Modeling conditional dependence between diagnostic tests: A multiple latent variable model

Applications of latent class analysis in diagnostic test studies have assumed that all tests are measuring a common binary latent variable, the true disease status. In this article we describe a new approach that recognizes that tests based on different biological phenomena measure different latent variables, which in turn measure the latent true disease status. This allows for adjustment of conditional dependence between tests within disease categories. The model further allows for the inclusion of measured covariates and unmeasured random effects affecting test performance within latent classes. We describe a Bayesian approach for model estimation and describe a new posterior predictive check for evaluating candidate models. The methods are motivated and illustrated by results from a study of diagnostic tests for Chlamydia trachomatis. Published in 2008 by John Wiley & Sons, Ltd.     

Binary latent variable modelling and its applicationin the study of air pollution in Hong Kong

A binary latent variable is constructed to account for the correlation between multiple binary outcomes and is treated as a dependent variable in modelling for covariate effects. This modelling method is similar to the structural equation modelling. Three models are considered: (1) all covariates affecting the binary latent variable directly; (2) some covariates affecting the binary latent variable while other affecting the manifestation of the binary outcomes; and (3) no covariates are included. Gibbs sampling, a special case of the Markov Chain Monte Carlo method, is used to estimate the parameters in the models. Simulation studies show that this method is valid and reliable in estimating covariate effects. But Model (1) fitted the data best with lowest value in the deviance information criteria. The method is illustrated by applying it to the data analysis of an environmental air pollution study. The results show that air pollution (i.e. the most versus the least polluted district) (odds ratio 1.20; 95% confidence interval 0.97-1.49; p=0.102), smoking (relative to not smoking) (2.75; 2.21-3.41; p < 0.001) and mosquito coil use (relative to non-use) (1.27; 0.99-1.62; p=0.058) had an impact on the respiratory health of male adults in Hong Kong.     

Women’s Longitudinal Smoking Patterns from Preconception Through Child’s Kindergarten Entry: Profiles of Biological Mothers of a 2001 US Birth Cohort

To identify longitudinal patterns of women’s smoking during the pre-conception, perinatal, and early parenting period and describe risk factors distinguishing the different profiles. We conducted longitudinal latent class analysis of maternal smoking status over a 6–7 year period in a sample of 8,650 biological mothers of the Early Childhood Longitudinal Study-Birth Cohort, nationally representative of US births in 2001. Five latent classes were identified: pregnancy-inspired quitters (4.3 %), delayed initiators (5.1 %), persistent smokers (8.5 %), temporary quitters (10.4 %), and nonsmokers (71.7 %). These classes were distinguished by age, race/ethnicity, education, poverty status, marital status, parity, drinking behavior, and depression. For example, when compared to those with college degrees, those with less than a high school degree were at least five times as likely to be in the delayed initiator, temporary quitter, or persistent smoker classes (vs. the nonsmoker class). Heterogeneous longitudinal smoking patterns indicate the need for both prevention messages and cessation treatment continuing past parturition, tailored to fit individual profiles in order to achieve better health outcomes for both mothers and children.     

Effects of pre- and postnatal exposure to parental smoking on early childhood respiratory health

Compelling evidence suggests a causal relation between parental smoking during and after pregnancy and adverse effects on respiratory health in the offspring. The authors' aim in this study was to disentangle the effects of prenatal and postnatal smoking on early childhood respiratory health. Most parents who smoke during pregnancy continue to smoke postpartum, and it is difficult to identify sufficiently large subgroups of children who were exclusively exposed in utero or after birth. This study was based on the first 22,390 children born between 2000 and 2004 in the Norwegian Mother and Child Cohort, a pregnancy cohort designed to eventually include 100,000 pregnancies. Data were collected through detailed questionnaires administered at various stages of pregnancy, starting in early pregnancy. Because of the large study population, the authors were able to disentangle the pre- and postnatal effects of parental smoking on wheeze and lower respiratory tract infection in the children's first 18 months of life. They found maternal smoking in pregnancy to be an independent risk factor for wheeze and respiratory infection. Postnatal paternal smoking was also associated with these outcomes, independently of maternal smoking in pregnancy.     

Parental smoking and respiratory illnesses in Australian children aged 0–4 years: ABS 1989–90 National Health Survey results

Objective: This study investigated the associations between parental smoking and respiratory infections in Australian children aged 0–4 years.
Methods: Data from the ABS 1989–90 National Health Survey were used. The exposure variables examined were maternal, paternal and combined family smoking. Outcome variables were parent-reported chronic or recent asthma, asthma wheeze, bronchitis, influenza, common cold, cough, otitis media and other respiratory conditions. Logistic regression techniques were used to control for confounding by socio-economic status, child's sex, maternal education, place of residence, ethnicity and family size.
Results: Of the 4,281 children in the sample, 45% lived in households with one or more current smokers and 29% had a mother who smoked. Maternal (but not paternal) smoking was significantly associated with asthma (OR 1.52, 95% CI 1.19–1.94) and asthma wheeze (OR 1.51, 95% CI 1.26–1.80). No other significant associations were observed. Positive and significant dose response relationships were found between the amount of maternal smoking and both asthma variables. Population attributable risks were calculated and almost 13% of asthma and asthma wheeze in 0–4 year old Australian children in 1989–90 was estimated to be due to maternal smoking.
Conclusion: Large numbers of Australian children live in households with smokers. This study, like others, has shown an association between maternal smoking and respiratory illnesses in young children. Further strategies are needed to prevent or reduce young children's exposure to environmental tobacco smoke in their homes.     

A robust goodness-of-fit test statistic with application to ordinal regression models

We propose a goodness-of-fit test statistic for linear regression with heterogeneous variance, which is asymptotically chi-square if the given model is correct. The test statistic is computed as a quadratic form of observed minus predicted responses. We apply the method to a linear regression for an ordinal categorical response, the wheezing status of a child (no wheeze, wheeze with cold, wheeze apart from cold) as a function of maternal smoking and city of residence.     

Influence of parental smoking on respiratory symptoms during the first decade of life: the Tucson Children's Respiratory Study.

Compelling evidence suggests a causal relation between exposure to parental cigarette smoking and respiratory symptoms during childhood. Still, the roles of prenatal versus postnatal parental smoking need clarification. In this study, the authors assessed the effects of passive smoking on respiratory symptoms in a cohort of over 1,000 children born during 1980-1984. The children were enrolled in the Tucson Children's Respiratory Study in Tucson, Arizona, and were followed from birth to age 11 years. The population was generally middle class and consisted of two main ethnic groups, non-Hispanic Whites (75%) and Hispanics (20%), reflecting Tucson's population. Information on parental smoking and on wheeze and cough in their children was elicited from parents by using questionnaires at five different surveys. Data were analyzed both cross-sectionally and by using the generalized estimation equation approach, a longitudinal mixed-effects model. The best-fitting model indicated that maternal prenatal but not postnatal smoking was associated with current wheeze (odds ratio = 2.3, 95% confidence interval 1.4-3.8) independently of a family history of asthma, socioeconomic factors, and birth weight. This effect was time dependent and significant only below age 3 years; although independent of gender, the association was stronger for girls (odds ratio = 3.6, 95% confidence interval 1.6-8.0). Cough was not associated with parental smoking during the first decade of life. This transitory effect of maternal prenatal smoking on wheezing could be due to changes that affect the early stages of lung development.     

Correlation between maternal age, social class and smoking, and birth weight]

An epidemiological survey was carried out in Ribeir?o Preto, Brazil, from June 1978 to May 1979. Interviews were held with mother of singleton live borne children, delivered in hospitals, which accounted for 98% of all births in the area. The higher percentages of low birthweight children related to the offspring of smokers, young mother and women belonging to the working class. The majority of young women were found in the working class and the prevalence of smoking was higher in the group of women below 20 years of age. There was no statistical difference in the smoking habit as between different social classes. A larger number of low birthweight children were observed in nonsmoking women of the working class than among women smokers of the middle class. A log model was adjusted to the data in order to study the possible multiple association of smoking, maternal age and social class with birthweight. The results indicated that maternal smoking, maternal age and social class had independent effects on birthweight. The was no interaction between them. These findings suggest that the higher prevalence of low birthweight in nonsmoking mothers of the working class in relation to smoking mothers of the middle class probably reflects clustering of other risk factors-such as poor education inadequate prenatal care, high parity and differences in reproductive behavior in women of the working class.     

Comparing self and maternal reports of adolescents’ general health status: Do self and proxy reports differ in their relationships with covariates?

Purpose

Given that mothers often—but do not always—report children’s health status in surveys, it is essential to gain an understanding of whether the relationship between children’s general health status and relevant covariates depends on who reports children’s general health status.

Methods

Using data from the first wave of the National Longitudinal Study of Youth 1997 cohort (N = 6,466), a nationally representative sample of adolescents in the United States ages 12 to 17 in 1997, the study first examined the concordance between self and maternal reports of adolescents’ general health status. Then, self and maternal reports of adolescents’ general health status were each regressed on health-relevant covariates, and tests of differences in coefficients across the models were estimated.

Results

Self and maternal reports of adolescents’ general health status are moderately concordant. Furthermore, the associations of adolescents’ general health status with adolescent BMI and the adolescent being female significantly differ across reporters, such that the negative relationships are even more negative with self compared to maternal reports of adolescents’ general health status. The associations of adolescents’ general health status with the measures of adolescents’ health limitations, maternal self-rated health, and certain sociodemographic covariates differ across reporters, such that each has a greater relationship with maternal compared to self-reports of adolescents’ general health status.

Conclusion

The results are important for interpreting research on the causes and consequences of child and adolescent health, as results across studies may not be comparable if the reporter is not the same.     

Prevalence of asthma and wheeze in relation to passive smoking in Japanese children

PURPOSE: Evidence remains inconclusive as to whether environmental tobacco smoke is a risk factor for allergic disorders in childhood. The present large-scale cross-sectional study examined the relationship between passive smoking at home and the prevalence of allergic disorders in Japanese schoolchildren. METHODS: Study subjects were 23,044 children aged 6 to 15 years in Okinawa. Outcomes were based on diagnostic criteria from the International Study of Asthma and Allergies in Childhood. Adjustment was made for sex, age, region of residence, number of siblings, paternal and maternal history of asthma, atopic eczema, or allergic rhinitis, as well as paternal and maternal educational level. RESULTS: The prevalence of wheeze, asthma, atopic eczema, and allergic rhinoconjunctivitis in the previous 12 months was 10.7%, 7.6%, 6.8%, and 7.7%, respectively. Current heavy passive smoking and 7.0 or more pack-years of smoking in the household were independently related to an increased prevalence of wheeze and asthma, especially in children 6 to 10 years of age and children with a positive parental allergic history. There was no dose-response relationship between pack-years of smoking in the household and atopic eczema or allergic rhinoconjunctivitis. CONCLUSIONS: Our findings suggested that environmental tobacco smoke might be associated with an increased prevalence of wheeze and asthma in Japanese children.     

Lack of association of mercury with risk of wheeze and eczema in Japanese children: the Osaka Maternal and Child Health Study

Mercury can have profound and complicated effects on the immune system, and epidemiological evidence regarding the relationship between mercury exposure and allergic disorders has been sparse. We investigated the associations between mercury levels in maternal and children's hair and the risk of wheeze and eczema in Japanese children at 29–39 months of age. Study subjects were 582 Japanese mother–child pairs. Presence or absence of wheeze and eczema symptoms was determined based on the criteria of the International Study of Asthma and Allergies in Childhood. Adjustment was made for maternal age; residential municipality at baseline; maternal and paternal education; maternal and paternal history of allergic disorders; maternal energy-adjusted fish intake during pregnancy; maternal smoking during pregnancy; number of child's older siblings; child's sex; household smoking in the same room as the child; breastfeeding duration; and children's fish intake at the fourth survey. The prevalence of wheeze and eczema was 18.6% and 17.2%, respectively. The range of hair mercury levels was 0.26−6.05 μg/g in mothers and 0.13−9.51 μg/g in children. Neither maternal nor children's hair mercury levels were related to the risk of wheeze or eczema. Maternal and children's hair mercury levels in the second quartile were non-significantly inversely related to the risk of wheeze (adjusted odds ratios [95% confidence intervals] were 0.77 [0.41−1.44] and 0.57 [0.29−1.11], respectively) while those in the third quartile were non-significantly inversely associated with the risk of eczema (adjusted odds ratios [95% confidence intervals] were 0.77 [0.40−1.45] and 0.66 [0.33−1.30], respectively). The current study provides no evidence that hair mercury levels in either mothers or children are positively associated with the risk of wheeze or eczema in children aged 29–39 months in Japan, where fish intake is high.     

The Biplot as a diagnostic tool of local dependence in latent class models. A medical application

Latent class models (LCMs) can be used to assess diagnostic test performance when no reference test (a gold standard) is available, considering two latent classes representing disease or non-disease status. One of the basic assumptions in such models is that of local or conditional independence: all indicator variables (tests) are statistically independent within each latent class. However, in practice this assumption is often violated; hence, the two-LCM fits the data poorly. In this paper, we propose the use of Biplot methods to identify the conditional dependence between pairs of manifest variables within each latent class. Additionally, we propose incorporating such dependence in the corresponding latent class using the log-linear formulation of the model.     

Early life patterns of common infection: a latent class analysis

Early life infection has been implicated in the aetiology of many chronic diseases, most often through proxy measures. Data on ten infectious symptoms were collected by parental questionnaire when children were 6 months old as part of the Avon Longitudinal Study of Parents and Children, United Kingdom. A latent class analysis was used to identify patterns of infection and their relationship to five factors commonly used as proxies: sex, other children in the home, maternal smoking, breastfeeding and maternal education. A total of 10,032 singleton children were included in the analysis. Five classes were identified with differing infectious disease patterns and children were assigned to the class for which they had a highest probability of membership based on their infectious symptom profile: ‘general infection’ (n = 1,252, 12.5%), ‘gastrointestinal’ (n = 1,902, 19.0%), ‘mild respiratory’ (n = 3,560, 35.5%), ‘colds/ear ache’ (n = 462, 4.6%) and ‘healthy’ (n = 2,856, 28.5%). Females had a reduced risk of being in all infectious classes, other children in the home were associated with an increased risk of being in the ‘general infection’, ‘mild respiratory’ or ‘colds/ear ache’ class. Breastfeeding reduced the risk of being in the ‘general infection’ and ‘gastrointestinal’ classes whereas maternal smoking increased the risk of membership. Higher maternal education was associated with an increased risk of being in the ‘mild respiratory’ group. Other children in the home had the greatest association with infectious class membership. Latent class analysis provided a flexible method of investigating the relationship between multiple symptoms and demographic and lifestyle factors.     

Attendance Trajectory Classes Among Divorced and Separated Mothers and Fathers in the New Beginnings Program

We examined attendance trajectories among mothers and fathers in the effectiveness trial of the New Beginnings Program, a parenting-focused prevention program for divorced and separated parents. We also investigated attendance trajectory class differences on two sets of pretest covariates: one set previously linked to participation in programs not specifically targeting divorced parents (i.e., sociodemographics, perceived parenting skills, child problem behaviors, parent psychological distress) and another that might be particularly salient to participation in the context of divorce (i.e., interparental conflict, level of parent-child contact, previous marital status to the ex-spouse). For mothers and fathers, results supported four attendance trajectory classes: (1) non-attenders (NA), (2) early dropouts (ED), (3) declining attenders (DA), and (4) sustained attenders (SA). In the final model testing multiple covariates simultaneously, mothers who were EDs and DAs were more likely to be Latina than SAs, and EDs reported more interparental conflict than SAs. Mother trajectory groups did not differ on parenting skills, child problem behavior, or mother-child contact in the final or preliminary models. In the final model for fathers, EDs rated their children higher on externalizing than DAs, had less contact with their children than DAs and NAs, and reported less distress than SAs. Father trajectory groups did not differ on fathers’ age, ethnicity, income, perceived parenting skills, or interparental conflict in the final or preliminary models. Results highlight qualitatively distinct latent classes of mothers and fathers who disengage from a parenting intervention at various points. We discuss implications for intervention engagement strategies and translational science.     

Bayesian latent class models with conditionally dependent diagnostic tests: a case study

In the assessment of the accuracy of diagnostic tests for infectious diseases, the true disease status of the subjects is often unknown due to the lack of a gold standard test. Latent class models with two latent classes, representing diseased and non-diseased subjects, are often used to analyze this type of data. In its basic format, latent class analysis requires the observed outcomes to be statistically independent conditional on the disease status. In most diagnostic settings, this assumption is highly questionable. During the last decade, several methods have been proposed to estimate latent class models with conditional dependence between the test results. A class of flexible fixed and random effects models were described by Dendukuri and Joseph in a Bayesian framework. We illustrate these models using the analysis of a diagnostic study of three field tests and an imperfect reference test for the diagnosis of visceral leishmaniasis. We show that, as observed earlier by Albert and Dodd, different dependence models may result in similar fits to the data while resulting in different inferences. Given this problem, selection of appropriate latent class models should be based on substantive subject matter knowledge. If several clinically plausible models are supported by the data, a sensitivity analysis should be performed by describing the results obtained from different models and using different priors. Copyright (c) 2008 John Wiley & Sons, Ltd.     

祖父母哮喘史与被动吸烟对儿童哮喘的影响效应

目的 探讨祖父母哮喘史与被动吸烟的交互作用对儿童哮喘及哮喘样症状的影响。方法 采用整群随机抽样法在辽宁省7个城市各抽取1所小学和2所幼儿园, 对所选学校所有在此居住时间≥2年且与祖父母居住时间≥3年的儿童的祖父母哮喘史、居住环境、被动吸烟等情况及呼吸系统疾病与症状的信息进行收集。采用Glimmix回归模型分析祖父母哮喘史与被动吸烟的交互作用对儿童哮喘及哮喘样症状(持续咳嗽、持续咳痰、喘鸣现患、过敏性鼻炎)的影响。结果 在调查的3 576名儿童中, 年龄为(8.11±2.87)岁, 祖父母有哮喘史的儿童有224名, 祖父母哮喘史率为6.26%;被动吸烟的儿童有970名, 被动吸烟率为27.13%。祖父母有哮喘史的儿童咳嗽、喘鸣现患和哮喘阳性率分别为16.07%(336/224), 14.29%(32/224)和13.39%(30/224), 祖父母无哮喘史的儿童为10.32%(346/3 352), 5.91%(198/3 352)和5.61%(188/3 352), 祖父母有哮喘史儿童发生咳嗽(OR=1.66, 95%CI:1.15～2.42)、喘鸣现患(OR=2.65, 95%CI:1.78～3.96)和哮喘(OR=2.60, 95%CI:1.72～3.93)的危险性较高。祖父母吸烟的儿童咳嗽和咳痰的阳性率分别为14.02%(136/970)和6.19%(60/970), 祖父母吸烟的儿童发生咳嗽(OR=1.56, 95%CI:1.25～1.96)和咳痰(OR=1.57, 95%CI:1.13～2.18)的危险性较高。Glimmix回归模型分析显示, 祖父母有哮喘史与被动吸烟对儿童哮喘及哮喘样症状存在交互影响:在儿童被动吸烟暴露下, 祖父母有哮喘史的儿童发生咳痰的危险性高于祖父母无哮喘史儿童(P<0.05)。结论 祖父母有哮喘史使被动吸烟的儿童发生哮喘及哮喘样症状的危险性升高。     

Passive smoking and respiratory conditions in primary school children.

The effect of passive smoking on respiratory symptoms of children aged 5 to 11 years was investigated in over 4000 English children and nearly 800 Scottish children participating in the National Study of Health and Growth in 1982. After adjusting for associations of respiratory symptoms with age, sex, and a number of potentially confounding variables, significant associations were found of wheeze, both occasional and persistent, day or night cough, and bronchitis attacks with number of cigarettes smoked by parents at home for English children and for occasional wheeze in Scottish children. Asthma attacks and cough first thing in the morning showed positive but not statistically significant associations in English children. The presence of at least one condition was statistically significant in both English and Scottish children. The largest relative risk for exposure to 20 cigarettes a day compared to no exposure was 1.60 for persistent wheeze in English children (95% confidence interval 1.17-2.18).     

Making sense of symptom checklists: a latent class approach to the first 9 years of the British Household Panel Survey

BACKGROUND: In health inequalities research there is a growing impetus to examine the development of inequalities in health over time. However, many of the sources of longitudinal data in Britain are not designed specifically for health research. Typically, health status is assessed by self-reported problems and the use of symptom checklists. METHODS: The British Household Panel Survey (BHPS) is an annual survey of approximately 5500 private households containing 9000 men and women, which began in 1991. Each year, the BHPS contains a checklist of 13 health problems and symptoms. The findings presented here are based on adult participants aged 16 years and over in 1991. Using eight waves of data from the BHPS, we use latent class analysis (LCA) to model latent health status from a set of observed binary variables. Individuals are assigned to a latent health class on the basis of LCA estimated probabilities of class membership given their response patterns and the estimated unconditional class frequencies. The predictive value of latent health class membership is assessed for self-reported health status and functioning, health and welfare service use, and mortality 1 year later. RESULTS: The LCA supported a suitable four-class model of health status representing good health, psychosomatic health problems, physical health problems and comorbid health problems. Members of the good latent health class were predicted to have better self-reported health and functioning, less health and welfare service use, and lower risk of mortality 1 year later than members of the three problem health classes. Those with comorbid health problems were predicted to have particularly poor outcomes. CONCLUSIONS: A latent class approach to modelling self-reported health problems and symptoms has allowed for both quantitative and qualitative dimensions of health status to be captured. This may motivate better informed models of health by users of general population surveys.     

Maternal factors, obstetric history and smoking stage of change.

BACKGROUND: Maternal smoking is known to be associated with low educational status, low social class and younger age groups. The aim of this study was to determine if maternal smoking and stage of change relating to smoking is associated with other maternal variables such as intention to breastfeed and attend antenatal classes, having a planned pregnancy, previous obstetric history and child health problems. METHOD: A cross-sectional survey was carried out of all women who attended antenatal clinics at the Leicester Royal Infirmary NHS Trust over a two-week period. The data comprised 254 completed questionnaires. Results Intention to breastfeed was more common among non-smokers as shown by smoking status (p < 0.001) and smoking stage of change (p < 0.05). Having a planned pregnancy was more common among non-smokers as determined by smoking status (p < 0.001) and stage of change (p < 0.05). Intention to attend antenatal classes showed no significant relationship with smoking status but the majority of those planning to attend antenatal classes were in the action-maintenance stage (p< 0.05). Previous obstetric complications were not associated with either smoking status or stage of change. Smokers were more likely to have at least one child with asthma (p < 0.05) or respiratory infections (p < 0.001). Having at least one child with asthma or respiratory infections was more common among precontemplators (p < 0.05). CONCLUSION: Smoking stage of change should be assessed in antenatal care so that appropriate information can be offered to pregnant smokers. The development and evaluation of stage-specific smoking cessation materials should offer considerable benefits to maternal and infant health.     

A dynamic trajectory class model for intensive longitudinal categorical outcome

This paper presents a novel dynamic latent class model for a longitudinal response that is frequently measured as in our prospective study of older adults with monthly data on activities of daily living for more than 10 years. The proposed method is especially useful when the longitudinal response is measured much more frequently than other relevant covariates. The trajectory classes are latent classes that represent distinct temporal patterns of the longitudinal response wherein an individual may remain in a trajectory class or switch to another as the class membership predictors are updated periodically over time. The identification of a common set of trajectory classes allows changes among the temporal patterns to be distinguished from local fluctuations in the response. Within a trajectory class, the longitudinal response is modeled by a class‐specific generalized linear mixed model. An informative event such as death is jointly modeled by class‐specific probability of the event through shared random effects with that for the longitudinal response. We do not impose the conditional independence assumption given the classes. We illustrate the method by analyzing the change over time in activities of daily living trajectory class among 754 older adults with 70,500 person‐months of follow‐up in the Precipitating Events Project. We also investigate the impact of jointly modeling the class‐specific probability of the event on the parameter estimates in a simulation study. The primary contribution of our paper is the periodic updating of trajectory classes for a longitudinal categorical response without assuming conditional independence. Copyright © 2014 John Wiley & Sons, Ltd.