A predictive Bayesian network that risk stratifies patients undergoing Barrett's surveillance for personalized risk of developing malignancy

Modeling repeated measurements data using the multilevel Bayesian network: A case of child morbidity

BACKGROUND AND OBJECTIVE

In epidemiological research, studying the long-term dependencies between multiple diseases is important. This study extends the multilevel Bayesian network (MBN) for repeated measures data that can estimate the rate of change in outcomes over time while quantifying the variabilities of these rates across higher-level units through various variance-covariance structures.

METHOD

The performance and reliability of a model are examined through a simulation study, and its practical application is demonstrated using child morbidity data. This data has a hierarchical structure in which children were randomly selected from clusters (villages) and their conditions were assessed quarterly from March 2015 to May 2016. MBN was used to explore the relationship between outcomes weight-for-age (WAZ), height-for-age (HAZ), the number of days a child suffers from diarrhea (NOD), and flu (NOF), and estimate the rate of change of these outcomes over time. Since the outcomes considered were hybrid in nature, the connected three-parent set block Gibbs sampler with a multilevel generalized Poisson regression, multilevel zero inflated Poisson regression, and linear mixed-effects models were considered during the structure and parametric learning of the MBN.

RESULT

The simulation study confirmed that a MBN using the time metric t as a node performed well for repeated measures data. The result from the structure learning of MBN shows a causal relationship between WAZ, HAZ, NOD and NOF. Furthermore, exclusive breastfeeding months and usage of micronutrient powder appeared as a strong predictor for all outcomes considered in this study.

CONCLUSION

This study reveals that MBN is suitable in modeling repeated measures data to study the relationship between outcomes and estimate rate of change of an outcome over time while quantifying the variability due to higher-level clustering variables. Furthermore, the study highlights the importance of focusing on monitoring children with low WAZ and HAZ scores together with good feeding practices against the frequency of getting flu and diarrhea.

Multilevel Bayesian network to model child morbidity using Gibbs sampling

Bayesian networks (BNs) are suitable models for studying complex interdependencies between multiple health outcomes, simultaneously. However, these models fail the assumption of independent observation in the case of hierarchical data. Therefore, this study proposes a two and three-level random intercept multilevel Bayesian network (MBN) models to study the conditional dependencies between multiple outcomes. The structure of MBN was learned using the connected three parent set block Gibbs sampler, where each local network was included based on Bayesian information criteria (BIC) score of multilevel regression. These models were examined using simulated data assuming features of both multilevel models and BNs. The estimated area under the receiver operating characteristics for both models were above 0.8, indicating good fit. The MBN was then applied to real child morbidity data from the 2016 Ethiopian Demographic Health Survey (EDHS). The result shows a complex causal dependencies between malnutrition indicators and child morbidities such as anemia, acute respiratory infection (ARI) and diarrhea. According to this result, families and health professionals should give special attention to children who suffer from malnutrition and also have one of these illnesses, as the co-occurrence of both can worsen the health of a child.

Development and validation of a prediction model for histologic progression in patients with nondysplastic Barrett's esophagus

OBJECTIVES

Patients with Barrett's esophagus (BE) are at risk of progression to esophageal adenocarcinoma (EAC). We developed a model to predict histologic progression in patients with nondysplastic BE (NDBE).

METHODS

A longitudinal study in three referral centers was performed between January 2010 and December 2019. As progression to low-grade dysplasia (LGD) can be considered an indication for ablative therapy, the study end-point was histopathologic progression to LGD, high-grade dysplasia, or EAC at 3 years after diagnosis. We used logistic regression to create the model. Seventy percent of the cohort were used to stem the model and the remaining 30% for internal validation.

RESULTS

A total of 542 patients were included, 69.4% of whom were male, mean age 62.2 years. Long-segment BE at index endoscopy was diagnosed in 20.8% of the patients. After a mean follow-up of 6.7 years, 133 patients (24.5%) had histologic progression. Our model identified a neutrophil-to-lymphocyte ratio (odds ratio [OR] 2.08, 95% confidence interval [CI] 1.77-2.32, P < 0.001), BE length (OR 1.22, 95% CI 1.09-1.36, P < 0.001), age (OR 1.03, 95% CI 1.02-1.05, P = 0.02), smoking (OR 1.66, 95% CI 1.09-2.75, P = 0.04), and renal failure (OR 1.51, 95% CI 0.93-2.43, P = 0.07) as predictors of histologic progression at 3 years. The areas under the receiver operating characteristic curves of this model were 0.88 and 0.76 in the training and validation cohorts, respectively.

CONCLUSION

This novel, internally validated model may predict histologic progression, even in patients with NDBE who generally have low rates of progression over time, and may contribute to enhanced patient selection for more intense surveillance programs.