Positive and negative affect data are often collected over time in psychiatric care settings, yet no generally accepted means are available to relate these data to useful diagnoses or treatments. Latent class analysis attempts data reduction by classifying subjects into one of K unobserved classes based on observed data. Latent class models have recently been extended to accommodate longitudinally observed data. We extend these approaches in a Bayesian framework to accommodate trajectories of both continuous and discrete data. We consider whether latent class models might be used to distinguish patients on the basis of trajectories of observed affect scores, reported events, and presence or absence of clinical depression.
Cardiovascular disease; Depression; DIC; General growth mixture modeling; Gibbs sampling; Label switching; Model choice
To understand one developmental process, it is often helpful to investigate its relations with other developmental processes. Statistical methods that model development in multiple processes simultaneously over time include latent growth curve models with time-varying covariates, multivariate latent growth curve models, and dual trajectory models. These models are designed for growth represented by continuous, unidimensional trajectories. The purpose of this article is to present a flexible approach to modeling relations in development among two or more discrete, multidimensional latent variables based on the general framework of loglinear modeling with latent variables called associative latent transition analysis (ALTA). Focus is given to the substantive interpretation of different associative latent transition models, and exactly what hypotheses are expressed in each model. An empirical demonstration of ALTA is presented to examine the association between the development of alcohol use and sexual risk behavior during adolescence.
latent class analysis; latent transition analysis; associative latent transition analysis; loglinear modeling; sexual behavior; alcohol use; adolescents
Although psychiatric diagnostic systems have conceptualized mania as a discrete phenomenon, appropriate latent structure investigations testing this conceptualization are lacking. In contrast to these diagnostic systems, several influential theories of mania have suggested a continuous conceptualization. The present study examined whether mania has a continuous or discrete latent structure using a comprehensive approach including taxometric, information-theoretic latent distribution modeling (ITLDM), and predictive validity methodologies in the Epidemiologic Catchment Area (ECA) study.
Eight dichotomous manic symptom items were submitted to a variety of latent structural analyses; including factor analyses, taxometric procedures, and ITLDM; in 10,105 ECA community participants. Additionally, a variety of continuous and discrete models of mania were compared in terms of their relative abilities to predict outcomes (i.e., health service utilization, internalizing and externalizing disorders, and suicidal behavior).
Taxometric and ITLDM analyses consistently supported a continuous conceptualization of mania. In ITLDM analyses, a continuous model of mania demonstrated 6:52:1 odds over the best fitting latent class model of mania. Factor analyses suggested that the continuous structure of mania was best represented by a single latent factor. Predictive validity analyses demonstrated a consistent superior ability of continuous models of mania relative to discrete models.
The present study provided three independent lines of support for a continuous conceptualization of mania. The implications of a continuous model of mania are discussed.
To assess variation in the quality of care in general practice and identify factors associated with high quality care.
Stratified random sample of 60 general practices in six areas of England.
Quality of management of chronic disease (angina, asthma in adults, and type 2 diabetes) and preventive care (rates of uptake for immunisation and cervical smear), access to care, continuity of care, and interpersonal care (general practice assessment survey). Multiple logistic regression with multilevel modelling was used to relate each of the outcome variables to practice size, routine booking interval for consultations, socioeconomic deprivation, and team climate.
Quality of clinical care varied substantially, and access to care, continuity of care, and interpersonal care varied moderately. Scores for asthma, diabetes, and angina were 67%, 21%, and 17% higher in practices with 10 minute booking intervals for consultations compared with practices with five minute booking intervals. Diabetes care was better in larger practices and in practices where staff reported better team climate. Access to care was better in small practices. Preventive care was worse in practices located in socioeconomically deprived areas. Scores for satisfaction, continuity of care, and access to care were higher in practices where staff reported better team climate.
Longer consultation times are essential for providing high quality clinical care. Good teamworking is a key part of providing high quality care across a range of areas and may need specific support if quality of care is to be improved. Additional support is needed to provide preventive care to deprived populations. No single type of practice has a monopoly on high quality care: different types of practice may have different strengths.
What is already known on this topicQuality of care varies in virtually all aspects of medicine that have been studiedMost studies look at quality of care from a single perspective or for a single conditionWhat this study addsQuality of care varies for both clinical care and assessments by patients of access and interpersonal carePractices with longer booking intervals provide better management of chronic disease; preventive care is less good in practices in deprived areasNo single type of practice has a monopoly on high quality care—small practices provide better access but poorer diabetes careGood team climate reported by staff is associated with a range of aspects of high quality care
Socio-economic variables are often measured on a discrete scale or rounded to protect confidentiality. Nevertheless, when exploring the effect of a relevant covariate on the outcome distribution of a discrete response variable, virtually all common quantile regression methods require the distribution of the covariate to be continuous. This paper departs from this basic requirement by presenting an algorithm for nonparametric estimation of conditional quantiles when both the response variable and the covariate are discrete. Moreover, we allow the variables of interest to be pairwise correlated. For computational efficiency, we aggregate the data into smaller subsets by a binning operation, and make inference on the resulting prebinned data. Specifically, we propose two kernel-based binned conditional quantile estimators, one for untransformed discrete response data and one for rank-transformed response data. We establish asymptotic properties of both estimators. A practical procedure for jointly selecting band- and binwidth parameters is also presented. Simulation results show excellent estimation accuracy in terms of bias, mean squared error, and confidence interval coverage. Typically prebinning the data leads to considerable computational savings when large datasets are under study, as compared to direct (un)conditional quantile kernel estimation of multivariate data. With this in mind, we illustrate the proposed methodology with an application to a large dataset concerning US hospital patients with congestive heart failure.
Binning; bootstrap; confidence interval; jittering; nonparametric
Contraceptive prevalence is relatively high in Brazil (55% among women of reproductive age). However, reversible methods account for less than half of the method mix and widespread differences persist across regions and social groups. This draws attention to the need for monitoring family planning service-related outcomes that might be linked with quality of care. The present study examines the factors associated with method discontinuation, failure and switching among current contraceptive users, with a focus on sub-national assessment.
Data for the analysis are drawn from the Brazil Demographic and Health Survey, notably the calendar module of reproductive events. Multilevel discrete-time competing risks hazard models are used to estimate the random- and fixed-effects on the probability of a woman making a specific transition after a given duration of contraceptive use.
Contraceptive continuation was found to be highest for the contraceptive pill, the most popular reversible method. Probabilities of abandonment while in need of family planning and of switching to another method were highest for injections. Failure, abandonment and switching were each higher among users in the Northeast region compared to the more prosperous Southeast and South.
Findings point to seemingly important disparities in the availability and quality of family planning and reproductive health care services across regions of the country. Expanding access to a range of contraceptive methods, improving knowledge among health agents of contraceptive technologies and increasing medical supervision of contraceptive practice may be considered key to expanding quality reproductive health care services for all.
Health-Related Quality of Life (HRQoL) measures are becoming increasingly used in clinical trials as primary outcome measures. Investigators are now asking statisticians for advice on how to analyse studies that have used HRQoL outcomes.
HRQoL outcomes, like the SF-36, are usually measured on an ordinal scale. However, most investigators assume that there exists an underlying continuous latent variable that measures HRQoL, and that the actual measured outcomes (the ordered categories), reflect contiguous intervals along this continuum.
The ordinal scaling of HRQoL measures means they tend to generate data that have discrete, bounded and skewed distributions. Thus, standard methods of analysis such as the t-test and linear regression that assume Normality and constant variance may not be appropriate. For this reason, conventional statistical advice would suggest that non-parametric methods be used to analyse HRQoL data. The bootstrap is one such computer intensive non-parametric method for analysing data.
We used the bootstrap for hypothesis testing and the estimation of standard errors and confidence intervals for parameters, in four datasets (which illustrate the different aspects of study design). We then compared and contrasted the bootstrap with standard methods of analysing HRQoL outcomes. The standard methods included t-tests, linear regression, summary measures and General Linear Models.
Overall, in the datasets we studied, using the SF-36 outcome, bootstrap methods produce results similar to conventional statistical methods. This is likely because the t-test and linear regression are robust to the violations of assumptions that HRQoL data are likely to cause (i.e. non-Normality). While particular to our datasets, these findings are likely to generalise to other HRQoL outcomes, which have discrete, bounded and skewed distributions. Future research with other HRQoL outcome measures, interventions and populations, is required to confirm this conclusion.
Health Related Quality of Life; SF-36; Bootstrap Simulation; Statistical Analysis.
The Department of Veterans Affairs (VA) provides rehabilitation for veterans with moderate to severe war injuries through four regional Polytrauma Rehabilitation Centers (PRCs). To standardize and improve care provided to these veterans′ family members, health services researchers partnered with program leaders and rehabilitation specialists to implement a family care quality improvement collaborative.
To describe practice changes associated with the Family Care Collaborative′s intervention.
Cross-site, mixed-method evaluation.
Rehabilitation interdisciplinary team members (n = 226) working at the four participating sites.
The collaborative developed and implemented in a 6-month pilot a web-based tool to standardize and promote family-centered care.
Provider survey of family care, satisfaction with family care, and perceived competence in working with families; specific practice changes at each site; provider and facilitator perceptions of the collaborative work; and a validated measure to predict likelihood of success of the selected intervention.
Family-centered practices and satisfaction improved at sites with lower baseline scores (P < 0.05) and was equivalent across sites after the pilot. Providers initiated specific family-centered practices that often began at one site and spread to the others through the collaborative. Sites standardized family education and collaboration. Providers believed that the collaborative produced a “culture change” from patient-centered to family-centered care and viewed program leadership and health services researchers' involvement as crucial for success. Scores on the measure to predict successful implementation of the intervention beyond the pilot were promising.
Collaboratives that bring together clinicians, program leaders, and researchers may be useful for fostering complex change involving interdisciplinary teams.
rehabilitation; family-centered care; veterans; collaborative; attitude of health personnel; patient care team; implementation evaluation; polytrauma; practice change
An increasing number of algorithms for biochemical network inference from experimental data require discrete data as input. For example, dynamic Bayesian network methods and methods that use the framework of finite dynamical systems, such as Boolean networks, all take discrete input. Experimental data, however, are typically continuous and represented by computer floating point numbers. The translation from continuous to discrete data is crucial in preserving the variable dependencies and thus has a significant impact on the performance of the network inference algorithms. We compare the performance of two such algorithms that use discrete data using several different discretization algorithms. One of the inference methods uses a dynamic Bayesian network framework, the other—a time-and state-discrete dynamical system framework. The discretization algorithms are quantile, interval discretization, and a new algorithm introduced in this article, SSD. SSD is especially designed for short time series data and is capable of determining the optimal number of discretization states. The experiments show that both inference methods perform better with SSD than with the other methods. In addition, SSD is demonstrated to preserve the dynamic features of the time series, as well as to be robust to noise in the experimental data. A C++ implementation of SSD is available from the authors at http://polymath.vbi.vt.edu/discretization.
gene networks; genetic algorithms; linear algebra; reverse engineering; time discrete dynamical systems
Analyses of serially-sampled data often begin with the assumption that the observations represent discrete samples from a latent continuous-time stochastic process. The continuous-time Markov chain (CTMC) is one such generative model whose popularity extends to a variety of disciplines ranging from computational finance to human genetics and genomics. A common theme among these diverse applications is the need to simulate sample paths of a CTMC conditional on realized data that is discretely observed. Here we present a general solution to this sampling problem when the CTMC is defined on a discrete and finite state space. Specifically, we consider the generation of sample paths, including intermediate states and times of transition, from a CTMC whose beginning and ending states are known across a time interval of length T. We first unify the literature through a discussion of the three predominant approaches: (1) modified rejection sampling, (2) direct sampling, and (3) uniformization. We then give analytical results for the complexity and efficiency of each method in terms of the instantaneous transition rate matrix Q of the CTMC, its beginning and ending states, and the length of sampling time T. In doing so, we show that no method dominates the others across all model specifications, and we give explicit proof of which method prevails for any given Q, T, and endpoints. Finally, we introduce and compare three applications of CTMCs to demonstrate the pitfalls of choosing an inefficient sampler.
Continuous-time Markov chains; simulation; molecular evolution
Reporting of quality indicators (QIs) in Veterans Health Administration Medical Centers is complicated by estimation error due to small numbers of eligible patients per facility. We applied multilevel modeling and empirical Bayes (EB) estimation in addressing this issue in performance reporting of stroke care quality in the Medical Centers.
Methods and Results
We studied a retrospective cohort of 3812 veterans admitted to 106 Medical Centers with ischemic stroke during fiscal year 2007. The median number of study patients per facility was 34 (range: 12-105). Inpatient stroke care quality was measured with thirteen evidence-based QIs. Eligible patients could either pass or fail each indicator. Multilevel modeling of a patient’s pass/fail on individual QIs was used to produce facility-level EB estimated QI pass rates and confidence intervals. The EB estimation reduced inter-facility variation in QI rates. Small facilities and those with exceptionally high or low rates were most affected. We recommended 8 of the 13 QIs for performance reporting: dysphagia screening, NIH Stroke Scale documentation, early ambulation, fall risk assessment, pressure ulcer risk assessment, Functional Independence Measure documentation, lipid management, and deep vein thrombosis prophylaxis. These QIs displayed sufficient variation across facilities, had room for improvement, and identified sites with performance that was significantly above or below the population average. The remaining 5 QIs were not recommended because of too few eligible patients or high pass rates with little variation.
Considerations of statistical uncertainty should inform the choice of QIs and their application to performance reporting.
acute stroke; performance measurement; quality indicators; stroke management; statistics
Continuing challenges to timely adoption of evidence-based clinical practices in healthcare have generated intense interest in the development and application of new implementation methods and frameworks. These challenges led the United States (U.S.) Department of Veterans Affairs (VA) to create the Quality Enhancement Research Initiative (QUERI) in the late 1990s. QUERI's purpose was to harness VA's health services research expertise and resources in an ongoing system-wide effort to improve the performance of the VA healthcare system and, thus, quality of care for veterans. QUERI in turn created a systematic means of involving VA researchers both in enhancing VA healthcare quality, by implementing evidence-based practices, and in contributing to the continuing development of implementation science.
The efforts of VA researchers to improve healthcare delivery practices through QUERI and related initiatives are documented in a growing body of literature. The scientific frameworks and methodological approaches developed and employed by QUERI are less well described. A QUERI Series of articles in Implementation Science will illustrate many of these QUERI tools. This Overview article introduces both QUERI and the Series.
The Overview briefly explains the purpose and context of the QUERI Program. It then describes the following: the key operational structure of QUERI Centers, guiding frameworks designed to enhance implementation and related research, QUERI's progress and promise to date, and the Series' general content. QUERI's frameworks include a core set of steps for diagnosing and closing quality gaps and, simultaneously, advancing implementation science. Throughout the paper, the envisioned involvement and activities of VA researchers within QUERI Centers also are highlighted. The Series is then described, illustrating the use of QUERI frameworks and other tools designed to respond to implementation challenges.
QUERI's simultaneous pursuit of improvement and research goals within a large healthcare system may be unique. However, descriptions of this still-evolving effort, including its conceptual frameworks, methodological approaches, and enabling processes, should have applicability to implementation researchers in a range of health care settings. Thus, the Series is offered as a resource for other implementation research programs and researchers pursuing common goals in improving care and developing the field of implementation science.
Finite mixture models have come to play a very prominent role in modelling data. The finite mixture model is predicated on the assumption that distinct latent groups exist in the population. The finite mixture model therefore is based on a categorical latent variable that distinguishes the different groups. Often in practice distinct sub-populations do not actually exist. For example, disease severity (e.g. depression) may vary continuously and therefore, a distinction of diseased and not-diseased may not be based on the existence of distinct sub-populations. Thus, what is needed is a generalization of the finite mixture’s discrete latent predictor to a continuous latent predictor. We cast the finite mixture model as a regression model with a latent Bernoulli predictor. A latent regression model is proposed by replacing the discrete Bernoulli predictor by a continuous latent predictor with a beta distribution. Motivation for the latent regression model arises from applications where distinct latent classes do not exist, but instead individuals vary according to a continuous latent variable. The shapes of the beta density are very flexible and can approximate the discrete Bernoulli distribution. Examples and a simulation are provided to illustrate the latent regression model. In particular, the latent regression model is used to model placebo effect among drug treated subjects in a depression study.
Beta distribution; EM algorithm; finite and infinite mixtures; quasi-Newton algorithms; placebo effect; skew normal distribution
Inflammatory bowel disease imposes psychosocial stress on the patient. Patients′ adaptive capacities may predict quality of life. We examined two adaptive capacity measures and their association with quality of life.
Cross-sectional mail survey of patients with inflammatory bowel disease. The Patient Activation Measure (PAM) assesses knowledge, skill, and confidence in self-health management. The Perceived Expectancies Index (PEI) measures perceived competence and dispositional optimism.
Four hundred and seventy-seven veterans at VA-Tennessee Valley Healthcare System.
Main Outcome Measure
Primary outcome was health-related quality of life (measured by the Short Inflammatory Bowel Disease Questionnaire). Bivariate analysis assessed unadjusted correlations. Sequential multivariate linear regression tested theoretical model relationships by calculating the variation in each dependent variable accounted for by independent variables (-squared statistic).
Two hundred and sixty surveys were returned with usable data (54.5%). Median age was 63 years (range 19–91); 90.8% were men and 86.9% self-identified as white. Fifty percent reported having ulcerative colitis, 36.5% Crohn’s disease, and 12.3% uncertain type. Unadjusted bivariate analysis revealed positive correlations between the PAM and PEI and the Short Inflammatory Bowel Disease Questionnaire (correlation coefficient = 0.35 and 0.60, respectively; < 0.0001). Multivariate model including the PAM accounted for 26% of the variation in Short Inflammatory Bowel Disease Questionnaire scores, while the model including the PEI accounted for 50% ( < 0.0001).
There are positive, highly significant correlations between adaptive capacities and health-related quality of life in patients with inflammatory bowel disease.
Electronic supplementary material
The online version of this article (doi:10.1007/s11606-009-1002-0) contains supplementary material, which is available to authorized users.
psychosocial stress; inflammatory bowel disease; Veterans; health outcomes
In data commonly used for health services research, a number of relevant variables are unobservable. These include population lifestyle and socio-economic status, physician practice behaviors, population tendency to use health care resources, and disease prevalence. These variables may be considered latent constructs of many observed variables. Using health care data from South Carolina, we show an application of spatial structural equation modeling to identify how these latent constructs are associated with access to primary health care, as measured by hospitalizations for ambulatory care sensitive conditions. We applied the confirmatory factor analysis approach, using the Bayesian paradigm, to identify the spatial distribution of these latent factors. We then applied cluster detection tools to identify counties that have a higher probability of hospitalization for each of the twelve adult ambulatory care sensitive conditions, using a multivariate approach that incorporated the correlation structure among the ambulatory care sensitive conditions into the model.
For the South Carolina population ages 18 and over, we found that counties with high rates of emergency department visits also had less access to primary health care. We also observed that in those counties there are no community health centers.
Locating such clusters will be useful to health services researchers and health policy makers; doing so enables targeted policy interventions to efficiently improve access to primary care.
Mental health, specifically mood/anxiety disorders, may be associated with value for health care attributes, but the association remains unclear. Examining the relation between mental health and attributes in a context where quality of care is low and exposure to suboptimal health conditions is increased, such as in Sub Saharan Africa (SSA), may elucidate the association.
We assessed whether preference weights for obstetric care attributes varied by mental health among 1006 women from Jimma Zone, Ethiopia, using estimates obtained through a discrete choice experiment (DCE), a method used to elicit preferences. Facilities were described by several attributes including provider attitude and performance and drug/equipment availability. Mental health measures included depressive symptoms and posttraumatic stress disorder (PTSD). We used Bayesian models to estimate preference weights for attributes and linear models to investigate whether these weights were associated with mental health. We found that women with high depressive symptoms valued a positive provider attitude [β = −0.43 (95% CI: −0.66, −0.21)] and drug/equipment availability [β = −0.43 (95% CI: −0.78, −0.07)] less compared to women without high depressive symptoms. Similar results were obtained for PTSD. Upon adjusting for both conditions, value for drug/equipment availability was lower only among women with both conditions [β = −0.89 (95% CI −1.4, −0.42)].
We found that women with psychopathology had lower preference weights for positive provider attitude and drug/equipment availability. Further work investigating why value for obstetric care attributes might vary by psychopathology in SSA is needed.
Item Response Theory (IRT) is increasingly applied in health research to combine information from multiple-item responses. IRT posits that a person's susceptibility to a symptom is driven by the interaction of the characteristics of the symptom and person. This article describes the statistical background of incorporating IRT into a multilevel framework and extends this approach to longitudinal health outcomes, where the self-report method is used to construct a multi-item scale.
A secondary analysis of data from 2 descriptive longitudinal studies is performed. The data include 21 symptoms reported across time by 350 women with breast cancer. A 3-level hierarchical linear model (HLM) was used for the analysis. Level 1 models the item responses, consisting of symptom presence or absence. Level 2 models the trajectory of each individual, representing change over time of the IRT-created latent variable symptom experience. Level 3 explains that trajectory using person-specific characteristics such as age and location of care. The purpose of the analysis is to examine if older and younger women with breast cancer differ in their symptom experience trajectory after controlling for location of care.
Fatigue and pain were the most prevalent symptoms. The symptom experience of women with breast cancer was found to improve over time. Neither age nor location of care was significantly associated with the symptom experience trajectory.
Embedding IRT into an HLM framework produces several benefits. The example provided demonstrates benefits through the creation of a latent symptom experience variable that can be used either as an outcome or as a covariate in another model, examining the latent symptom experience trajectory and its relationship with covariates at the individual level, and managing symptom nonresponse.
cancer symptoms; hierarchical linear model; Item Response Theory; women with breast cancer
Suicide is a global public health problem. Recently in the U.S., much attention has been given to preventing suicide and other premature mortality in veterans returning from Iraq and Afghanistan. A strong predictor of suicide is a past suicide attempt, and suicide attempters have multiple physical and mental comorbidities that put them at risk for additional causes of death. We examined mortality among U.S. military veterans after hospitalization for attempted suicide.
A retrospective cohort study was conducted with all military veterans receiving inpatient treatment during 1993-1998 at United States Veterans Affairs (VA) medical facilities following a suicide attempt. Deaths occurring during 1993-2002, the most recent available year at the time, were identified through VA Beneficiary and Records Locator System data and National Death Index data. Mortality data for the general U.S. adult population were also obtained from the National Center for Health Statistics. Comparisons within the veteran cohort, between genders, and against the U.S. population were conducted with descriptive statistics and standardized mortality ratios. The actuarial method was used estimate the proportion of veterans in the cohort we expect would have survived through 2002 had they experienced the same rate of death that occurred over the study period in the U.S. population having the age and sex characteristics.
During 1993-1998, 10,163 veterans were treated and discharged at a VA medical center after a suicide attempt (mean age = 44 years; 91% male). There was a high prevalence of diagnosed alcohol disorder or abuse (31.8%), drug dependence or abuse (21.8%), psychoses (21.2%), depression (18.5%), and hypertension (14.2%). A total of 1,836 (18.1%) veterans died during follow up (2,941.4/100,000 person years). The cumulative survival probability after 10 years was 78.0% (95% CI = 72.9, 83.1). Hence the 10-year cumulative mortality risk was 22.0%, which was 3.0 times greater than expected. The leading causes overall were heart disease (20.2%), suicide (13.1%), and unintentional injury (12.7%). Whereas suicide was the ninth leading cause of death in the U.S. population overall (1.8%) during the study period, suicide was the leading and second leading cause among women (25.0%) and men (12.7%) in the cohort, respectively.
Veterans who have attempted suicide face elevated risks of all-cause mortality with suicide being prominent. This represents an important population for prevention activities.
Evaluation of outcomes can help improve the quality of provision of services within a healthcare setting. There is limited report on patient satisfaction in private-sector in India although they provide three-quarters of healthcare services.
The study was designed to report the level of satisfaction among inpatients of a private tertiary care hospital in India.
Materials and Methods:
A total of 102 participants were recruited and their socio-demographic, health-seeking behavior, and satisfaction rating on various aspects of healthcare were elicited. A five item Likert scale was used to obtain the satisfaction rating. Data analysis was done with the help of Stata version-9. Proportions for the discrete variables and means with Standard Deviation for the continuous variables were obtained.
All the participants were urban and from upper-middle or upper socio-economic strata. The participants reported a high level of overall satisfaction (93%) as well as high satisfaction with physicians (95%), the doctor's interpersonal skills (99%), nursing-care (93%), general services (94%), and pharmacy (88.1%).
There was a high level of satisfaction reported by the participants at this tertiary level hospital. This might reflect the actual good quality services being provided by the provider or the nonannoying response, which cannot be ruled out.
Client satisfaction; patient satisfaction; private hospital; tertiary care
Reverse engineering cellular networks is currently one of the most challenging problems in systems biology. Dynamic Bayesian networks (DBNs) seem to be particularly suitable for inferring relationships between cellular variables from the analysis of time series measurements of mRNA or protein concentrations. As evaluating inference results on a real dataset is controversial, the use of simulated data has been proposed. However, DBN approaches that use continuous variables, thus avoiding the information loss associated with discretization, have not yet been extensively assessed, and most of the proposed approaches have dealt with linear Gaussian models.
We propose a generalization of dynamic Gaussian networks to accommodate nonlinear dependencies between variables. As a benchmark dataset to test the new approach, we used data from a mathematical model of cell cycle control in budding yeast that realistically reproduces the complexity of a cellular system. We evaluated the ability of the networks to describe the dynamics of cellular systems and their precision in reconstructing the true underlying causal relationships between variables. We also tested the robustness of the results by analyzing the effect of noise on the data, and the impact of a different sampling time.
The results confirmed that DBNs with Gaussian models can be effectively exploited for a first level analysis of data from complex cellular systems. The inferred models are parsimonious and have a satisfying goodness of fit. Furthermore, the networks not only offer a phenomenological description of the dynamics of cellular systems, but are also able to suggest hypotheses concerning the causal interactions between variables. The proposed nonlinear generalization of Gaussian models yielded models characterized by a slightly lower goodness of fit than the linear model, but a better ability to recover the true underlying connections between variables.
Gaussian factor models have proven widely useful for parsimoniously characterizing dependence in multivariate data. There is a rich literature on their extension to mixed categorical and continuous variables, using latent Gaussian variables or through generalized latent trait models acommodating measurements in the exponential family. However, when generalizing to non-Gaussian measured variables the latent variables typically influence both the dependence structure and the form of the marginal distributions, complicating interpretation and introducing artifacts. To address this problem we propose a novel class of Bayesian Gaussian copula factor models which decouple the latent factors from the marginal distributions. A semiparametric specification for the marginals based on the extended rank likelihood yields straightforward implementation and substantial computational gains. We provide new theoretical and empirical justifications for using this likelihood in Bayesian inference. We propose new default priors for the factor loadings and develop efficient parameter-expanded Gibbs sampling for posterior computation. The methods are evaluated through simulations and applied to a dataset in political science. The models in this paper are implemented in the R package bfa.1
Factor analysis; Latent variables; Semiparametric; Extended rank likelihood; Parameter expansion; High dimensional
Siblings of children with autism (sibs-A) are at increased genetic risk for autism spectrum disorders (ASD) and milder impairments. To elucidate diversity and contour of early developmental trajectories exhibited by sibs-A, regardless of diagnostic classification, latent class modeling was used.
Sibs-A (n=204) were assessed with the Mullen Scales of Early Learning from age 6–36 months. Mullen T scores served as dependent variables. Outcome classifications at age 36 months included: ASD (n=52); non-ASD social/communication delay (broader autism phenotype; BAP) (n=31); and unaffected (n=121). Child-specific patterns of performance were studied using latent class growth analysis. Latent class membership was then related to diagnostic outcome through estimation of within-class proportions of children assigned to each diagnostic classification.
A 4-class model was favored. Class 1 represented accelerated development and consisted of 25.7% of the sample, primarily unaffected children. Class 2 (40.0% of the sample), was characterized by normative development with above-average nonverbal cognitive outcome. Class 3 (22.3% of the sample) was characterized by receptive language, and gross and fine motor delay. Class 4 (12.0% of the sample), was characterized by widespread delayed skill acquisition, reflected by declining trajectories. Children with an outcome diagnosis of ASD were spread across Classes 2, 3, and 4.
Results support a category of ASD that involves slowing in early non-social development. Receptive language and motor development is vulnerable to early delay in sibs-A with and without ASD outcomes. Non-ASD sibs-A are largely distributed across classes depicting average or accelerated development. Developmental trajectories of motor, language, and cognition appear independent of communication and social delays in non-ASD sibs-A.
autism; trajectories; broader autism phenotype
The Institute of Medicine proposed recently that, while current pay for performance measures should target multiple dimensions of care, including measures of technical quality, they should transition toward longitudinal and health-outcome measures across systems of care. This article describes the development of the Diabetes Epidemiology Cohorts (DEpiC), which facilitates evaluation of intermediate “quality of care” outcomes and surveillance of adverse outcomes for veterans with diabetes served by the Veterans Health Administration (VHA) over multiple years.
The Diabetes Epidemiology Cohorts is a longitudinal research database containing records for all diabetes patients in the VHA since 1998. It is constructed using data from a variety of national computerized data files in the VHA (including medical encounters, prescriptions, laboratory tests, and mortality files), Medicare claims data for VHA patients, and large patient surveys conducted by the VHA. Rigorous methodology is applied in linking and processing data into longitudinal patient records to assure data quality.
Validity is demonstrated in the construction of the DEpiC. Adjusted comparisons of disease prevalence with general population estimates are made. Further analyses of intermediate outcomes of care demonstrate the utility of the database. In the first example, using growth curve models, we demonstrated that hemoglobin A1c trends exhibit marked seasonality and that serial cross-sectional outcomes overestimate the improvement in population A1c levels compared to longitudinal cohort evaluation. In the second example, the use of individual level data enabled the mapping of regional performance in amputation prevention into four quadrants using calculated observed to expected major and minor amputation rates. Simultaneous evaluation of outliers in all categories of amputation may improve the oversight of foot care surveillance programs.
The use of linked, patient level longitudinal data resolves confounding case mix issues inherent in the use of serial cross-sectional data. Policy makers should be aware of the limitations of cross-sectional data and should make use of longitudinal patient databases to evaluate clinical outcomes.
A1c; amputations; databases; diabetes; registry
To test the hypothesis that higher levels of participation and functioning in cross-functional psychiatric treatment teams will be related to improved patient outcomes.
Data Sources/Study Setting
Primary data were collected during the period 1992–1999. The study was conducted in 40 teams within units treating seriously mentally ill patients in 16 Veterans Affairs hospitals across the U.S.
A longitudinal, multilevel analysis assessed the relationship between individual- and team-level variables and patients' ability to perform activities of daily living (ADL) over time. Team data were collected in 1992, 1994, and 1995. The number of times patient data were collected was dependent on the length of time the patient was treated and varied from 1 to 14 between 1992 and 1999. Key variables included: patients' ADL scores (the dependent variable); measures of team participation and team functioning; the number of days from baseline on which a patient's ADLs were assessed; and several control variables.
Data Collection Methods
Team data were obtained via self-administered questionnaires distributed to staff on the study teams. Additional team data were obtained via questionnaires completed by unit directors contemporaneously with the staff survey. Patient data were collected by trained clinicians at regular intervals using a standard assessment instrument.
Results indicated that patients treated in teams with higher levels of staff participation experienced greater improvement in ADL over time. No differences in ADL change were noted for patients treated in teams with higher levels of team functioning.
Findings support our premise that team process has important implications for patient outcomes. The results suggest that the level of participation by the team as a whole may be a more important process attribute, in terms of patient improvements in ADLs, than the team's smooth functioning. These findings indicate the potential appropriateness of managerial interventions to encourage member investment in team processes.
Cross-functional teams; mental health; activities of daily living; participation on teams; team functioning
This study aims to accurately predict patient mortality in the ICU. Given all physiologic measurements in the first 48 hours of the ICU stay, the Bayesian model of the study predicts outcome with a posterior probability.
This study modeled the outcome as a binary random variable dependent on trends of daily physiologic measures of the patient, where trends were conditionally independent given the outcome. A two-day trend is a sequence of two discrete values, one for each day. Each value (low, medium, high or unmeasured) is a function of the arithmetic mean of that measure on the corresponding day.
The prediction performance of the model was measured as the minimum of sensitivity and positive predictive values. The model yielded a score of 0.39 along with a Hosmer-Lemeshow H statistic of 36, which measures calibration. The perfect scores would be 1.0 and 0, respectively.
The prediction performance of the study was an improvement over the established ICU scoring metric SAPS-I, whose score was 0.32. Calibration of the model outputs was comparable to that of SAPS-I.