Generalized Linear Latent And Mixed Models Research Articles

Exogenous shock: hierarchical effect of the industrial entrepreneur confidence in the mist of the COVID-19 pandemic

PurposeIn the context of investment decisions, the intricate interplay between exogenous shocks and their influence on investor confidence significantly shapes their behaviors and, consequently, their outcomes. Investment decisions are influenced by uncertainties, exogenous shocks as well as the sentiments and confidence of investors, factors typically overlooked by decision-makers. This study will meticulously examine these multifaceted influences and discern their intricate hierarchical nuances in the sentiments of industrial entrepreneurs during the COVID-19 pandemic.Design/methodology/approachEmploying the robust framework of the generalized linear latent and mixed models (GLLAMM), this research will thoroughly investigate individual and group idiosyncrasies present in diverse data compilations. Additionally, it will delve deeply into the exogeneity of disturbances across different sectors and regions.FindingsRelevant insights gleaned from this research elucidate the adverse influence of exogenous forces, including pandemics and financial crises, on the confidence of industrial entrepreneurs. Furthermore, a significant discovery emerges in the regional analysis, revealing a notable homogeneity in the propagation patterns of industrial entrepreneurs' perceptions within the sectoral and regional context. This finding suggests a mitigation of regional effects in situations of global exogenous shocks.Originality/valueWithin the realm of academic inquiry, this study offers an innovative perspective in unveiling the intricate interaction between external shocks and their significant impacts on the sentiment of industrial entrepreneurs. Furthermore, the utilization of the robust GLLAMM captures the hierarchical dimension of this relationship, enhancing the precision of analyses. This approach provides a significant impetus for data-informed strategic directions.

Understanding Child Wasting in Ethiopia: Cross-sectional Analysis of 2019 Ethiopian Demographic and Health Survey Data Using Generalized Linear Latent and Mixed Models

Wasting is an immediate, visible, and life-threatening form of undernutrition in children aged <5 years. Within a short time, wasting causes recurrent sickness, delayed physical and mental growth, impatience, poor feeding, and low body weight. The long-term consequences of wasting and undernutrition are stunting, inability to learn, poor health status, and poor work performance. Wasting remains a public health problem in Ethiopia. According to the World Health Organization, countries have to reduce undernutrition including child wasting to below 5% by 2025. Ethiopia is attempting to attain national and international targets of undernutrition while struggling with many problems. This study aimed to identify the prevalence and associated factors of wasting to provide information for further renewing policy commitments. We used community-based, cross-sectional data from the Ethiopian Mini Demographic and Health Survey. The survey was conducted in 9 regions and 2 city administrations. Two-stage cluster sampling was used to recruit study participants. In the first stage, enumerations areas were selected, and 28-35 households per enumeration area were selected in the second stage. Our analysis included 2016 women with children aged <5 years from the 2019 EMDHS data set. We dropped incomplete records and included all women who fulfilled the eligibility criteria. We used multilevel ordinal regression using Generalized Linear Latent and Mixed Models (GLLAMM) and predicted probability with log-likelihood ratio tests. Fulfilling the proportional odds model's assumption during the application of multilevel ordinary logistic regression was a cumbersome task. GLLAMM enabled us to perform the multilevel proportional odds model using an alternative method. In our analysis, wasting was 7.68% (95% CI 6.56%-8.93%). Around 26.82% of mothers never used antenatal care for their current child. Most mothers (52.2%) did not have formal education, and 86.8% did not have postnatal care for their children. Additionally, half (50.93%) of the mothers have ≥6 household members. Wasting was associated with feeding diverse foods (coefficient 4.90, 95% CI 4.90-4.98), female sex of the household head (-40.40, 95% CI -40.41 to -40.32), home delivery (-35.51, 95% CI -35.55 to -35.47), first (16.66, 95% CI, 16.60-16.72) and second (16.65, 95% CI 16.60-16.70) birth order, female child (-12.65, 95% CI -12.69 to -12.62), and household size of 1 to 3 (10.86, 95% CI 10.80-10.92). According to the target set by World Health Organization for reducing undernutrition in children aged <5 years to below 5% by 2025, child wasting of 7.68% in Ethiopia should spark an immediate reaction from the government and stakeholders. Informed policy decisions, technology-based child-feeding education, and food self-sufficiency support could improve the current challenges. Additional effort is important to improve low maternal education, family planning, awareness of sex preferences, women empowerment, and maternal health services.

Prevalence and factors associated with complementary feeding practices among children aged 6\u201323\u2009months in India: a regional analysis

BackgroundInappropriate complementary feeding practices significantly contribute to undernutrition among children under 2 years of age in India. However, there is limited up-to-date evidence on the prevalence and factors associated with complementary feeding practices to guide policy actions at the subnational level in India. We investigated the regional prevalence and factors associated with complementary feeding practices in India.MethodsThis study used a sample of 69,464 maternal responses from the 2015–16 National Family Health Survey in India. The prevalence of complementary feeding indicators was estimated using data for each administrative region, namely: North (n = 8469), South (n = 12,828), East (n = 18,141), West (n = 8940), North-East (n = 2422) and Central (n = 18,664). Factors associated with complementary feeding by region in India were investigated using logistic regression Generalized Linear Latent and Mixed Models (GLLAMM) with a logit link and binomial family that adjusted for clustering and sampling weights.ResultsThe study showed a wide variation in the prevalence of introduction of solid, semi-solid or soft foods (complementary foods) among infants aged 6–8 months in regional India; highest in the South (61%) and lowest in the Central and Northern regions (38%). Similarly, minimum dietary diversity (MDD) was highest in the South (33%) and lowest in the Central region (12%). Both minimum meal frequency (MMF) and minimum acceptable diet (MAD) varied substantially across the regions. The factors associated with complementary feeding practices also differed across Indian regions. Significant modifiable factors associated with complementary feeding practices included higher household wealth index for the introduction of complementary foods in the North and Eastern India; higher maternal education for MMF and MDD in the North and Central regions; and frequent antenatal care visits (≥4 visits) for all indicators but for different regions.ConclusionOur study indicates that there are wide differences in regional prevalence and factors associated with complementary feeding practices in India. The improvement of complementary feeding practices in India would require national and sub-national efforts that target vulnerable mothers, including those with no education and limited health service contacts.

Childhood undernutrition in three disadvantaged East African Districts: a multinomial analysis

BackgroundUndernutrition is an important public health indicator for monitoring nutritional status and survival. In spite of its importance, undernutrition is a significant problem health problem in many East African communities. The aim of this study was to identify factors associated with childhood undernutrition in three disadvantaged East African Districts.MethodsWe examined data for 9270 children aged 0–59 months using cross-sectional survey from Gicumbi District in Rwanda, Kitgum District in Uganda and Kilindi District in Tanzania. We considered the level of undernutrition (stunting, wasting and underweight) as the outcome variables with four ordinal categories (severely undernourished, moderately undernourished, mildly undernourished, and nourished). Generalized linear latent and mixed models (GLLAMM) with the mlogit link and binomial family that adjusted for clustering and sampling weights were used to identify factors associated with undernutrition among children aged 0–59 months in three disadvantaged East African Districts.ResultsAfter adjusting for potential confounding factors, the odds of a child being stunted were higher in Gicumbi District in Rwanda while the odds of a child being wasted and underweight were higher in Kitgum District in Uganda. Having diarrhoea two weeks prior to the survey was significantly associated with severe undernutrition. Wealth index (least poor household), increasing child’s age, sex of the child (male) and unavailability of water all year were reported to be associated with moderate or severe stunting/wasting. Children of women who did not attend monthly child growth monitoring sessions and children who had Acute Respiratory Infection (ARI) symptoms were significantly associated with moderate or severe underweight.ConclusionsFindings from our study indicated that having diarrhoea, having ARI, not having water availability all year and not attending monthly child growth monitoring sessions were associated with undernutrition among children aged 0–59 months. Interventions aimed at improving undernutrition in these disadvantaged communities should target all children especially those children from households with poor sanitation practices.

Factors associated with perinatal mortality in Nepal: evidence from Nepal demographic and health survey 2001\u20132016

BackgroundPerinatal mortality is a devastating pregnancy outcome affecting millions of families in many low and middle-income countries including Nepal. This paper examined the more distant factors associated with perinatal mortality in Nepal.MethodsA sample of 23,335 pregnancies > 28 weeks’ gestation from the Nepal Demographic and Health Survey datasets for the period (2001–2016) was analysed. Perinatal Mortality (PM) is defined as the sum of stillbirth (fetal deaths in pregnancies > 28 weeks’ gestation) and early neonatal mortality (deaths within the first week of life), while Extended Perinatal Mortality (EPM) is denoted as the sum of stillbirth and neonatal mortality (deaths within the first 28 days of life). Rates of PM and EPM were calculated. Logistic regression generalized linear latent and mixed models (GLLAMM) that adjusted for clustering and sampling weight was used to examine the factor associated with perinatal mortality.ResultsOver the study period, the PMR was 42 [95% Confidence Interval (CI): 39, 44] per 1000 births for the five-year before each survey; while corresponding EPMR was 49 (95% CI, 46, 51) per 1000 births. Multivariable analyses revealed that women residing in the mountains, who did not use contraceptives, women aged 15–18 years or 19–24 years, and women having no education were associated with increased PM and EPM. The study also identified households using biomass as cooking fuel, and households who reported unimproved sanitation or open defecation were significantly more likely to experience PM and EPM.ConclusionsInterventions aimed to improve use of contraceptives, and reduce biomass as a source of cooking fuel are needed to achieve the recommended target of < 12 perinatal deaths per 1000 births by 2030.

Gender differences in factors associated with prehypertension and hypertension in Nepal: A nationwide survey.

BackgroundNepal has one of the highest prevalence of hypertension in South Asia. However, no national studies have examined the gender differences in the determinants of prehypertension and hypertension in the country to inform targeted interventions. This study aimed to investigate gender differences in factors associated with prehypertension and hypertension in Nepal using the 2016 Nepal Demographic and Health Survey (NDHS).MethodsSociodemographic, behavioural, anthropometric and health status data and information on hypertension were obtained from 14,857 (males: 6,245 and females: 8,612) individuals aged 15 years or above from the biomarker sample of the 2016 NDHS. Factors associated with prehypertension and hypertension by gender were investigated using generalized linear latent and mixed models (GLLAM) with the mlogit link and binomial family that adjusted for clustering and sampling weights.ResultsThe overall prevalence of prehypertension and hypertension was 26.9% [95% confidence interval (CI): 25.7, 28.1] and 17.2% (95% CI 16.1, 18.3), respectively. Prehypertension was present in 30.4% (95%CI: 28.7, 32.2) of males and 24.3% (95% CI: 23.1, 25.6) of females, while hypertension was present in 20.4%, (95% CI 18.9, 22.0) of males and 14.8% (95% CI: 13.7, 16.0) of females. Key modifiable factors that were strongly associated with prehypertension and hypertension in both genders included overweight and obesity, caffeine intake, tobacco use, no schooling, previously informed of hypertension in a health facility, and alcohol consumption (for males). Other significant factors associated with prehypertension and hypertension included increasing age (> 30 years), ecological zone (Hill), Developmental zone (Western) and being married.ConclusionOur results suggest that prehypertension and hypertension were higher in males compared to females. Interventions to improve awareness, screening, treatment and control of prehypertension and hypertension in Nepal are warranted and should target key modifiable factors, as well as people aged 30 years and above.

Infections in hospitalized lupus nephritis patients: characteristics, risk factors, and outcomes.

Background There are limited studies documenting infection epidemiology in lupus nephritis (LN) patients. We aimed to study infection incidences and risk characteristics and ascertain possible predictors of infective outcomes in a multiethnic Asian population. Methods This retrospective study from 2006 to 2012 included newly diagnosed LN patients. We collected admissions, immunosuppression, and infection data until end-stage renal failure, death, last follow-up, or the year 2015. The infection incidence rates were evaluated and a generalized linear latent and mixed models (GLLAMM) analysis was carried out to evaluate potential predictors of infection. Results Our cohort of 101 patients consisted of mainly Chinese (75.2%), females (78.2%), and a median age of 38 years (IQR 26-49). The median time of follow-up was 72.8 months (IQR 49.5-94.0). Six out of 14 patient deaths were infection related. Seventy-eight out of 200 admissions, with a total of 102 episodes of infection, occurred in 46 patients. The incidence of infection rates leading to admissions was 10.3 per 100 person-years and the incidence of infection episodes rates was 17.6 per 100 person-years. Cytomegalovirus (CMV) was the most common organism identified. Univariate GLLAMM analysis showed Malay compared to Chinese ethnicity (HR 2.833 (95% CI 1.088-7.373)), use of oral cyclophosphamide (HR 6.618 (95% CI 1.015-43.154)) and rituximab (HR 3.967 (95% CI 1.157-13.603)) as predictors of infection-related admissions. Malay ethnicity and rituximab remained significant in the multivariate GLLAMM analysis. Conclusions Our study highlights substantial infection incidences in patients with LN and its contribution to deaths.

Reduce mortality risk above all else: a discrete-choice experiment in acute coronary syndrome patients.

Cardiovascular disease is the main cause of death in Germany and other industrialized countries. However, until now, little has been known about how people with acute coronary syndrome (ACS) value aspects of their medical treatment. The objective of this study was to evaluate patients' preferences regarding different antiplatelet medication options following an ACS. After identification of patient-relevant treatment attributes (a literature review and qualitative interviews), a discrete-choice experiment (DCE) including five patient-relevant attributes was conducted. The DCE used a forced-choice approach in which no "opt out" was present, as no treatment is not an option after ACS. The attribute and level combinations were created using a fractional-factorial NGene design with priors. Data analysis was performed using a random-effects logit model. An additional generalized linear latent and mixed models (GLLAMM) analysis was performed to evaluate subgroup differences. ACS patients (N=683) participated in computer-assisted personal interviews. Preference analysis showed a clear dominance of the attribute "mortality risk" (coefficient: 0.803). Ranked second was "side effect: dyspnea" (coefficient: 0.550) followed by "risk of a new myocardial infarction" (coefficient: 0.464) and "side effect: bleeding" (coefficient: 0.400). "Frequency of intake" was less important (coefficient: 0.025). Within the 3-class GLLAMM, the variables "marital status" (p=0.008), "highest level of education" (p=0.003), and "body-mass index" (according to World Health Organization cluster; p=0.014) showed a significant impact on the estimated class probabilities. Our study found "mortality risk" to be of the highest value for patients. Patient-centered care and decision making requires consideration of patient preferences; moreover, the information on preferences can be used to develop effective therapies after an ACS. The data generated will enable healthcare decision makers and stakeholders to understand patient preferences to promote patients' benefit.

What determines individuals' preferences for efficiency over equity-based wages?

Purpose – The paper aims to examine factors that influence individuals' preferences between wages indexed on job performance or efficiency over equity-based wages. Design/methodology/approach – Generalized linear latent and mixed models (GLLAMM) are estimated on the 2005 wave of the World Values Survey on employed individuals from 43 countries. Findings – Results suggest that employees' preference for efficiency-based wages increases with education and globalization, while it decreases with unemployment rates. Research limitations/implications – Institutions and specifically public policies that promote education, and globalization, along with policies that reduce unemployment rates could be used to promote wage setting policies that reward performance or efficiency. Originality/value – The originality of the study lies in its use of both individual- and country-level data to estimate GLLAMM that take into account the multi-level nature of the dataset. This study can inform firms and policymakers on what measures to adopt to promote preferences for efficiency-based wages among individuals.

Factors associated with mosquito net use by individuals in households owning nets in Ethiopia

Ownership of insecticidal mosquito nets has dramatically increased in Ethiopia since 2006, but the proportion of persons with access to such nets who use them has declined. It is important to understand individual level net use factors in the context of the home to modify programmes so as to maximize net use. Generalized linear latent and mixed models (GLLAMM) were used to investigate net use using individual level data from people living in net-owning households from two surveys in Ethiopia: baseline 2006 included 12,678 individuals from 2,468 households and a sub-sample of the Malaria Indicator Survey (MIS) in 2007 included 14,663 individuals from 3,353 households. Individual factors (age, sex, pregnancy); net factors (condition, age, net density); household factors (number of rooms [2006] or sleeping spaces [2007], IRS, women's knowledge and school attendance [2007 only], wealth, altitude); and cluster level factors (rural or urban) were investigated in univariate and multi-variable models for each survey. In 2006, increased net use was associated with: age 25-49 years (adjusted (a) OR = 1.4, 95% confidence interval (CI) 1.2-1.7) compared to children U5; female gender (aOR = 1.4; 95% CI 1.2-1.5); fewer nets with holes (Ptrend = 0.002); and increasing net density (Ptrend < 0.001). Reduced net use was associated with: age 5-24 years (aOR = 0.2; 95% CI 0.2-0.3). In 2007, increased net use was associated with: female gender (aOR = 1.3; 95% CI 1.1-1.6); fewer nets with holes (aOR [all nets in HH good] = 1.6; 95% CI 1.2-2.1); increasing net density (Ptrend < 0.001); increased women's malaria knowledge (Ptrend < 0.001); and urban clusters (aOR = 2.5; 95% CI 1.5-4.1). Reduced net use was associated with: age 5-24 years (aOR = 0.3; 95% CI 0.2-0.4); number of sleeping spaces (aOR [per additional space] = 0.6, 95% CI 0.5-0.7); more old nets (aOR [all nets in HH older than 12 months] = 0.5; 95% CI 0.3-0.7); and increasing household altitude (Ptrend < 0.001). In both surveys, net use was more likely by women, if nets had fewer holes and were at higher net per person density within households. School-age children and young adults were much less likely to use a net. Increasing availability of nets within households (i.e. increasing net density), and improving net condition while focusing on education and promotion of net use, especially in school-age children and young adults in rural areas, are crucial areas for intervention to ensure maximum net use and consequent reduction of malaria transmission.

Which nets are being used: factors associated with mosquito net use in Amhara, Oromia and Southern Nations, Nationalities and Peoples' Regions of Ethiopia

BackgroundThere has been recent large scale-up of malaria control interventions in Ethiopia where transmission is unstable. While household ownership of long-lasting insecticidal nets (LLIN) has increased greatly, there are concerns about inadequate net use. This study aimed to investigate factors associated with net use at two time points, before and after mass distribution of nets.MethodsTwo cross sectional surveys were carried out in 2006 and 2007 in Amhara, Oromia and SNNP regions. The latter was a sub-sample of the national Malaria Indicator Survey (MIS 3R). Each survey wave used multi-stage cluster random sampling with 25 households per cluster (224 clusters with 5,730 households in Baseline 2006 and 245 clusters with 5,910 households in MIS 3R 2007). Net ownership was assessed by visual inspection while net utilization was reported as use of the net the previous night. This net level analysis was restricted to households owning at least one net of any type. Logistic regression models of association between net use and explanatory variables including net type, age, condition, cost and other household characteristics were undertaken using generalized linear latent and mixed models (GLLAMM).ResultsA total of 3,784 nets in 2,430 households were included in the baseline 2006 analysis while the MIS 3R 2007 analysis comprised 5,413 nets in 3,328 households. The proportion of nets used the previous night decreased from 85.1% to 56.0% between baseline 2006 and MIS 3R 2007, respectively. Factors independently associated with increased proportion of nets used were: LLIN net type (at baseline 2006); indoor residual spraying (at MIS 3R 2007); and increasing wealth index at both surveys. At both baseline 2006 and MIS 3R 2007, reduced proportion of nets used was independently associated with increasing net age, increasing damage of nets, increasing household net density, and increasing altitude (>2,000 m).ConclusionThis study identified modifiable factors affecting use of nets that were consistent across both surveys. While net replacement remains important, the findings suggest that: more education about use and care of nets; making nets more resistant to damage; and encouraging net mending are likely to maximize the huge investment in scale up of net ownership by ensuring they are used. Without this step, the widespread benefits of LLIN cannot be realized.

The Role of Health Literacy and Written Medicine Information in Nonsteroidal Antiinflammatory Drug Risk Awareness

Despite widespread use of prescription nonsteroidal antiinflammatory drugs (NSAIDs), patients are commonly unaware of their risks. Pharmacies regularly distribute written medicine information (WMI) describing the risks and benefits of NSAID therapy at the time of dispensing. To clarify the relationship among common sociodemographic factors, education, health literacy, reading of WMI routinely distributed at pharmacies, and NSAID risk awareness. Generalized linear latent and mixed models (GLLAMM) ordered logistic regression and confirmatory path analysis were used to evaluate multivariable relationships in a cross-sectional dataset of 382 patients in the second phase of the Alabama NSAID Patient Safety Study. The majority of the analytical sample was female (72.0%) with 38.7% African American, 38.1% age 65 years or older, and 43.3% reporting at least some college education. Health literacy was positively associated with reading of WMI (p = 0.001) and NSAID risk awareness (p = 0.025), while age was negatively associated with reading WMI (p = 0.001) and NSAID risk awareness (p = 0.005). Medicaid/uninsured status was negatively associated with risk awareness (p = 0.013). Reading of WMI was not associated with NSAID risk awareness (p = 0.659). The final path model demonstrated excellent fit. The lack of relationship between reading of WMI and NSAID risk awareness questions the current strategy of distribution of patient-targeted print education materials at pharmacies. To maximize limited resources, future research should identify more effective strategies to deliver risk information to patients and ensure its retention, especially in high-risk groups such as the elderly, the indigent, and those with inadequate health literacy.

Some applications of generalized linear latent and mixed models in epidemiology: Repeated measures, measurement error and multilevel modeling

-

Generalized multilevel structural equation modeling

A unifying framework for generalized multilevel structural equation modeling is introduced. The models in the framework, called generalized linear latent and mixed models (GLLAMM), combine features of generalized linear mixed models (GLMM) and structural equation models (SEM) and consist of a response model and a structural model for the latent variables. The response model generalizes GLMMs to incorporate factor structures in addition to random intercepts and coefficients. As in GLMMs, the data can have an arbitrary number of levels and can be highly unbalanced with different numbers of lower-level units in the higher-level units and missing data. A wide range of response processes can be modeled including ordered and unordered categorical responses, counts, and responses of mixed types. The structural model is similar to the structural part of a SEM except that it may include latent and observed variables varying at different levels. For example, unit-level latent variables (factors or random coefficients) can be regressed on cluster-level latent variables. Special cases of this framework are explored and data from the British Social Attitudes Survey are used for illustration. Maximum likelihood estimation and empirical Bayes latent score prediction within the GLLAMM framework can be performed using adaptive quadrature in gllamm, a freely available program running in Stata.

Correcting for covariate measurement error in logistic regression using nonparametric maximum likelihood estimation

When covariates are measured with error, inference based on conventional generalized linear models can yield biased estimates of regression parameters. This problem can potentially be rectified by using generalized linear latent and mixed models (GLLAMM), including a measurement model for the relationship between observed and true covariates. However, the models are typically estimated under the assumption that both the true covariates and the measurement errors are normally distributed, although skewed covariate distributions are often observed in practice. In this article we relax the normality assumption for the true covariates by developing nonparametric maximum likelihood estimation (NPMLE) for GLLAMMs. The methodology is applied to estimating the effect of dietary fibre intake on coronary heart disease. We also assess the performance of estimation of regression parameters and empirical Bayes prediction of the true covariate. Normal as well as skewed covariate distributions are simulated and inference is performed based on both maximum likelihood assuming normality and NPMLE. Both estimators are unbiased and have similar root mean square errors when the true covariate is normal. With a skewed covariate, the conventional estimator is biased but has a smaller mean square error than the NPMLE. NPMLE produces substantially improved empirical Bayes predictions of the true covariate when its distribution is skewed.