most handwashing (χ23=309.8, P<.001). Multivariable logistic regression analyses indicated that self-efficacy was a positive predictor for handwashing (odds ratio [OR]United States 1.58, 95% CI 1.21-2.07; ORMexico 1.5, 95% CI 1.21-1.96; ORHong Kong 2.48, 95% CI 1.80-3.44; ORTaiwan 2.30, 95% CI 2.21-2.39) and social distancing practices (ORUnited States 1.77, 95% CI 1.24-2.49; ORMexico 1.77, 95% CI 1.40-2.25; ORHong Kong 3.25, 95% CI 2.32-4.62; ORTaiwan 2.58, 95% CI 2.47-2.68) in all countries. Handwashing was positively associated with perceived susceptibility in Mexico, Hong Kong, and Taiwan, while social distancing was positively associated with perceived severity in the United States, Mexico, and Taiwan.

CONCLUSIONS

Social media recruitment strategies can be used to reach a large audience during a pandemic. Self-efficacy was the strongest predictor for handwashing and social distancing. Policies that address relevant health beliefs can facilitate adoption of necessary actions for preventing COVID-19. Our findings may be explained by the timing of government policies, the number of cases reported in each country, individual beliefs, and cultural context.

Child Health and Telehealth in Global, Underresourced Settings

Telehealth has improved delivery of health care worldwide by improving access to and the quality of health care and by improving the global shortage of health professionals through collaboration and training. Although many telehealth efforts have been reported in adult health care settings, it is important to examine telehealth efforts in the pediatric setting. Children who are most commonly ill and malnourished are often those of underserved populations of the developing world. This article examines current uses of pediatric telehealth in a global setting and discusses key approaches to how telehealth may become successfully integrated and scaled in those settings.

Associations Between Body Fat, Muscle Mass, and Nonalcoholic Fatty Liver Disease: A Population-Based Study

Nonalcoholic fatty liver disease (NAFLD) is one of the most common forms of liver disease worldwide and has emerged as a significant public health concern in China. A better understanding of the etiology of NAFLD can inform effective management strategies for this disease. We examined factors associated with NAFLD in two districts of Hangzhou, China, focusing on the relationship of regional body fat distribution, muscle mass, and NAFLD. We used baseline data to carry out a cross‐sectional analysis among 3,589 participants from the Wellness Living Laboratory (WELL) China study, a longitudinal population‐based study that aims to investigate and promote well‐being among the Chinese population. NAFLD was defined using the widely validated fatty liver index (FLI). Multivariate logistic regressions were performed to assess independent associations between NAFLD and metabolic risk factors (e.g., insulin resistance) and dual x‐ray absorptiometry (DXA)‐derived measures (e.g., android fat ratio [AFR] and skeletal muscle index [SMI]). Of the 3,589 participants, 476 (13.3%) were classified as having FLI‐defined NAFLD (FLI ≥60). Among those, 58.0% were men. According to our analysis, AFR (odds ratio [OR], 10.0; 95% confidence interval [CI], 5.8‐18.5), insulin resistance (OR, 4.0; 95% CI, 3.0‐5.3), high alanine aminotransferase levels (OR, 7.6; 95% CI, 5.8‐10.0), smoking (OR, 2.0; 95% CI, 1.4‐3.0), and male sex (OR, 2.9; 95% CI, 2.0‐4.2) were positively associated with NAFLD risk, while SMI (OR, 0.1; 95% CI, 0.07‐0.13) was inversely associated with NAFLD risk. Conclusion: In addition to known metabolic risk factors, DXA‐derived AFR and SMI may provide additional insights to the understanding of NAFLD. Interventions that aim to decrease AFR and increase SMI may be important to reduce the burden of NAFLD in this population.