Search

Introduction Dans la recherche en surveillance de la santé de la population, les données d’enquête sont couramment analysées à l’aide de méthodes de régression. Or ces méthodes disposent d’une capacité limitée à analyser les relations complexes. À l’opposé, les modèles d’arbres décisionnels sont parfaitement adaptés pour segmenter les populations et étudier les interactions complexes entre facteurs, et leur utilisation dans la recherche en santé est en pleine croissance. Cet article fournit un aperçu de la méthodologie des arbres décisionnels et de leur application aux données d’enquête sur la santé mentale des jeunes. Methods La performance de deux techniques courantes de construction d’arbres décisionnels, soit l’arbre de classification et de régression (CART) et l’arbre d’inférence conditionnelle (CTREE), est comparée aux modèles classiques de régression linéaire et logistique par le biais d’une application aux résultats en santé mentale des jeunes de l’étude COMPASS. Les données ont été recueillies auprès de 74 501 élèves de 136 écoles au Canada. L’anxiété, la dépression et le bien‑être psychologique ont été mesurés, de même que 23 variables sociodémographiques et facteurs de prédiction des comportements liés à la santé. La performance du modèle a été évaluée au moyen de mesures de prédiction de l’exactitude, de la parcimonie et de l’importance relative des variables. Results Les modèles d’arbres décisionnels et les modèles de régression ont systématiquement mis en évidence les mêmes ensembles de facteurs de prédiction les plus importants pour chaque résultat, ce qui indique un niveau général de concordance entre méthodes. Trois modèles ont présenté une exactitude prédictive plus faible, mais se caractérisent par une plus grande parcimonie et accordent une importance relative plus élevée aux principaux facteurs de différenciation. Conclusion Les arbres décisionnels permettent de cerner les sous-groupes à risque élevé qu’il convient de cibler dans le cadre des efforts de prévention et d’intervention. Ils constituent donc un outil utile pour répondre aux questions de recherche auxquelles les méthodes de régression classiques ne peuvent pas répondre.

Large cohort studies now routinely involve biobanks in which biospecimens are stored for use in future biomarker studies. In such settings, two-phase response-dependent sampling designs involve subsampling individuals in the cohort, assaying their biospecimen to measure an expensive biomarker, and using this data to estimate key parameters of interest under budgetary constraints. When analyses are based on inverse probability weighted estimating functions, recent work has described adaptive two-phase designs in which a preliminary phase of subsampling based on a standard design facilitates approximation of an optimal selection model for a second subsampling phase. In this article, we refine the definition of an optimal subsampling scheme within the framework of adaptive two-phase designs, describe how adaptive two-phase designs can be used when analyses are based on likelihood or conditional likelihood, and consider the setting of a continuous biomarker where the nuisance covariate distribution is estimated nonparametrically at the design stage and analysis stage as required; efficiency and robustness issues are investigated. We also explore these methods for the surrogate variable problem and describe a generalization to accommodate multiple stages of phase II subsampling. A study involving individuals with psoriatic arthritis is considered for illustration, where the aim is to assess the association between the biomarker MMP-3 and the development of joint damage.

Modifiable environmental and behavioural factors influence youth mental health; however, past studies have primarily used regression models that quantify population average effects. Decision trees are an analytic technique that examine complex relationships between factors and identify high-risk subgroups to whom intervention measures can be targeted. This study used decision trees to examine associations of various risk factors with youth anxiety, depression, and flourishing. Data were collected from 74,501 students across Canadian high schools participating in the 2018–2019 COMPASS Study. Students completed a questionnaire including validated mental health scales and 23 covariates. Decision trees were grown to identify key factors and subgroups for anxiety, depression, and flourishing outcomes. Females lacking both happy home life and sense of connection to school were at greatest risk for higher anxiety and depression levels. In contrast with previous literature, behavioural factors such as diet, movement and substance use did not emerge as differentiators. This study highlights the influence of home and school environments on youth mental health using a novel decision tree analysis. While having a happy home life is most important in protecting against youth anxiety and depression, a sense of connection to school may mitigate the negative influence of a poor home environment.

Borrowing information from populations with similar structural mortality patterns and trajectories has been well recognized as a useful strategy for the mortality forecasting of a target population. This paper presents a flexible framework for the selection of populations from a given candidate pool to assist a target population in mortality forecasting. The defining feature of the framework is the deletion-substitution-addition (DSA) algorithm, which is entirely data-driven and versatile to work with any multiple-population model for mortality prediction. In numerical studies, the framework with an extended augmented common factor model is applied to the Human Mortality Database, and the superiority of the proposed framework is evident in mortality forecasting performance.

This article proposes a novel paradigm for building regression trees and ensemble learning in survival analysis. Generalizations of the classification and regression trees (CART) and random forests (RF) algorithms for general loss functions, and in the latter case more general bootstrap procedures, are both introduced. These results, in combination with an extension of the theory of censoring unbiased transformations (CUTs) applicable to loss functions, underpin the development of two new classes of algorithms for constructing survival trees and survival forests: censoring unbiased regression trees and censoring unbiased regression ensembles. For a certain “doubly robust” CUT of squared error loss, we further show how these new algorithms can be implemented using existing software (e.g., CART, RF). Comparisons of these methods to existing ensemble procedures for predicting survival probabilities are provided in both simulated settings and through applications to four datasets. It is shown that these new methods either improve upon, or remain competitive with, existing implementations of random survival forests, conditional inference forests, and recursively imputed survival trees.

Estimating a patient's mortality risk is important in making treatment decisions. Survival trees are a useful tool and employ recursive partitioning to separate patients into different risk groups. Existing 'loss based' recursive partitioning procedures that would be used in the absence of censoring have previously been extended to the setting of right censored outcomes using inverse probability censoring weighted estimators of loss functions. In this paper, we propose new 'doubly robust' extensions of these loss estimators motivated by semiparametric efficiency theory for missing data that better utilize available data. Simulations and a data analysis demonstrate strong performance of the doubly robust survival trees compared with previously used methods. Copyright © 2016 John Wiley & Sons, Ltd.

This paper applies machine learning techniques to credibility theory and proposes a regression-tree-based algorithm to integrate covariate information into credibility premium prediction. The recursive binary algorithm partitions a collective of individual risks into mutually exclusive sub-collectives, and applies the classical Buhlmann-Straub credibility formula for the prediction of individual net premiums. The algorithm provides a flexible way to integrate covariate information into individual net premiums prediction. It is appealing for capturing non-linear and/or interaction covariate effects. It automatically selects influential covariate variables for premium prediction and requires no additional ex-ante variable selection procedure. The superiority in the prediction accuracy of the proposed algorithm is demonstrated by extensive simulation studies. The proposed method is applied to the U.S. Medicare data for illustration purposes.

Many chronic diseases feature recurring clinically important events. In addition, however, there often exists a random variable which is realized upon the occurrence of each event reflecting the severity of the event, a cost associated with it, or possibly a short term response indicating the effect of a therapeutic intervention. We describe a novel model for a marked point process which incorporates a dependence between continuous marks and the event process through the use of a copula function. The copula formulation ensures that event times can be modeled by any intensity function for point processes, and any multivariate model can be specified for the continuous marks. The relative efficiency of joint versus separate analyses of the event times and the marks is examined through simulation under random censoring. An application to data from a recent trial in transfusion medicine is given for illustration.

Estimating a patient's mortality risk is important in making treatment decisions. Survival trees are a useful tool and employ recursive partitioning to separate patients into different risk groups. Existing 'loss based' recursive partitioning procedures that would be used in the absence of censoring have previously been extended to the setting of right censored outcomes using inverse probability censoring weighted estimators of loss functions. In this paper, we propose new 'doubly robust' extensions of these loss estimators motivated by semiparametric efficiency theory for missing data that better utilize available data. Simulations and a data analysis demonstrate strong performance of the doubly robust survival trees compared with previously used methods. Copyright © 2016 John WileySons, Ltd.

A copula-based model is described which enables joint analysis of multiple progressive multistate processes. Unlike intensity-based or frailty-based approaches to joint modeling, the copula formulation proposed herein ensures that a wide range of marginal multistate processes can be specified and the joint model will retain these marginal features. The copula formulation also facilitates a variety of approaches to estimation and inference including composite likelihood and two-stage estimation procedures. We consider processes with Markov margins in detail, which are often suitable when chronic diseases are progressive in nature. We give special attention to the setting in which individuals are examined intermittently and transition times are consequently interval-censored. Simulation studies give empirical insight into the different methods of analysis and an application involving progression in joint damage in psoriatic arthritis provides further illustration.