Validation of a Bayesian learning model to predict the risk for cannabis use disorder

BACKGROUND

Cannabis use disorder (CUD) is a growing public health problem. Early identification of adolescents and young adults at risk of developing CUD in the future may help stem this trend. A logistic regression model fitted using a Bayesian learning approach was developed recently to predict the risk of future CUD based on seven risk factors in adolescence and youth. A nationally representative longitudinal dataset, Add Health was used to train the model (henceforth referred as Add Health model).

METHODS

We validated the Add Health model on two cohorts, namely, Michigan Longitudinal Study (MLS) and Christchurch Health and Development Study (CHDS) using longitudinal data from participants until they were approximately 30 years old (to be consistent with the training data from Add Health). If a participant was diagnosed with CUD at any age during this period, they were considered a case. We calculated the area under the curve (AUC) and the ratio of expected and observed number of cases (E/O). We also explored recalibrating the model to account for differences in population prevalence.

RESULTS

The cohort sizes used for validation were 424 (53 cases) for MLS and 637 (105 cases) for CHDS. AUCs for the two cohorts were 0.66 (MLS) and 0.73 (CHDS) and the corresponding E/O ratios (after recalibration) were 0.995 and 0.999.

CONCLUSION

The external validation of the Add Health model on two different cohorts lends confidence to the model's ability to identify adolescent or young adult cannabis users at high risk of developing CUD in later life.

Discrimination between Gaussian process models: active learning and static constructions

The paper covers the design and analysis of experiments to discriminate between two Gaussian process models with different covariance kernels, such as those widely used in computer experiments, kriging, sensor location and machine learning. Two frameworks are considered. First, we study sequential constructions, where successive design (observation) points are selected, either as additional points to an existing design or from the beginning of observation. The selection relies on the maximisation of the difference between the symmetric Kullback Leibler divergences for the two models, which depends on the observations, or on the mean squared error of both models, which does not. Then, we consider static criteria, such as the familiar log-likelihood ratios and the Fréchet distance between the covariance functions of the two models. Other distance-based criteria, simpler to compute than previous ones, are also introduced, for which, considering the framework of approximate design, a necessary condition for the optimality of a design measure is provided. The paper includes a study of the mathematical links between different criteria and numerical illustrations are provided.

Association between NMR metabolomic signatures of healthy lifestyle and incident coronary artery disease

AIMS

To identify metabolites associated with a healthy lifestyle and explore the possible mechanisms of lifestyle in coronary artery disease (CAD).

METHODS AND RESULTS

The nuclear magnetic resonance metabolomics platform was applied to perform metabolomic profiling of baseline plasma samples from a randomly selected subset of 121 733 UK Biobank participants. Cox proportional hazards models with covariate adjustments were used to investigate the associations between validated lifestyle-associated metabolites and incident CAD and to estimate the accuracy of the inclusion of metabolites to predict CAD compared with traditional prediction models. The discriminatory ability of each model was evaluated using Harrell's C statistic, integrated discrimination improvement (IDI), and continuous net reclassification improvement (NRI) indexes. During a median of 8.6 years of follow-up, 5513 incident CAD cases were documented. Among the 111 lifestyle-associated metabolites, 65 were significantly associated with incident CAD after multivariate adjustment (Bonferroni P < 3.11 × 10-04). The addition of these metabolites to classic risk prediction models [Framingham Risk Score (FRS) using lipids; FRS using body mass index] improved CAD prediction accuracy as assessed by the C statistic (increasing to 0.739 [95% CI, 0.731-0.747] and 0.752 [95% CI, 0.746-0.758]), respectively; continuous NRI (0.274 [0.227-0.325] and 0.266 [0.223-0.317]) and IDI (0.003 [0.002-0.004] and 0.003 [0.002-0.004]).

CONCLUSION

Healthy lifestyle-associated metabolites are associated with the incidence of CAD and may help improve the prediction of CAD risk. The use of metabolite information combined with the FRS model warrants further investigation before clinical implementation.

Concordance probability as a meaningful contrast across disparate survival times

The performance of time-to-event models is frequently assessed in part by estimating the concordance probability, which evaluates the probabilistic pairwise ordering of the model-based risk scores and survival times. The standard definition of this probability conditions on any survival time pair ordering, irrespective of whether the times are meaningfully separated. Inclusion of survival times that would be deemed clinically similar attenuates the concordance and moves the estimate away from the contrast-of-interest: comparing the risk scores between individuals with disparate survival times. In this manuscript, we propose a concordance definition and corresponding method to estimate the probability conditional on survival times being separated by at least a minimum difference. The proposed estimate requires direct input from the analyst to identify a separable survival region and, in doing so, is analogous to the clinically defined subgroups used for binary outcome area under the curve estimates. The method is illustrated in two cancer examples: a prognostic score in clear cell renal cell carcinoma and two biomarkers in metastatic prostate cancer.

Kinetic modeling of heterogeneous esterification reaction using initial reaction rate analysis: data extraction and evaluation of mass transfer criteria

This data article provides detailed guidance to obtain heterogeneous reaction rate expressions and the corresponding initial reaction rates and their application. Explanation is provided to deal with specific criteria to rule out internal and external concentration gradients, so that the usage of intrinsic catalytic data is guaranteed. Overall, the main goal is to provide an easy tool to evaluate both aforementioned results by simple plug-and-play of available reaction data.

Value of the Data

•

The presented data and corresponding data treatment can be put forward by other researchers in order to guarantee the acquisition of intrinsic experimental data for catalytic reactions.

•

The presented data can be used as an example to set up typical heterogeneous esterification reactions by researchers working on catalytic systems with the specific purpose of kinetic modeling.

•

Data treatment in order to calculate initial reaction rates is explained in detail with example. This has a high applicability and very easy practicability for users in the research field of heterogeneous catalysis.

•

Concentration gradients, which might destroy the intrinsic character of the experimental data, can be ruled out via simple criteria. How to use the data is explained in this manuscript.

A C-index for recurrent event data: Application to hospitalizations among dialysis patients

We propose a C-index (index of concordance) applicable to recurrent event data. The present work addresses the dearth of measures for quantifying a regression model's ability to discriminate with respect to recurrent event risk. The data which motivated the methods arise from the Dialysis Outcomes and Practice Patterns Study (DOPPS), a long-running prospective international study of end-stage renal disease patients on hemodialysis. We derive the theoretical properties of the measure under the proportional rates model (Lin et al., 2000), and propose computationally convenient inference procedures based on perturbed influence functions. The methods are shown through simulations to perform well in moderate samples. Analysis of hospitalizations among a cohort of DOPPS patients reveals substantial improvement in discrimination upon adding country indicators to a model already containing basic clinical and demographic covariates, and further improvement upon adding a relatively large set of comorbidity indicators.

Dual substrate limitation modeling and implications for mainstream deammonification

Substrate limitation occurs frequently in wastewater treatment and knowledge about microbial behavior at limiting conditions is essential for the use of biokinetic models in system design and optimization. Monod kinetics are well-accepted for modeling growth rates when a single substrate is limiting, but several models exist for treating two or more limiting substrates simultaneously. In this study three dual limitation models (multiplicative, minimum, and Bertolazzi) were compared based on experiments using nitrite-oxidizing bacteria (limited by dissolved oxygen and nitrite) and ANaerobic AMMonia-OXidizing bacteria or Aanammox (limited by ammonium and nitrite) within mixed liquor from deammonification pilots. A deterministic likelihood-based parameter estimation followed by Bayesian inference was used to estimate model-specific parameters. The minimum model outperformed the other two by a slight margin in three separate analyses. 1) Parameters estimated using the minimum model were closest to parameters estimated from single limitation batch tests. 2) Among simulations based on each model's own estimated parameters, the minimum model best described the experimental observations. 3) Among simulations based on parameters estimated from single limitation, the minimum model best described the experimental observations. The dual substrate model selected among the three studied can effect a 75% process performance variation based on simulations of a full-scale mainstream deammonification system.

How many longitudinal covariate measurements are needed for risk prediction?

OBJECTIVE

In epidemiologic follow-up studies, many key covariates, such as smoking, use of medication, blood pressure, and cholesterol, are time varying. Because of practical and financial limitations, time-varying covariates cannot be measured continuously, but only at certain prespecified time points. We study how the number of these longitudinal measurements can be chosen cost-efficiently by evaluating the usefulness of the measurements for risk prediction.

STUDY DESIGN AND SETTING

The usefulness is addressed by measuring the improvement in model discrimination between models using different amounts of longitudinal information. We use simulated follow-up data and the data from the Finnish East-West study, a follow-up study, with eight longitudinal covariate measurements carried out between 1959 and 1999.

RESULTS

In a simulation study, we show how the variability and the hazard ratio of a time-varying covariate are connected to the importance of remeasurements. In the East-West study, it is seen that for older people, the risk predictions obtained using only every other measurement are almost equivalent to the predictions obtained using all eight measurements.

CONCLUSION

Decisions about the study design have significant effects on the costs. The cost-efficiency can be improved by applying the measures of model discrimination to data from previous studies and simulations.

Optimal experimental design for discriminating between microbial growth models as function of suboptimal temperature

In the field of predictive microbiology, mathematical models play an important role for describing microbial growth, survival and inactivation. Often different models are available for describing the microbial dynamics in a similar way. However, the model that describes the system in the best way is desired. Optimal experimental design for model discrimination (OED-MD) is an efficient tool for discriminating among rival models. In this work the T12-criterion proposed by Atkinson and Fedorov (1975) [1] and applied efficiently by Ucinski and Bogacka (2005) [2] and the Schwaab-approach proposed by Schwaab et al. (2008) [3] and Donckels et al. (2009) [4] will be applied for discriminating among rival models for the microbial growth rate as a function of temperature. The two methods will be tested in silico and their performances will be compared. Results from a simulation study indicate that it is possible to validate the case that one of the proposed models is more accurate for describing the temperature effect on the microbial growth rate. Both methods are able to design inputs with a sufficient discrimination potential. However, it has been observed that the Schwaab-approach provides inputs with a higher discrimination potential in combination with more accurate parameter estimates.