Zero-inflated Negative Binomial Research Articles

Measles Disease Analysis in Bengkulu Province Using Zero Inflated Poisson Regression and Zero Inflated Negative Binomial Regression

Zero Inflated Poisson regression (ZIP) and Zero Inflated Negative Binomial (ZINB) regression were used if there was overdispersion and no multicollinearity in the data. This study aims to analyze measles in Bengkulu Province using the ZIP and ZINB regression models. Among them are selecting the best model, seeing the influential variables from the best model, and predicting the results of the best model. The data used is one dependent variable, namely the number of measles cases (Y) in each puskesmas and six independent variables (X) namely the percentage of measles immunization, the amount of malnutrition, the percentage of exclusive breastfeeding, the percentage of vitamin A, the percentage of proper sanitation, and the percentage of healthy house. The results of this study, the ZIP regression model formed is a discrete model for , namely ln()=-5.042-0.007X1-0.014X3+0.094X4 and a zero inflation model for , namely logit()= -3.656+0.101X4-0.054X6, while the ZINB regression model formed is a discrete model for , namely ln()=-9,289+0.120X4 and a zero inflation model for , namely logit()=- 17.841+0.205X4. The AIC value of the ZINB regression model is 255.249, which is smaller than the AIC value of the ZIP regression model of 331.467, so the ZINB regression model is better to use. The influential variable in this study is the percentage of vitamin A administered. There is not much difference between predicted results and the actual data.

Liu-type shrinkage strategies in zero-inflated negative binomial models with application to Expenditure and Default Data

In modeling count data with overdispersion and extra zeros, zero-inflated negative binomial (ZINB) regression model is useful. In a regression model, the multicollinearity problem arises when there are some high correlations between predictor variables. This problem leads to the maximum likelihood method will not be an efficient estimator. The ridge and Liu-type estimators have been proposed to combat the multicollinearity problem so that the Liu-type estimator is better. In this paper, we proposed the Liu-type shrinkage estimators, namely linear shrinkage, preliminary test, shrinkage preliminary test, Stein-type, and positive Stein-type Liu estimators to estimate the count parameters in the ZINB model, when some of the predictor variables have not a significant effect to predict the response variable so that a sub-model may be sufficient. The asymptotic distributional biases and variances of the proposed estimators are nicely demonstrated. We also compared the performance of the Liu-type shrinkage estimators along with the Liu-type unrestricted estimator by using an extensive Monte Carlo simulation study. The results show that the performances of the proposed estimators are superior to those based on Liu-type unrestricted estimators. We also applied the proposed estimation methods to Expenditure and Default Data.

A comparison of statistical methods for modeling count data with an application to hospital length of stay

BackgroundHospital length of stay (LOS) is a key indicator of hospital care management efficiency, cost of care, and hospital planning. Hospital LOS is often used as a measure of a post-medical procedure outcome, as a guide to the benefit of a treatment of interest, or as an important risk factor for adverse events. Therefore, understanding hospital LOS variability is always an important healthcare focus. Hospital LOS data can be treated as count data, with discrete and non-negative values, typically right skewed, and often exhibiting excessive zeros. In this study, we compared the performance of the Poisson, negative binomial (NB), zero-inflated Poisson (ZIP), and zero-inflated negative binomial (ZINB) regression models using simulated and empirical data.MethodsData were generated under different simulation scenarios with varying sample sizes, proportions of zeros, and levels of overdispersion. Analysis of hospital LOS was conducted using empirical data from the Medical Information Mart for Intensive Care database.ResultsResults showed that Poisson and ZIP models performed poorly in overdispersed data. ZIP outperformed the rest of the regression models when the overdispersion is due to zero-inflation only. NB and ZINB regression models faced substantial convergence issues when incorrectly used to model equidispersed data. NB model provided the best fit in overdispersed data and outperformed the ZINB model in many simulation scenarios with combinations of zero-inflation and overdispersion, regardless of the sample size. In the empirical data analysis, we demonstrated that fitting incorrect models to overdispersed data leaded to incorrect regression coefficients estimates and overstated significance of some of the predictors.ConclusionsBased on this study, we recommend to the researchers that they consider the ZIP models for count data with zero-inflation only and NB models for overdispersed data or data with combinations of zero-inflation and overdispersion. If the researcher believes there are two different data generating mechanisms producing zeros, then the ZINB regression model may provide greater flexibility when modeling the zero-inflation and overdispersion.

Influencing factors of multiple adverse outcomes among schizophrenia patients using count regression models: a cross-sectional study

BackgroundSchizophrenia patients have increased risks of adverse outcomes, including violent crime, aggressiveness, and suicide. However, studies of different adverse outcomes in schizophrenia patients are limited and the influencing factors for these outcomes need clarification by appropriate models. This study aimed to identify influencing factors of these adverse outcomes by examining and comparing different count regression models.MethodsThis study included schizophrenia patients who had at least one follow-up record in the Guangdong Mental Health Center Network Medical System during 2020. Three types of adverse outcomes were included: a) aggressiveness with police dispatch or violent crime, b) aggressiveness without police dispatch, and c) self-harm or suicide attempts. The incidence density of these adverse outcomes was investigated using the Poisson, negative binomial (NB), zero-inflated Poisson (ZIP), and zero-inflated negative binomial (ZINB) models, accordingly. The best model was chosen based on goodness-of-fit tests. We further analyzed associations between the number of occurrences of adverse outcomes and sociodemographic, clinical factors with the best model.ResultsA total of 130,474 schizophrenia patients were enrolled. Adverse outcomes rates were reported to be less than 1% for schizophrenia patients in 2020, in Guangdong. The NB model performed the best in terms of goodness-of-fit and interpretation when fitting for the number of occurrences of aggressiveness without police dispatch, whereas the ZINB models performed better for the other two outcomes. Age, sex, and history of adverse outcomes were influencing factors shared across these adverse outcomes. Higher education and employment were protective factors for aggressive and violent behaviors. Disease onset aged ≥ 18 years served as a significant risk factor for aggressiveness without police dispatch, and self-harm or suicide attempts. Family history of mental diseases was a risk factor for self-harm or suicide attempts individually.ConclusionsNB and ZINB models were selected for fitting the number of occurrences of adverse outcomes among schizophrenia patients in our studies. Influencing factors for the incidence density of adverse outcomes included both those shared across different types and those individual to specific types. Therefore, comprehensive and customized tools in risk assessment and intervention might be necessary.

Finding the Right Distribution for Highly Skewed Zero-inflated Clinical Data


 Discrete, highly skewed distributions with excess numbers of zeros often result in biased estimates and misleading inferences if the zeros are not properly addressed. A clinical example of children with electrophysiologic disorders in which many of the children are treated without surgery is provided. The purpose of the current study was to identify the optimal modeling strategy for highly skewed, zeroinflated data often observed in the clinical setting by: (a) simulating skewed, zero-inflated count data; (b) fitting simulated data with Poisson, Negative Binomial, Zero-Inflated Poisson (ZIP) and Zero-inflated Negative Binomial (ZINB) models; and, (c) applying the aforementioned models to actual, highlyskewed, clinical data of children with an EP disorder. The ZIP model was observed to be the optimal model based on traditional fit statistics as well as estimates of bias, mean-squared error, and coverage. 

ZINB-Based Graph Embedding Autoencoder for Single-Cell RNA-Seq Interpretations

Single-cell RNA sequencing (scRNA-seq) provides high-throughput information about the genome-wide gene expression levels at the single-cell resolution, bringing a precise understanding on the transcriptome of individual cells. Unfortunately, the rapidly growing scRNA-seq data and the prevalence of dropout events pose substantial challenges for cell type annotation. Here, we propose a single-cell model-based deep graph embedding clustering (scTAG) method, which simultaneously learns cell–cell topology representations and identifies cell clusters based on deep graph convolutional network. scTAG integrates the zero-inflated negative binomial (ZINB) model into a topology adaptive graph convolutional autoencoder to learn the low-dimensional latent representation and adopts Kullback–Leibler (KL) divergence for the clustering tasks. By simultaneously optimizing the clustering loss, ZINB loss, and the cell graph reconstruction loss, scTAG jointly optimizes cluster label assignment and feature learning with the topological structures preserved in an end-to-end manner. Extensive experiments on 16 single-cell RNA-seq datasets from diverse yet representative single-cell sequencing platforms demonstrate the superiority of scTAG over various state-of-the-art clustering methods.

Prevalence trends and risk factors associated with HIV, syphilis, and hepatitis C virus among pregnant women in Southwest China, 2009–2018

ObjectiveThis study investigated prevalence trends and identified the associated factors of HIV, syphilis and hepatitis C virus (HCV) among pregnant women in the Guangxi Zhuang Autonomous Region (Guangxi), Southwest China.MethodsSerial cross-sectional surveys were performed annually among pregnant women in Guangxi from 2009 to 2018. Blood specimens were collected to test the prevalence of HIV, syphilis and HCV. Cochran–Armitage analysis was used to assess the trends of HIV, syphilis and HCV prevalence, as well as the sociodemographic and behavioural data. In this study, we used zero-inflated negative binomial (ZINB) regression models to identify factors associated with HIV, syphilis and HCV infection.ResultsA total of 23,879 pregnant women were included in the study. The prevalence of HIV, syphilis and HCV was 0.24%, 0.85% and 0.19%, respectively. There was a decrease in HIV prevalence from 0.54% to 0.10%, a decrease in HCV prevalence from 0.40% to 0.05% and a decrease in syphilis prevalence from 1.53% to 0.30%. The findings based on the ZINB model revealed that pregnant women who had a history of STI had significantly increased risks of HIV (OR 6.63; 95% CI 1.33–32.90) and syphilis (OR 9.06; 95% CI 3.85–21.30) infection, while pregnant women who were unmarried/widowed/divorced were more likely to have HIV (OR 2.81; 95% CI 1.20–6.54) and HCV (OR 58.12; 95% CI, 3.14–1076.99) infection. Furthermore, pregnant women whose husband had a history of STI (OR 5.62; 95% CI 1.24–25.38) or drug use (OR 7.36; 95% CI 1.25–43.43) showed an increased risk of HIV infection.ConclusionsThere was a relatively low prevalence of HIV, syphilis and HCV among pregnant women. Although decreasing trends in HIV, syphilis and HCV infections were observed, effort is needed to promote STI testing in both premarital medical check-ups and antenatal care, especially targeting couples with a history of STI or drug use.

The Frequency of Primary Healthcare Contacts Preceding the Diagnosis of Lower-Extremity Arterial Disease: Do Women Consult General Practice Differently?

Background. Women with lower-extremity arterial disease (LEAD) are often underdiagnosed, present themselves with more advanced disease at diagnosis, and fare worse than men. Objective. To investigate to what extent potential gender differences exist in the frequency and reasons for general practitioner (GP) consultation six months prior to the diagnosis of LEAD, as potential indicators of diagnostic delay. Methods. Individuals older than 18 years diagnosed with LEAD, sampled from the Julius General Practitioner’s Network (JGPN), were included and compared with a reference population, matched (1:2.6 ratio) in terms of age, sex, and general practice. We applied a zero-inflated negative binomial (ZINB) regression model. Results. The study population comprised 4044 patients with LEAD (43.5% women) and 10,486 subjects in the reference population (46.3% women). In the LEAD cohort, the number of GP contacts was 2.70 (95% CI: 2.42, 3.02) in women and 2.54 (2.29, 2.82) in men. In the reference cohort, 1.77 (95% CI: 1.62, 1.94) in women and 1.63 (95% CI: 1.50, 1.78) in men. In the LEAD cohort, 21.9% of GP contacts occurred one month prior to diagnosis. In both cohorts and both sexes, the most common cause of consultation during the last month before the index date was cardiovascular problems. Conclusions. Six months preceding the initial diagnosis of LEAD, patients visit the GP more often than a similar population without LEAD, regardless of gender. Reported gender differences in the severity of LEAD at diagnosis do not seem to be explained by a delay in presentation to the GP.

FREQUENTIST AND BAYESIAN ZERO-INFLATED REGRESSION MODELS ON INSURANCE CLAIM FREQUENCY: A COMPARISON STUDY USING MALAYSIA’S MOTOR INSURANCE DATA

A no-claim event is a common scenario in insurance and the abundance of no-claim events can be described adequately by zero-inflated models. The zero-inflated Poisson (ZIP) and zero-inflated negative binomial (ZINB) regression models from frequentist and Bayesian approaches are considered for fitting to Malaysia’s motor insurance data. The results from the fittings are compared using mean absolute deviation and mean squared prediction error. The data is categorized into three claim types and the factors considered for regression modelling are coverage type, vehicle age, vehicle cubic capacity and vehicle make. The results from the fittings showed that the ZIP model from both approaches provide better fit than the ZINB model. Also, both ZIP and ZINB models from the Bayesian approach provide better fitting than the frequentist models. Therefore, Bayesian ZIP is the best model in explaining motor insurance claim frequency in Malaysia for all three claim types. From the best regression models, vehicle age, coverage type and vehicle make are the most influential factors in determining the frequency of claim for each claim type. Vehicle age and coverage type have positive effect on the frequency of claim whereas the vehicle make has negative effect on the frequency of claim.

Analyzing the overall effects of the microbiome abundance data with a Bayesian predictive value approach.

The microbiome abundance data is known to be over-dispersed and sparse count data. Among various zero-inflated models, zero-inflated negative binomial (ZINB) model and zero-inflated beta binomial (ZIBB) model are the methods to analyze the microbiome abundance data. ZINB and ZIBB have two sets of parameters, which are for modeling the zero-inflation part and the count part separately. Most previous methods have focused on making inferences in terms of separate case-control effect for the zero-inflation part and the count part. However, in a case-control study, the primary interest is normally focused on the inference and a single interpretation of the overall unconditional mean (also known as the overall effect) of the microbiome abundance in microbiome studies. Here, we propose a Bayesian predictive value (BPV) approach to estimate the overall effect of the microbiome abundance. This approach is implemented based on R package brms. Hence, the parameters in the models will be estimated with two Markov chain Monte Carlo (MCMC) algorithms used in Stan. We performed simulations and real data applications to compare the proposed approach and R package glmmTMB with simulation method in the estimation and inference in terms of the ratio function between the overall effects from two groups in a case-control study. The results show that the performance of the BPV approach is better than R package glmmTMB with the simulation method in terms of lower absolute biases and relative absolute biases, and coverage probability being closer to the nominal level especially when the sample size is small and zero-inflation rate is high.

Application of Count Regression Models for Factors of Antenatal Care (ANC) Utilization in Ethiopia

Adequate antenatal care is necessary to improve the health status of peoples in a country as it contributes the reduction of infant and maternal mortalities. The coverage and the quality of antenatal care plays significant role in maternal health. Ethiopia has witnessed an encouraging improvement in the ANC coverage (at least one visit) in the last decade from 62% in 2016 to 74% in 2019. But only 43% of women had at least four ANC visits during their last pregnancy. The purpose of this study is to identify factors that affect antenatal care visits based on 2019 Ethiopia Mini Demographic and Health Survey (EMDHS) dataset using count regression model. A total of 8,885 women aged 15-49 years were asked for the EMDHS-2019 out of which 3,927 women were illegible and asked for their ANC experiences for their last pregnancy. This empirical study aimed at investigating the impacts of the socioeconomic and demographic characteristics on the utilization of antenatal care. To investigate the impacts of the factors on the number of ANC visits, count models such as Poisson, Negative Binomial (NB), Zero Inflated Poisson (ZIP) and Zero-inflated negative binomial (ZINB) models were examined to select the best fitted model. The selection criterion Akaike information criterion (AIC), Bayesian information criterion (BIC) and Log Likelihood were applied to select the best model. The ZIP model was found to be the best fitted model for the data. Based on ZIP model, the explanatory variables region, age, place of residence, education, religion, wealth index and marital status were statistically significant factors on the frequency of ANC visits. Hence, efforts are still required to put on creating and expanding awareness on the utilization of ANC. The policy makers along with both private sectors and civil societies have to facilitate the increase of awareness and the experiences of the utilization of ANC in consideration of households within different groups of a society. Keywords: EMDHS-2019, antenatal care, Count Regression model DOI: 10.7176/JHMN/101-02 Publication date: June 30 th 2022

Statistical Safety Performance Models considering Pavement and Roadway Characteristics

Transportation agencies build statistical models and predict the average crash frequency to identify hazardous road sections and make informed decisions to reduce crashes. In this paper, safety performance models (SPFs) were built and evaluated considering various pavement and roadway characteristics, including pavement friction, which is seldom available for analysis. Four count data models—Poisson model, negative binomial (NB) model, hurdle-NB model (HNB), and zero-inflated NB (ZINB) model—were built based on roadway characteristics and crash data provided by the Oklahoma Department of Transportation (ODOT). Pavement friction, roadway geometry, surface condition characteristics, and traffic exposure were considered the contributing factors to traffic crashes. Established models were compared in terms of the goodness-of-fit, zero inflation, and statistical significance of factors. The HNB model exhibited promising fitting performance with a manageable number of influencing variables. Coefficients in the HNB model suggest that adequate pavement friction and the presence of shoulders can significantly reduce the crash frequency and thus improve roadway safety performance. Potential issues of the statistical models, such as unobserved heterogeneity and multicollinearity, were also discussed. The relation between roadway infrastructure characteristics (including pavement friction) and roadway safety revealed in this study could assist in choosing the proper statistical model for better decision-making and selecting appropriate preventive treatments for improved roadway safety.

A new Stein estimator for the zero‐inflated negative binomial regression model

AbstractThe Zero‐inflated negative binomial (ZINB) regression models are mainly applied for count data that shows over‐dispersion and extra zeros. Multicollinearity is considered to be a significant problem in the estimation of parameters in the ZINB regression model. Thus, in order to alleviate the serious effects of multicollinearity, a new estimator is proposed which is called ZINB Stein estimator (ZINBSE). We also proposed various biasing parameters for the ZINBSE. A theoretical comparison is also conducted with some existing estimators in the literature. A Monte Carlo simulation study has been considered in order to judge the superiority of the proposed and other estimators, where the estimated mean squared error and mean absolute error are the evaluation criterion. An empirical application is also considered for illustration purposes. Based on the simulation and application results, it is observed that the new ZINBSE with proposed biasing parameters are superior over the other competitors' estimators.

Prediction of Angiography Results Using Logistic Regression and Zeroinflated Negative Binomial Models

Background and Objectives Angiography is a common and invasive method in diagnosing cardiovascular diseases. Some patients refuse to perform angiography due to reasons such as fear, high cost, and lack of confidence in the decision of physician for angiography. This study aims to determine the factors predicting coronary artery occlusion to predict the outcome of angiography. Subjects and Methods In this cross-sectional study, participants were 1187 patients received angiography in Ghaem Hospital in Mashhad, Iran. Demographic data, lipid profile, blood sugar level, and history of underlying disorders were used in two prediction models of logistic regression and zero-inflated negative binomial (NB), fitted using R3.6.1 software. Then, their sensitivity and specificity were compared. Results Of 1187 patients, 404 (34%) had negative angiography. The results of both models showed that the risk of positive angiography was significantly higher in male and diabetic patients. The risk increased with the increase of age. The area under the ROC curve (sensitivity and specificity) for logistic regression and zero-inflated NB models were 78.4(70.4%, 70.5%) and 78.2(71.4%, 71.5%). Conclusion Age, gender, smoking, and history of diabetes are significant predictors of the angiography outcome. There is no significant difference between logistic regression and zero-inflated NB models in predicting the outcome of angiography. Due to the ease of use of logistic regression model, it can be used to predict the results of angiography.

Effects of early preventive dental visits and its associations with dental caries experience: a cross-sectional study

ObjectivesLimited information is known about preventive dental visits (PDVs) before seven years of age among children in China. This study aimed to examine the early PDV rate, identify the impact of PDV on dental caries and untreated dental caries, and explore the factors related to PDV among Chinese sampled children under seven years old.MethodsA cross-sectional survey was conducted in five selected primary health care facilities in Chengdu, China, from May to August 2021. Parent–child dyads during regular systematic medical management were recruited to participate. Children's dental caries were identified through dental examinations and documented as decayed, missing and filled teeth index (dmft) by trained primary care physicians. Dental-related information was collected through a questionnaire. Zero-inflated negative binomial (ZINB) regression was used to test the effect of early PDV on the dmft value, and logistic regression was used to analyse impact factors on the early PDV.ResultsA total of 2028 out of 2377 parent–child dyads were qualified for analysis. Half of the children (50.4%) were male, with a mean age of 4.8 years. Among all the children, 12.1% had their first dental visit for preventive purposes, 34.4% had their first dental visit for symptomatic purposes, and more than half had never visited a dentist. The results showed that a lower dmft value (adjusted OR: 0.69, 95% CI: 0.48–0.84), a higher rate of caries-free (aOR: 6.5, 95% CI: 3.93–10.58), and a lower rate of untreated dental caries (aOR: 0.40, 95% CI: 0.21–0.76) were associated with early PDV utilization. Children who had a higher rate of PDV were positively associated with living in a family with better parental behaviours (aOR: 2.30, 95% CI: 1.71–3.08), better parental oral health perception (aOR: 1.23, 95% CI: 1.06–1.32), fathers who had no untreated caries (aOR: 0.68, 95% CI: 0.47–0.97), families with higher socioeconomic status (aOR: 1.09, 95% CI: 1.04–1.16), and dental health advice received from well-child care physicians (aOR: 1.47, 95% CI: 1.08–2.00).ConclusionsEarly PDV was associated with a lower rate of dental caries prevalence and untreated dental caries among sampled children younger than seven in Western China. Underutilization and social inequities existed in PDV utilization. Public health strategies should be developed to increase preventive dental visits and eliminate social disparities that prevent dental care utilization.

Consumption of Added Sugar among Chinese Toddlers and Its Association with Picky Eating and Daily Screen Time.

Objectives: This study aimed to examine the relationship between daily screen time, picky eating, and consumption frequency of sugared foods and sugar-sweetened beverages (SSBs). Methods: The research data came from the Young Investigation (YI) study conducted in 10 cities in China. This study used sociodemographic information, feeding behavior, picky eating reported by parents, and the consumption frequency of sugared foods and SSBs of 879 toddlers aged 1–3 years. The relationship between daily screen time and picky eating behavior was assessed using logistic regression. The zero-inflated negative binomial (ZINB) model was used to fit the consumption frequencies of sugared foods and SSBs. Results: In all, 13.1% (n = 115) of toddlers did not have sugared foods 1 month before the survey, while 73.3% (n = 644) of toddlers did not have SSBs 1 month before the survey. The consumption rate of sugared foods was relatively higher than SSBs (χ2 = 661.25, p < 0.001). After adjusting for social demographic information, no relationship was found between daily screen time and picky eating (OR = 1.437; 95% CI: 0.990,2.092). The ZINB model showed that, among children who ate sugared foods, children who were picky eaters ate them more often (IRR = 1.133; 95% CI: 1.095,1.172), but no association was found between picky eating and the chance of avoiding sugared foods (OR = 0.949; 95% CI: 0.613,1.471). Children who were picky eaters were less likely not to drink SSBs (OR = 0.664; 95% CI: 0.478,0.921). However, among children who consumed SSBs, picky eaters drank them less frequently (IRR = 0.599; 95% CI: 0.552,0.650). Children with a screen time of no less than 1 h/d ate sugared foods more frequently (IRR = 1.383; 95% CI: 1.164,1.644), and they were less likely to avoid sugared foods (OR = 0.223; 95% CI: 0.085,0.587). The longer the screen time per day was, the less likely children did not have SSBs (<1 h/d: OR = 0.272; 95% CI: 0.130, 0.569; ≥1 h/d: OR = 0.136; 95% CI: 0.057, 0.328). Conclusions: The consumption rate of sugared foods was higher than that of SSBs. Picky eating and daily screen time were related to the consumption frequency of added sugar among Chinese toddlers aged 1–3 years. Picky eaters consumed sugared foods more frequently and were more likely to drink SSBs. Children whose daily screen time reached 1 h/d were more likely to eat sugared foods and drink SSBs.

Identifying features of source and message that influence the retweeting of health information on social media during the COVID-19 pandemic

BackgroundSocial media has become an essential tool to implement risk communication, giving health information could gain more exposure by retweeting during the COVID-19 pandemic.MethodsContent analysis was conducted to scrutinize the official (national and provincial) public health agencies’ Weibo posts (n = 4396) to identify features of information sources and message features (structure, style content). The Zero-Inflated Negative Binomial (ZINB) model was adopted to analyze the association between these features and the frequency of the retweeted messages.ResultsResults indicated that features of source and health information, such as structure, style, and content, were correlated to retweeting. The results of IRR further suggested that compared to provincial accounts, messages from national health authorities’ accounts gained more retweeting. Regarding the information features, messages with hashtags#, picture, video have been retweeted more often than messages without any of these features respectively, while messages with hyperlinks received fewer retweets than messages without hyperlinks. In terms of the information structure, messages with the sentiment (!) have been retweeted more frequently than messages without sentiment. Concerning content, messages containing severity, reassurance, efficacy, and action frame have been retweeted with higher frequency, while messages with uncertainty frames have been retweeted less often.ConclusionsHealth organizations and medical professionals should pay close attention to the features of health information sources, structures, style, and content to satisfy the public’s information needs and preferences to promote the public's health engagement. Designing suitable information systems and promoting health communication strategies during different pandemic stages may improve public awareness of the COVID-19, alleviate negative emotions, and promote preventive measures to curb the spread of the virus.

Discovering single-cell eQTLs from scRNA-seq data only

eQTL studies are essential for understanding genomic regulation. The effects of genetic variations on gene regulation are cell-type-specific and cellular-context-related, so studying eQTLs at a single-cell level is crucial. The ideal solution is to use both mutation and expression data from the same cells. However, the current technology of such paired data in single cells is still immature. We present a new method, eQTLsingle, to discover eQTLs only with single-cell RNA-seq (scRNA-seq) data, without genomic data. It detects mutations from scRNA-seq data and models gene expression of different genotypes with the zero-inflated negative binomial (ZINB) model to find associations between genotypes and phenotypes at the single-cell level. On a glioblastoma and gliomasphere scRNA-seq dataset, eQTLsingle discovered hundreds of cell-type-specific tumor-related eQTLs, most of which cannot be found in bulk eQTL studies. Detailed analyses on examples of the discovered eQTLs revealed important underlying regulatory mechanisms. eQTLsingle is a uniquely powerful tool for utilizing the vast scRNA-seq resources for single-cell eQTL studies, and it is available for free academic use at https://github.com/horsedayday/eQTLsingle.

Relationship between physical activity and mental health in a national representative cross-section study: Its variations according to obesity and comorbidity

BackgroundPhysical activity is positively associated with mental health in obese people with chronic comorbidities. However, how physical activity types (PATs), physical activity frequency (PAF), and physical activity duration (PAD) are associated with mental health need to be further clarified. The current study investigated and compared the effectiveness of PATs, physical activity frequency PAF and PAD for mental health in obese people with various chronic comorbid conditions. MethodsThis cross-sectional study included 871,919 adults who participated in the Behavioral Risk Factor Surveillance System (BRFSS). They were divided into four groups: healthy people, obese people with 0, 1, and 2+ chronic comorbid conditions. The zero-inflated negative binomial (ZINB) regression model and the generalized additive model were used to explore the association between physical activity and mental health burden in the four groups, respectively. ResultsJogging (30.00%), hiking (28.36%) and bicycling (28.32%) have greater improvement in mental health of healthy people; jogging (19.25%), golf (19.95%) and bicycling machine exercise (19.13%) showed a greater improvement in mental health of obese people with no chronic comorbid condition; and aerobic exercise videos or class showed a greater improvement in mental health of obese people with one chronic comorbid condition (22.14%) and obese people with two or more chronic comorbid conditions (19.60%). Non-linear relationships were observed between PAF, PAD, and energy expenditure and mental health. The healthy participants who exercised about 10–15 times a month and 40–50 min per session or about 400–600 METs-min per week had greater benefits for mental health. However, the lowest point of the smooth curve moved to the left with an increasing number of chronic comorbid conditions in obese people. ConclusionsAlmost all PATs were associated with better mental health, but their benefits decreased with increasing number of chronic comorbid conditions in obese people. There were U-shaped relationships between mental health and weekly physical activity frequency, duration, and METs-min.

Distribution‐free model selection for longitudinal zero‐inflated count data with missing responses and covariates

In many medical and social science studies, count responses with excess zeros are very common and often the primary outcome of interest. Such count responses are usually generated under some clustered correlation structures due to longitudinal observations of subjects. To model such longitudinal count data with excess zeros, the zero-inflated binomial (ZIB) models for bounded outcomes, and the zero-inflated negative binomial (ZINB) and zero-inflated poisson (ZIP) models for unbounded outcomes all are popular methods. To alleviate the effects of deviations from model assumptions, a semiparametric (or, distribution-free) weighted generalized estimating equations has been proposed to estimate model parameters when data are subject to missingness. In this article, we further explore important covariates for the response variable. Without assumptions on the data distribution, a model selection criterion based on the expected weighted quadratic loss is proposed to select an appropriate subset of covariates, especially when count responses have excess zeros and data are subject to nonmonotone missingness in both responses and covariates. To understand the selection effects of the percentages of excess zeros and missingness, we design various scenarios for covariate selection in the mean model via simulation studies and a real data example regarding the study of cardiovascular disease is also presented for illustration.