Bivariate Poisson Regression Models Research Articles

Characterizing Mobility and its Association with HIV Outcomes in Refugee Settlements in Uganda.

A better understanding of refugee mobility is needed to optimize HIV care in refugee settlements. We aimed to characterize mobility patterns among people living with HIV in refugee settlements in Uganda and evaluate the association between mobility and retention in HIV care. Refugees and Ugandan nationals accessing HIV services at seven health centers in refugee settlements across Uganda, with access to a phone, were recruited and followed for six months. Participants received an intake survey and monthly phone surveys on mobility and HIV. Clinic visit and viral suppression data were extracted from clinic registers. Mobility and HIV data were presented descriptively, and an alluvial plot was generated characterizing mobility for participants' most recent trip. Bivariate Poisson regression models were used to describe the associations between long-term mobility (≥1 continuous month away in the past year) and demographic characteristics, retention (≥1 clinic visit/6 months) and long-term mobility, and retention and general mobility (during any follow-up month: ≥2 trips, travel outside the district or further, or spending >1-2 weeks (8-14 nights) away). Mobility data were provided by 479 participants. At baseline, 67 participants (14%) were considered long-term mobile. Male sex was associated with an increased probability of long-term mobility (RR 2.02; 95%CI: 1.30-3.14, p < 0.01). In follow-up, 185 participants (60% of respondents) were considered generally mobile. Reasons for travel included obtaining food or supporting farming activities (45% of trips) and work or trade (33% of trips). Retention in HIV care was found for 417 (87%) participants. Long-term mobility was associated with a 14% (RR 0.86; 95%CI: 0.75-0.98) lower likelihood of retention (p = 0.03). Refugees and Ugandan nationals accessing HIV care in refugee settlements frequently travel to support their survival needs. Mobility is associated with inferior retention and should be considered in interventions to optimize HIV care.

Analyzing Correlated Road Accident Count Data Using Zero Truncated Bivariate Poisson Regression Model

This paper aims to determine the significant factors which influence two correlated count responses, namely the total number of cars involved in an accident and the total number of fatalities due to that accident, of United Kingdom (UK) road accident count data. The bivariate Poisson (BVP) of two different forms and zero truncated bivariate Poisson regression (ZTBVP) models are considered to analyze UK road accident count data and the best model is selected based on the AIC and BIC values. From the data analysis, it is observed that the ZTBVP model provides the best fit (AIC value: 20563.26) for the UK road accident count data compared to all two variants of the BVP model (AIC value: &gt;20563.26). From the results obtained from ZTBVP model, it is also observed that sex of driver, area, serious severity, and light condition are the significant covariates for the total number of cars involved in an accident while area, fatal severity, serious severity, light condition and year 2021 are the significant covariates for the total number of fatalities due to that accident. Dhaka Univ. J. Sci. 72(1): 24-29, 2024 (January)

The newborn screening tests in Brazil: regional and socioeconomic prevalence and inequalities in 2013 and 2019

ObjectiveTo analyze the prevalence evolution of Guthrie, hearing, and eye screening testing among newborns in Brazil, between 2013 and 2019, according to demographic and socioeconomic characteristics. MethodsThis is a cross-sectional study with data from 5231 infants from the Pesquisa Nacional de Saúde (PNS), in 2013, and 6637 infants, in 2019, for the Guthrie test, hearing, and red reflex tests. The authors analyzed the outcomes according to the region of residence, self-reported color/race, having health insurance, and per capita household income. By using bivariate and multivariate Poisson regression models, the prevalence ratios and their respective 95 % Confidence Intervals (CI95%) were calculated for each year. ResultsIn 2013, Guthrie test, hearing, and red reflex tests were performed in 96.5 % (95%CI 95,8;97,0), 65.8 % (95%CI 63,9;67,7), and 60.4 % (95%CI 58,5;62,3) of infants, respectively. In 2019, the prevalence was 97.8 % (95%CI 97,3;98,2) in the Guthrie test, 81.6 % (95%CI 80,3;82,9) in the hearing test, and 78.6 % (95%CI 77,1;79,9) in the red reflex test. The testing frequency was higher among residents of the Southeast and South regions of Brazil, among infants whose mother or guardian was white, had health insurance, and was in the higher income strata; and the most evident differences were in the eye and hearing testing. ConclusionsThe coverage inequalities according to the region of residence, income, and having health insurance highlight the need to use strategies that enable exams to be carried out, with more information about their importance, encompassing actions from primary care, prenatal care to the puerperium, aiming at universal access and equity.

Estimation of zero-inflated bivariate Poisson regression with missing covariates

Zero-inflated bivariate Poisson regression (ZIBP) models are most often applied to correlated bivariate count data that contain a large proportion of zeros. These models have been applied in various fields, such as medical research, health economics, insurance, sports, etc. Because, in practice, some of the covariates involved in ZIBP regression modeling often have missing values, we propose methods for estimating the parameters of the ZIBP regression model with missing at random covariates. Assuming that the selection probability is unknown and estimated non parametrically (by a kernel estimator), we propose inverse probability weighting (IPW) and multiple imputation (MI) methods for estimating the parameters of the ZIBP regression model with missing at random covariates. Asymptotic properties of the proposed estimators are studied under certain regularity conditions. A simulation study is conducted to evaluate the performance of the proposed methods. Finally, the practical application of the proposed methodology is illustrated with data from the health-care field.

Are seniors in mountainous areas able to realize their desired trips? A novel approach to estimate trip demand

Public transport services in mountainous areas are limited, and seniors living in such areas experience various inconveniences in their daily lives. Therefore, the planning of adequate services is crucial in improving the quality of life. Conventionally, public transport services are planned based on observed trip demands, which are referred to as realized trip demands in this study. However, this approach is inadequate because seniors living in an area with insufficient public transport services are generally unable to realize all their desired trips. This study proposes a model to predict the desired trip demand based on the analysis of the travel conditions of seniors in mountainous areas. The proposed model predicts the desired trip demands based on both the realized and unrealized trip demands. The bivariate Poisson regression model is applied to the distribution of desired trip demands and mobility constraints, and the joint probability model, which considers the relationship between them, is proposed. Additionally, in the case where respondents are unable to report the unrealized trip demands accurately, an alternative method is implemented to estimate the desired trip demand based only on the information on whether unrealized trip demand exists, and its estimation results are discussed. The analysis results demonstrate that the conventional trip demand model, established using realized trip demands, underestimates the trip demand in the target area and show that the estimated parameters are biased. The conventional trip demand model neglects 30–50% of trip demands in the target area and cannot provide the trip demand required for appropriate transportation planning. The reproducibility of the desired trip demands can be further improved by extending the model to consider overdispersion or the case where unrealized trip demands have a large share of zero trips or both.

Rest or rust? Complex influence of schedule congestion on the home advantage in the National Basketball Association

The aim of this study was to examine the short-term and long-term effects of schedule congestion on the home advantage (HA) in the National Basketball Association (NBA). A total of 1214 games were considered from teams participating in NBA 2021–22 regular season. A bivariate Poisson regression model were used to estimate the effects of the schedule congestion on the HA. The results revealed that the relationships between the HA and the rest time of home and away team were non-linear in the schedule congestion cycle. Too much rest time might be detrimental to the team performance, especially for the away team. In addition, the accumulation of fatigue and incomplete physical recovery due to schedule congestion could have a season-long impact on the HA. The findings offered a new insight about the effect of schedule congestion on the HA and teams’ performance in NBA and might be useful for load management to reduce the disadvantage from schedule congestion, as well as the risk of athletes’ injury.

Sports Data Analysis by using Bivariate Poisson Models in the Bayesian Framework

Bivariate distribution models are commonly used to analyze sports data and data from various fields. These models are used to analyze discrete count data with two dependent variables in the data. In this research article, we have used Bivariate Poisson and Diagonally Inflated Bivariate Poisson regression models. We have proposed an estimation procedure in the Bayesian framework in conjunction with the augmentation of data. For parameter estimation, we use Gaussian priors and beta priors for both models. To illustrate the fitting performances of our suggested models we have performed real data analysis on English Premier League soccer data.

Psychotropic drug use and suggestive depression symptoms associated with NR3C1 DNA methylation

Introduction: Psychiatric disorders have become a global problem that leads millions of people to use psychotropic medications, especially benzodiazepines. The effects of these substances are widely known regarding tolerance and chemical dependence, however, from epigenetics perspective, there are still little known.&#x0D; Objective: To evaluate the association between psychotropic drug use, NR3C1 gene methylation and its relation with symptoms suggestive of depression in adult individuals assisted in the public health system.&#x0D; Methods: 385 adult volunteers (20-59 years) users of the Brazilian Unified Health System were recruited to evaluate socioeconomic, health, lifestyle conditions in a cross sectional study. BDI-II evaluated symptoms suggestive of depression and pyrosequencing evaluated NR3C1 DNA methylation. Bivariate and multivariate Poisson regression model with robust variance (p &lt; 0.05) evaluated the association between psychotropic drug use and NR3C1 gene methylation.&#x0D; Results: Specific depressive symptoms such as irritability, insomnia and fatigability were associated with psychotropic drug use. Symptoms of past failure, indecision and loss of appetite were associated with hypermethylation patterns in CpGs 40 to 47 of NR3C1 gene. Moreover, psychotropic drug use is associated with 50% reduction in NR3C1 gene methylation, through model adjusted with socioeconomic, health and lifestyle confounding variables.&#x0D; Conclusions: Psychotropic drug use and depressive symptoms was associated with changes in NR3C1 DNA methylation. In this context, epigenetic modification resulting from psychotropic drug use and depressive symptoms could be considered, mainly in population studies with epigenetic evaluation, where these factors may be influencing the findings of future studies.

Zero-Dependent Bivariate Poisson Distribution with Applications

The bivariate Poisson model is the most widely used model for bivariate counts, and in recent years, several bivariate Poisson regression models have been developed in order to analyse two response variables that are possibly correlated. In this paper, a particular class of bivariate Poisson model, developed from the bivariate Bernoulli model, will be presented and investigated. The proposed bivariate Poisson models use dependence parameters that can model positively and negatively correlated data, whereas more well-known models, such as Holgate’s bivariate Poisson model, can only be used for positively correlated data. As a result, the proposed model contributes to improving the properties of the more common bivariate Poisson regression models. Furthermore, some of the properties of the new bivariate Poisson model are outlined. The method of maximum likelihood and moment method were used to estimate the parameters of the proposed model. Additionally, real data from the healthcare utilization sector were used. As in the case of healthcare utilization, dependence between the two variables may be positive or negative in order to assess the performance of the proposed model, in comparison to traditional bivariate count models. All computations and graphs shown in this paper were produced using R programming language.

Frailty and Hospitalization Burden in Patients With Chronic Heart Failure

Frailty is associated with increased mortality and hospitalizations in patients with heart failure (HF). However, there is little evidence regarding the burden of morbidity. In this study, we aimed to assess the association between frailty and recurrent all-cause HF hospitalizations in patients with stable chronic HF. This was an observational and prospective study that enrolled HF outpatients followed in a specialized HF unit of a single tertiary care center from 2017 to 2019. Frailty was assessed by Fried criteria. Robustness, prefrailty, and frailty were defined as 0, 1 to 2, and ≥3, respectively. The independent association between frailty status and recurrent hospitalizations was assessed through Famoye's bivariate Poisson regression model, and risk estimates were expressed as incidence rate ratios (IRR). A total of 277 patients were included. The mean age was 74 ± 10 years, 118 were women (42.6%), and 131 patients (47.3%) had left ventricular ejection fraction ≥50. According to Fried's score 61 patients (22%) were robust, 95 patients (34%) were prefrail, and 121 patients (44%) were frail. After a median follow-up of 2.21 (1.6 to 2.8) years, 52 patients (19%) died. We registered 348 all-cause hospitalizations in 144 patients (52%) and 178 HF hospitalizations in 108 patients (39%). Compared with robust patients, frailty was associated with a higher risk of all-cause and HF recurrent hospitalizations in multivariable analysis (IRR 2.01, 95% confidence interval 1.14 to 3.57, p=0.017 and IRR 2.25, 95% confidence interval 1.16-4.36, p=0.016, respectively). In conclusion, in patients with chronic HF, frailty identifies patients with an increased risk of total and recurrent all-cause and HF hospitalizations.

Patterns of progression of cerebral small vessel disease markers in older adults of Amerindian ancestry: a population-based, longitudinal prospective cohort study.

BackgroundProgression of cerebral small vessel disease (cSVD) markers has been studied in different races/ethnic groups. However, information from individuals of Amerindian ancestry is lacking. We sought to evaluate progression patterns of cSVD markers in community-dwelling older adults of Amerindian ancestry.MethodsFollowing a longitudinal prospective study design, participants of the Atahualpa Project Cohort aged ≥ 60 years received a baseline brain MRI and clinical interviews. Those who also received a brain MRI at the end of the study were included. Poisson regression models were fitted to assess cSVD markers progression according to their baseline load after a median follow-up of 6.5 ± 1.4 years. Logistic regression models were fitted to assess interrelations in the progression of the different cSVD markers at the end of the study.ResultsThe study included 263 individuals (mean age: 65.7 ± 6.2 years). Progression of white matter hyperintensities (WMH) was noticed in 103 (39%) subjects, cerebral microbleeds in 25 (12%), lacunes in 12 (5%), and enlarged basal ganglia-perivascular spaces (BG-PVS) in 56 (21%). Bivariate Poisson regression models showed significant associations between WMH severity at baseline and progression of WMH and enlarged BG-PVS. These associations became non-significant in multivariate models adjusted for clinical covariates. Logistic regression models showed interrelated progressions of WMH, cerebral microbleeds and enlarged BG-PVS. The progression of lacunes was independent.ConclusionsPatterns of cSVD marker progression in this population of Amerindians are different than those reported in other races/ethnic groups. The independent progression of lacunes suggests different pathogenic mechanisms with other cSVD markers.

Prevalence of anti-SARS-CoV-2 antibodies and associated factors in healthcare workers of a Mexican Covid-19 hospital

To determine the prevalence of SARS-CoV-2 antibodies among healthcare workers (HCW) and to identify factors associated with infection. Materials and meth-ods. A cross-sectional study was conducted in a Covid-19 hospital in Morelos, Mexico. Antibodies against SARS-CoV-2 spike and nucleocapsid proteins were detected by ELISA. A bivariate and multivariable Poisson regression model were performed to identify factors associated with infection. Among all participants, 31% had anti-SARS-CoV-2 antibodies, while only 13.1% had reported a history of positive RT-PCR. Individuals who reported cohabiting with someone with Covid-19, and those who had a previous RT-PCR test, were more likely to be seropositive. Laboratory personnel had the lowest seroprevalence (12.0%), while social workers had the highest (35.7%). The results of this study show the seroprevalence of SARS-CoV-2 antibodies among HCW in a hospital in Mexico, and underline the importance of serological tests for a better estimate of prevalence in health systems where only symptomatic cases are recorded.

COVID-19 Vaccine Uptake among US Adults According to Standard Occupational Groups.

This cross-sectional ecological study examined the relationship between neighborhood-level standard occupational groups in the USA and COVID-19 vaccine uptake using 774 census tract data, each consisting of approximately 1600 housing units. The neighborhood-level COVID-19 vaccination uptake data were retrieved from Harris County Public Health, Harris County, Texas. The standard occupational group data were from the US Census Bureau. We calculated the incidence rate ratios (IRRs) for vaccine uptake using bivariate and multivariable Poisson regression models. In the adjusted models, we found that the healthcare practitioner/technician (IRR: 1.008; 95% CI: 1.003–1.014; p = 0.001), business/management/legal (IRR: 1.011; 95% CI: 1.008–1.013; p < 0.001), computer/engineering/life/physical/social science (IRR: 1.018; 95% CI: 1.013–1.023; p < 0.001), and arts/design/entertainment/sports/media (IRR: 1.031; 95% CI: 1.018–1.044; p < 0.001) occupational groups were more likely to have received the full regimen of a COVID-19 vaccine. On the contrary, the building/installation/maintenance/repair (IRR: 0.991; 95% CI: 0.987–0.995; p < 0.001), construction/extraction/production (IRR: 0.991; 95% CI: 0.988–0.995; p < 0.001), transportation/material moving (IRR: 0.992; 95% CI: 0.987–0.997; p = 0.002), food preparation/serving related (IRR: 0.995; 95% CI: 0.990–0.999; p = 0.023), and personal care/services (IRR: 0.991; 95% CI: 0.985–0.998; p = 0.017) groups were less likely to have received the complete dose of a COVID-19 vaccine. White-collar workers were more likely to be vaccinated than blue-collar workers. We adjusted for age, sex, and race/ethnicity in the multivariable analysis. The low vaccine uptake among certain occupational groups remains a barrier to pandemic control. Engaging labor-centered stakeholders in the development of vaccination interventions may increase uptake.

Copula-based bivariate finite mixture regression models with an application for insurance claim count data

Modeling bivariate (or multivariate) count data has received increased interest in recent years. The aim is to model the number of different but correlated counts taking into account covariate information. Bivariate Poisson regression models based on the shock model approach are widely used because of their simple form and interpretation. However, these models do not allow for overdispersion or negative correlation, and thus, other models have been proposed in the literature to avoid these limitations. The present paper proposes copula-based bivariate finite mixture of regression models. These models offer some advantages since they have all the benefits of a finite mixture, allowing for unobserved heterogeneity and clustering effects, while the copula-based derivation can produce more flexible structures, including negative correlations and regressors. In this paper, the new approach is defined, estimation through an EM algorithm is presented, and then different models are applied to a Spanish insurance claim count database.

Risk factors associated with hand tractor related injuries among rice farmers in Thailand

Hand tractors are important machines used in the soil preparation process before sowing rice in the paddy’s of Thailand. Previous research on injuries in rice farming indicated that working with a hand tractor may be one of main causes of injuries. This study investigated the prevalence of hand tractor-related injuries, their association risk factors and characteristic of injuries. The data were collected by face-to-face interview to 377 farmers from nearly all districts in Phra Nakhon Si Ayutthaya province. Among these rice farmers, the prevalence of injury was 41%. The most common type of injury was cuts (43.9%) and the most common source of injuries was stepping on a golden apple snail shell (33.1%). The bivariate Poisson regression models showed that self-reported normal working conditions that were associated with the risk of injury included sleeping problems (RR = 1.39), “Hustle Work (working quickly to complete plowing)” (RR = 1.48), feeling fatigue before work (RR = 1.60), and normal use of a leveler as attached plow equipment (RR = 1.41). However, a multivariate model showed only normal use of a leveler as attached equipment was associated with an increased use of injuries (RR = 1.47) after controlling for the other factors that were significant in the bivariate models. These results suggest that job stress protection should be recommended or redesign of the leveler attachment developed for hand tractors used in rice paddy fields.

Bivariate Mixed Poisson Regression Models with Varying Dispersion

The main purpose of this article is to present a new class of bivariate mixed Poisson regression models with varying dispersion that offers sufficient flexibility for accommodating overdispersion and accounting for the positive correlation between the number of claims from third-party liability bodily injury and property damage. Maximum likelihood estimation for this family of models is achieved through an expectation-maximization algorithm that is shown to have a satisfactory performance when three members of this family, namely, the bivariate negative binomial, bivariate Poisson–inverse Gaussian, and bivariate Poisson–Lognormal distributions with regression specifications on every parameter are fitted on two-dimensional motor insurance data from a European motor insurer. The a posteriori, or bonus-malus, premium rates that are determined by these models are calculated via the expected value and variance principles and are compared to those based only on the a posteriori criteria. Finally, we present an extension of the proposed approach with varying dispersion by developing a bivariate Normal copula-based mixed Poisson regression model with varying dispersion and dependence parameters. This approach allows us to consider the influence of individual and coverage-specific risk factors on the mean, dispersion, and copula parameters when modeling different types of claims from different types of coverage. For expository purposes, the Normal copula paired with negative binomial distributions for marginals and regressors on the mean, dispersion, and copula parameters is fitted on a simulated dataset via maximum likelihood.

Intimate partner violence against women and child maltreatment in a Brazilian birth cohort study: co-occurrence and shared risk factors

BackgroundIntimate partner violence (IPV) against women and child maltreatment (CM) are major public health problems and human rights issues and may have shared causes. However, their overlap is understudied. We...

Estimation in the zero-inflated bivariate Poisson model, with an application to health-care utilization data

Data on the demand for medical care is usually measured by a number of different counts. These count data are most often correlated and subject to high proportions of zeros. However, excess zeros and the dependence between these data can jointly affect several utilization measures.In this paper, the zero-inflated bivariate Poisson regression model (ZIBP) was used to analyze health-care utilization data. First, the asymptotic properties of the maximum likelihood estimator (MLE) of this model were investigated theoretically.Then, a simulation study is conducted to evaluate the behaviour of the estimator in finite samples. Finally, an application of the ZIBP model to health care demand data is provided by way of illustration.

Estimation in the zero-inflated bivariate Poisson model, with an application to health-care utilization data

Data on the demand for medical care is usually measured by a number of different counts. These count data are most often correlated and subject to high proportions of zeros. However, excess zeros and the dependence between these data can jointly affect several utilization measures.In this paper, the zero-inflated bivariate Poisson regression model (ZIBP) was used to analyze health-care utilization data. First, the asymptotic properties of the maximum likelihood estimator (MLE) of this model were investigated theoretically. Then, a simulation study is conducted to evaluate the behaviour of the estimator in finite samples. Finally, an application of the ZIBP model to health care demand data is provided by way of illustration.

Geographically weighted bivariate generalized Poisson regression: application to infant and maternal mortality data

Bivariate generalized Poisson regression (BGPR) is an extension of bivariate Poisson regression which deals overdipersion or underdispersion problem. This model gives global regression coefficients for all observations (locations) in the analysis. The BGPR model is then extended to take into account spatial heterogeneity, called geographically weighted bivariate generalized Poisson regression model, that yields varying regression coefficients locally. The regression model is applied to analyse factors affecting number of infant and maternal mortality in East Java, Indonesia.