Predicting haemoglobin deferral using machine learning models: Can we use the same prediction model across countries?

BACKGROUND AND OBJECTIVES: Personalized donation strategies based on haemoglobin (Hb) prediction models may reduce Hb deferrals and hence costs of donation, meanwhile improving commitment of donors. We previously found that prediction models perform better in validation data with a high Hb deferral rate. We therefore investigate how Hb deferral prediction models perform when exchanged with other blood establishments. MATERIALS AND METHODS: Donation data from the past 5 years from random samples of 10,000 donors from Australia, Belgium, Finland, the Netherlands and South Africa were used to fit random forest models for Hb deferral prediction. Trained models were exchanged between blood establishments. Model performance was evaluated using the area under the precision-recall curve (AUPR). Variable importance was assessed using SHapley Additive exPlanations (SHAP) values. RESULTS: Across the validation datasets and exchanged models, the AUPR ranged from 0.05 to 0.43. Exchanged models performed similarly within validation datasets, irrespective of the origin of the training data. Apart from subtle differences, the importance of most predictor variables was similar in all trained models. CONCLUSION: Our results suggest that Hb deferral prediction models trained in different blood establishments perform similarly within different validation datasets, regardless of the deferral rate of their training data. Models learn similar associations in different blood establishments.

Spontaneous recovery from neuromuscular block after a single dose of a muscle relaxant in pediatric patients: A systematic review using a network meta-analytic and meta-regression approach.

BACKGROUND: Age-related differences in the pharmacokinetics and pharmacodynamics of neuromuscular blocking agents (NMBAs) and the short duration of many surgical procedures put pediatric patients at risk of postoperative residual curarization (PORC). To date, the duration of neuromuscular blocking agent effect in children has not been analyzed in a quantitative review. The current meta-analysis aimed to compare spontaneous recovery following administration of various types and doses of neuromuscular blocking agents and to quantify the effect of prognostic variables associated with the recovery time in pediatric patients. METHOD: We searched for randomized controlled trials (RCTs) and controlled clinical trials (CCTs) that compared the time to 25% T1 (t25), from 25% to 75% T1 (RI25-75), and to ≥90% train-of-four (tTOF90) neuromuscular recovery between common neuromuscular blocking agent treatments administered as a single bolus to healthy pediatric participants. We compared spontaneous t25, RI25-75, and tTOF90 between (1) neuromuscular blocking agent treatments and (2) age groups receiving a given neuromuscular blocking agent intervention and anesthesia technique. Bayesian random-effects network and pairwise meta-analyses along with meta-regression were used to evaluate the results. RESULTS: We used data from 71 randomized controlled trials/controlled clinical trials including 4319 participants. Network meta-analysis allowed for the juxtaposition and ranking of spontaneous t25, RI25-75, and tTOF90 following common neuromuscular blocking agent interventions. For all neuromuscular blocking agents a log-linear relationship between dose and duration of action was found. With the neuromuscular blocking agent treatments studied, the average tTOF90 (mean[CrI95]) in children (>2-11 y) was 41.96 [14.35, 69.50] and 17.06 [5.99, 28.30] min shorter than in neonates (<28 d) and infants (28 d-12 M), respectively. We found a negative log-linear correlation between age and duration of neuromuscular blocking agent effect. The difference in the tTOF90 (mean[CrI95]) between children and other age groups increased by 21.66 [8.82, 34.53] min with the use of aminosteroid neuromuscular blocking agents and by 24.73 [7.92, 41.43] min with the addition of sevoflurane/isoflurane for anesthesia maintenance. CONCLUSIONS: The times to neuromuscular recovery are highly variable. These can decrease significantly with age and are prolonged when volatile anesthetics are administered. This variability, combined with the short duration of many pediatric surgical procedures, makes quantitative neuromuscular monitoring mandatory even after a single dose of neuromuscular blocking agent.

An international comparison of haemoglobin deferral prediction models for blood banking.

BACKGROUND AND OBJECTIVES: Blood banks use a haemoglobin (Hb) threshold before blood donation to minimize donors' risk of anaemia. Hb prediction models may guide decisions on which donors to invite, and should ideally also be generally applicable, thus in different countries and settings. In this paper, we compare the outcome of various prediction models in different settings and highlight differences and similarities. MATERIALS AND METHODS: Donation data of repeat donors from the past 5 years of Australia, Belgium, Finland, the Netherlands and South Africa were used to fit five identical prediction models: logistic regression, random forest, support vector machine, linear mixed model and dynamic linear mixed model. Only donors with five or more donation attempts were included to ensure having informative data from all donors. Analyses were performed for men and women separately and outcomes compared. RESULTS: Within countries and overall, different models perform similarly well. However, there are substantial differences in model performance between countries, and there is a positive association between the deferral rate in a country and the ability to predict donor deferral. Nonetheless, the importance of predictor variables across countries is similar and is highest for the previous Hb level. CONCLUSION: The limited impact of model architecture and country indicates that all models show similar relationships between the predictor variables and donor deferral. Donor deferral is found to be better predictable in countries with high deferral rates. Therefore, such countries may benefit more from deferral prediction models than those with low deferral rates.

A combined multilevel factor analysis and covariance regression model with mixed effects in the mean and variance structure.

Li et al developed a multilevel covariance regression (MCR) model as an extension of the covariance regression model of Hoff and Niu. This model assumes a hierarchical structure for the mean and the covariance matrix. Here, we propose the combined multilevel factor analysis and covariance regression model in a Bayesian framework, simultaneously modeling the MCR model and a multilevel factor analysis (MFA) model. The proposed model replaces the responses in the MCR part with the factor scores coming from an MFA model. Via a simulation study and the analysis of real data, we show that the proposed model is quite efficient when the responses of the MCR model are not measured directly but are latent variables such as the patient experience measurements in our motivating dataset.

Modeling local dependence in latent vector autoregressive models.

We propose a Bayesian latent vector autoregressive (LVAR) model to analyze multivariate longitudinal data of binary and ordinal variables (items) as a function of a small number of continuous latent variables. We focus on the evolution of the latent variables while taking into account the correlation structure of the responses. Often local independence is assumed in this context. Local independence implies that, given the latent variables, the responses are assumed mutually independent cross-sectionally and longitudinally. However, in practice conditioning on the latent variables may not remove the dependence of the responses. We address local dependence by further conditioning on item-specific random effects. A simulation study shows that wrongly assuming local independence may give biased estimates for the regression coefficients of the LVAR process as well as the item-specific parameters. Novel features of our proposal include (i) correcting biased estimates of the model parameters, especially the regression coefficients of the LVAR process, obtained when local dependence is ignored and (ii) measuring the magnitude of local dependence. We applied our model on data obtained from a registry on the elderly population in Belgium. The purpose was to examine the values of oral health information on top of general health information.

The power prior with multiple historical controls for the linear regression model.

Combining historical control data with current control data may reduce the necessary study size of a clinical trial. However, this only applies when the historical control data are similar enough to the current control data. Several Bayesian approaches for incorporating historical data in a dynamic way have been proposed, such as the meta-analytic-predictive (MAP) prior and the modified power prior (MPP). Here we discuss the generalization of the MPP approach for multiple historical control groups for the linear regression model. This approach is useful when the controls differ more than in a random way, but become again (approximately) exchangeable conditional on covariates. The proposed approach builds on the approach previously developed for binary outcomes by some of the current authors. Two MPP approaches have been developed with multiple controls. The first approach assumes independent powers, while in the second approach the powers have a hierarchical structure. We conducted several simulation studies to investigate the frequentist characteristics of borrowing methods and analyze a real-life data set. When there is between-study variation in the slopes of the model or in the covariate distributions, the MPP approach achieves approximately nominal type I error rates and greater power than the MAP prior, provided that the covariates are included in the model. When the intercepts vary, the MPP yields a slightly inflated type I error rate, whereas the MAP does not. We conclude that our approach is a worthy competitor to the MAP approach for the linear regression case.

Incorporating historical controls in clinical trials with longitudinal outcomes using the modified power prior.

Several dynamic borrowing methods, such as the modified power prior (MPP), the commensurate prior, have been proposed to increase statistical power and reduce the required sample size in clinical trials where comparable historical controls are available. Most methods have focused on cross-sectional endpoints, and appropriate methodology for longitudinal outcomes is lacking. In this study, we extend the MPP to the linear mixed model (LMM). An important question is whether the MPP should use the conditional version of the LMM (given the random effects) or the marginal version (averaged over the distribution of the random effects), which we refer to as the conditional MPP and the marginal MPP, respectively. We evaluated the MPP for one historical control arm via a simulation study and an analysis of the data of Alzheimer's Disease Cooperative Study (ADCS) with the commensurate prior as the comparator. The conditional MPP led to inflated type I error rate when there existed moderate or high between-study heterogeneity. The marginal MPP and the commensurate prior yielded a power gain (3.6%-10.4% vs. 0.6%-4.6%) with the type I error rates close to 5% (5.2%-6.2% vs. 3.8%-6.2%) when the between-study heterogeneity is not excessively high. For the ADCS data, all the borrowing methods improved the precision of estimates and provided the same clinical conclusions. The marginal MPP and the commensurate prior are useful for borrowing historical controls in longitudinal data analysis, while the conditional MPP is not recommended due to inflated type I error rates.

Latent Ornstein-Uhlenbeck models for Bayesian analysis of multivariate longitudinal categorical responses.

We propose a Bayesian latent Ornstein-Uhlenbeck (OU) model to analyze unbalanced longitudinal data of binary and ordinal variables, which are manifestations of fewer continuous latent variables. We focus on the evolution of such latent variables when they continuously change over time. Existing approaches are limited to data collected at regular time intervals. Our proposal makes use of an OU process for the latent variables to overcome this limitation. We show that assuming real eigenvalues for the drift matrix of the OU process, as is frequently done in practice, can lead to biased estimates and/or misleading inference when the true process is oscillating. In contrast, our proposal allows for both real and complex eigenvalues. We illustrate our proposed model with a motivating dataset, containing patients with amyotrophic lateral sclerosis disease. We were interested in how bulbar, cervical, and lumbar functions evolve over time.

Nonlinear mixed-effects modeling of longitudinal count data: Bayesian inference about median counts based on the marginal zero-inflated discrete Weibull distribution.

This article proposes a Bayesian regression model for nonlinear zero-inflated longitudinal count data that models the median count as an alternative to the mean count. The nonlinear model generalizes a recently introduced linear mixed-effects model based on the zero-inflated discrete Weibull (ZIDW) distribution. The ZIDW distribution is more robust to severe skewness in the data than conventional zero-inflated count distributions such as the zero-inflated negative binomial (ZINB) distribution. Moreover, the ZIDW distribution is attractive because of its convenience to model the median counts given its closed-form quantile function. The median is a more robust measure of central tendency than the mean when the data, for instance, zero-inflated counts, are right-skewed. In an application of the model we consider a biphasic mixed-effects model consisting of an intercept term and two slope terms. Conventionally, the ZIDW model separately specifies the predictors for the zero-inflation probability and the counting process's median count. In our application, the two latent class interpretations are not clinically plausible. Therefore, we propose a marginal ZIDW model that directly models the biphasic median counts marginally. We also consider the marginal ZINB model to make inferences about the nonlinear mean counts over time. Our simulation study shows that the models have good properties in terms of accuracy and confidence interval coverage.

Serial correlation structures in latent linear mixed models for analysis of multivariate longitudinal ordinal responses.

We propose a latent linear mixed model to analyze multivariate longitudinal data of multiple ordinal variables, which are manifestations of fewer continuous latent variables. We focus on the latent level where the effects of observed covariates on the latent variables are of interest. We incorporate serial correlation into the variance component rather than assuming independent residuals. We show that misleading inference may be drawn when misspecifying the variance component. Furthermore, we provide a graphical tool depicting latent empirical semi-variograms to detect serial correlation for latent stationary linear mixed models. We apply our proposed model to examine the treatment effect on patients having the amyotrophic lateral sclerosis disease. The result shows that the treatment can slow down progression of latent cervical and lumbar functions.

Assessment of oral health in older adults by non-dental professional caregivers-development and validation of a photograph-supported oral health-related section for the interRAI suite of instruments.

OBJECTIVES: An optimized oral health-related section and a video training were developed and validated for the interRAI suite of instruments. The latter is completed by professional non-dental caregivers and used in more than 40 countries to assess care needs of older adults. METHODS: The optimized oral health-related section (ohr-interRAI) consists of nine items and a video training that were developed in consecutive phases. To evaluate psychometric properties, a study was conducted in 260 long-term care residents. Each resident was assessed by a dentist and by four caregivers (two who received the video training, two who did not). RESULTS: Mean kappa values and percent agreement between caregivers and dentist ranged between κ = 0.60 (80.2%) for dry mouth and κ = 0.13 (54.0%) for oral hygiene. The highest inter-caregiver agreement was found for dry mouth with κ = 0.63 [95% CI: 0.56-0.70] (81.6%), while for the item palate/lips/cheeks only κ = 0.27 [95% CI: 0.18-0.36] (76.7%) was achieved. Intra-caregiver agreement ranged between κ = 0.93 [95% CI: 0.79-1.00] (96.4%) for dry mouth and κ = 0.45 [95% CI: 0.06-0.84] (82.8%) for gums. Logistic regression analysis showed only small differences between caregivers who watched the video training and those who did not. CONCLUSIONS: Psychometric properties of the optimized ohr-interRAI section were improved compared to previous versions. Nevertheless, particularly the items based on inspection of the mouth require further refinement and caregiver training needs to be improved. CLINICAL RELEVANCE: Valid assessment of oral health by professional caregivers is essential due to the impaired accessibility of regular dental care for care-dependent older adults.

Summarising salient information on historical controls: A structured assessment of validity and comparability across studies.

BACKGROUND: While placebo-controlled randomised controlled trials remain the standard way to evaluate drugs for efficacy, historical data are used extensively across the development cycle. This ranges from supplementing contemporary data to increase the power of trials to cross-trial comparisons in estimating comparative efficacy. In many cases, these approaches are performed without in-depth review of the context of data, which may lead to bias and incorrect conclusions. METHODS: We discuss the original 'Pocock' criteria for the use of historical data and how the use of historical data has evolved over time. Based on these factors and personal experience, we created a series of questions that may be asked of historical data, prior to their use. Based on the answers to these questions, various statistical approaches are recommended. The strategy is illustrated with a case study in colorectal cancer. RESULTS: A number of areas need to be considered with historical data, which we split into three categories: outcome measurement, study/patient characteristics (including setting and inclusion/exclusion criteria), and disease process/intervention effects. Each of these areas may introduce issues if not appropriately handled, while some may preclude the use of historical data entirely. We present a tool (in the form of a table) for highlighting any such issues. Application of the tool to a colorectal cancer data set demonstrates under what conditions historical data could be used and what the limitations of such an analysis would be. CONCLUSION: Historical data can be a powerful tool to augment or compare with contemporary trial data, though caution is required. We present some of the issues that may be considered when involving historical data and what (if any) statistical approaches may account for differences between studies. We recommend that, where historical data are to be used in analyses, potential differences between studies are addressed explicitly.

Assessment of oral health conditions presented in photographs - is there a difference between dentists and non-dental professional caregivers?

BACKGROUND: Photographs can help non-dental professional caregivers to identify problems when inspecting the mouth of care-dependent older individuals. This study evaluated whether the assessment of oral health-related conditions presented in photographs differed between dentists and non-dental professional caregivers. MATERIALS AND METHODS: One-hundred-and-seventy-nine photographs were taken from long-term care facility residents and from patients at the Department of Dentistry of a University Hospital. The following oral health aspects were depicted: denture hygiene, oral hygiene, teeth, gums, tongue and palate/lips/cheeks. Collection continued until for each oral health aspect a pool of photographs was available that showed conditions from perfect health and hygiene to severe problems. A segmented Visual Analogue Scale was applied to assess the conditions presented in the photographs. Each photograph was assessed by each participant of this study. The benchmark was established by three dentists with academic-clinical expertise in gerodontology, special needs dentistry and periodontology. For each photograph, they provided a collective score after reaching consensus. Photographs were assessed individually by 32 general dentists and by 164 non-dental professional caregivers. Linear mixed effects models and generalized linear mixed effects models were fitted and mean squared errors were computed to quantify differences between both groups. RESULTS: For the different oral health aspects, absolute distances from the benchmark scores were 1.13 (95%CI:1.03-1.23) to 1.51 (95%CI:1.39-1.65) times higher for the caregivers than for the dentists. The odds to overestimate the condition were higher for the caregivers than the dentists for oral hygiene (OR = 0.72, 95%CI = 0.62-0.84) and teeth (OR = 0.74; 95%CI = 0.61-0.88). The odds to underestimate the condition were higher for the caregivers than the dentists for gums (OR = 1.39; 95%CI:1.22-1.59) and palate/lips/cheeks (OR = 1.22; 95%CI = 1.07-1.40). Over all assessments, the variance in caregiver scores was 1.9 (95%CI:1.62-2.23) times higher than that for the dentists. CONCLUSION: Small but significant differences were found between dentists and non-dental professional caregivers assessing oral health-related conditions presented in photographs. When photographs are used to aid non-dental professional caregivers with the oral health assessment, these visualizations should be complemented with comments to facilitate accurate interpretation.

Contribution of HCAHPS Specific Care Experiences to Global Ratings Varies Across 7 Countries: What Can be Learned for Reporting These Global Ratings?

BACKGROUND: Given the increased international interest in improvement strategies for patient experiences with care, it is important to understand whether the same specific care experiences affect global ratings across countries. Moreover, reporting of these global ratings currently substantially varies in both research and public reporting. OBJECTIVES: The objectives of this study were to examine the differential magnitude of associations between Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) specific care experiences and global ratings, and to explore implications for reporting these global ratings. METHODS: HCAHPS data were collected from 11,289 patients across 7 European countries. We studied the association between 11 experience items and 2 global ratings (0-10 hospital rating, hospital recommendation) using multilevel ordered logistic regression analysis. Using interaction terms, we examined consistency of these associations across countries. Assuming homogeneous use and interpretations of response categories of these specific experiences across patients within and between countries, we investigated what the associations between specific experiences and global ratings imply for reporting global ratings across countries. RESULTS: All specific experiences were associated with both global ratings. "Being talked to about care after leaving the hospital" showed the strongest association. There were relatively little differential effects across countries. A reporting strategy with different cutpoints across countries might provide a more equitable comparison of global ratings. CONCLUSIONS: Our findings highlight a differential contribution of HCAHPS specific experiences to overall ratings across countries. Research and public reporting may consider examining in more detail within and across populations global rating cutpoints that represent the same true level of positive patient experiences.

Modified power prior with multiple historical trials for binary endpoints.

Including historical data may increase the power of the analysis of a current clinical trial and reduce the sample size of the study. Recently, several Bayesian methods for incorporating historical data have been proposed. One of the methods consists of specifying a so-called power prior whereby the historical likelihood is downweighted with a weight parameter. When the weight parameter is also estimated from the data, the modified power prior (MPP) is needed. This method has been used primarily when a single historical trial is available. We have adapted the MPP for incorporating multiple historical control arms into a current clinical trial, each with a separate weight parameter. Three priors for the weights are considered: (1) independent, (2) dependent, and (3) robustified dependent. The latter is developed to account for the possibility of a conflict between the historical data and the current data. We analyze two real-life data sets and perform simulation studies to compare the performance of competing Bayesian methods that allow to incorporate historical control patients in the analysis of a current trial. The dependent power prior borrows more information from comparable historical studies and thereby can improve the statistical power. Robustifying the dependent power prior seems to protect against prior-data conflict.

Power Distance and Physician-Nurse Collegial Relations Across 14 European Countries: National Culture is Not Merely a Nuisance Factor in International Comparative Research.

PURPOSE: This study illustrates the huge untapped potential of quantifying the impact of culture in making meaningful comparisons across groups. Our focus is on cross-national differences in nurses' reports of their relations with physicians, and how the measurement of this complex construct and the evaluation of true differences are related to dimensions of national culture. DESIGN: We examine across 14 European countries the association between indices of national culture from the seminal work of Hofstede and 39,435 nurses' ratings of their relations with physicians. Multilevel confirmatory factor analysis was used to evaluate strong factorial invariance across countries and to examine the influence of power distance and masculinity. FINDINGS: There was wide variation across countries in nurses' reports of their relations with physicians. Strong factorial invariance was shown for a one-factor model, which confirmed that across countries the seven survey items measure a common factor of physician-nurse relations. This model showed no country bias for any of the seven survey items, which suggests that differences across countries reflect true differences. These true differences were significantly associated with variation in country values of power distance, which showed a significant negative correlation with physician-nurse relations. CONCLUSIONS: Continuously pursuing a better understanding of characteristics that impact the studied indicators, such as national culture, is elementary to better understand the construct under study. In this application, country values of power distance negatively impacted nurse-reported relations with physicians, which strongly varied across countries. CLINICAL RELEVANCE: Better nurse-reported relations between nurses and physicians link to higher nurse job satisfaction, lower emotional exhaustion, better nurse-perceived quality of care, and lower patient mortality. The Practice Environment Scale of the Nursing Work Index is an excellent instrument to characterize variation in working relations between nurses and physicians as well as physicians' professional posture towards nurses.

The oral health-related section of the interRAI: Evaluation of test content validity by expert rating and assessment of potential reasons for inaccurate assessments based on focus group discussions with caregivers.

OBJECTIVES: To explore the failure of the oral health-related section of the interRAI (ohr-interRAI), this study investigated test content validity (A.) and reasons for inaccurate assessments (B.). BACKGROUND: Poor oral health negatively affects quality of life and is associated with a number of systemic diseases. The interRAI instruments, internationally used for geriatric assessment, should accurately detect oral conditions that require care. Previous research showed that the ohr-interRAI and related precursor versions do not achieve this goal. MATERIALS AND METHODS: (A.) A group of 12 experts rated completeness, relevance, clarity of wording and feasibility of the ohr-interRAI. Content validity indices were calculated per item (threshold 0.78). (B.) Focus group discussions with 23 caregivers were organized. A semi-structured question guide made sure that all topics of interest were covered. Qualitative content structuring analysis was applied after transcription. RESULTS: (A.) Experts agreed on the relevance of the items on chewing, pain, gingival inflammation and damaged teeth. They regarded none of the items as worded clearly and only prosthesis use and pain were considered to be assessable by untrained caregivers. All experts agreed that the ohr-interRAI was incomplete. (B.) Focus group discussions revealed that in the care environment oral health had low priority. Aspects related to the ohr-interRAI itself and aspects related to the assessment situation impeded the oral health assessment. The approach of the caregivers to complete the ohr-interRAI was inappropriate to accurately detect oral care needs. CONCLUSIONS: Findings challenge test content validity of the ohr-interRAI and reveal reasons for inaccurate assessments.

Noninferiority Clinical Trials: The Good, the Bad, and the Ugly.

For decades, the superiority trial has been the most popular design to assess the efficacy of newly developed drugs in a randomized controlled clinical trial. In a superiority trial, the aim is to show that the new (experimental) treatment is better than the standard treatment or placebo. However, it becomes increasingly difficult to improve the efficacy upon that of existing drugs. For this reason, noninferiority designs have been suggested. In a noninferiority study, one aims to show that the experimental treatment does not lower the efficacy of the standard treatment too much, but this loss of efficacy should be compensated by other better properties. In this article, the design, aims, and properties of the superiority and the noninferiority trial are contrasted and illustrated on recently published studies to treat patients with advanced hepatocellular carcinoma. The author discusses the reasons why noninferiority studies are becoming popular, but also why the results of noninferiority studies may be difficult to interpret and can be easily misused. Since only a few noninferiority studies in hepatocellular cancer have been organized, also examples from other therapeutic areas were taken. Finally, it is indicated how to appreciate the qualities of published noninferiority studies.

Comparing hierarchical models via the marginalized deviance information criterion.

Hierarchical models are extensively used in pharmacokinetics and longitudinal studies. When the estimation is performed from a Bayesian approach, model comparison is often based on the deviance information criterion (DIC). In hierarchical models with latent variables, there are several versions of this statistic: the conditional DIC (cDIC) that incorporates the latent variables in the focus of the analysis and the marginalized DIC (mDIC) that integrates them out. Regardless of the asymptotic and coherency difficulties of cDIC, this alternative is usually used in Markov chain Monte Carlo (MCMC) methods for hierarchical models because of practical convenience. The mDIC criterion is more appropriate in most cases but requires integration of the likelihood, which is computationally demanding and not implemented in Bayesian software. Therefore, we consider a method to compute mDIC by generating replicate samples of the latent variables that need to be integrated out. This alternative can be easily conducted from the MCMC output of Bayesian packages and is widely applicable to hierarchical models in general. Additionally, we propose some approximations in order to reduce the computational complexity for large-sample situations. The method is illustrated with simulated data sets and 2 medical studies, evidencing that cDIC may be misleading whilst mDIC appears pertinent.

Multilevel moderated mediation model with ordinal outcome.

Although increasingly complex models have been proposed in mediation literature, there is no model nor software that incorporates the multiple possible generalizations of the simple mediation model jointly. We propose a flexible moderated mediation model allowing for (1) a hierarchical structure of clustered data, (2) more and possibly correlated mediators, and (3) an ordinal outcome. The motivating data set is obtained from a European study in nursing research. Patients' willingness to recommend their treating hospital was recorded in an ordinal way. The research question is whether such recommendation directly depends on system-level features in the organization of nursing care, or whether these associations are mediated by 2 measurements of nursing care left undone and possibly moderated by nurse education. We have developed a Bayesian approach and accompanying program that takes all the above generalizations into account.