Identification of subgroups of patients for whom treatment A is more effective than treatment B, and vice versa, is of key importance to the development of personalized medicine. Tree-based algorithms are helpful tools for the detection of such interactions, but none of the available algorithms allow for taking into account clustered or nested dataset structures, which are particularly common in psychological research. Therefore, we propose the generalized linear mixed-effects model tree (GLMM tree) algorithm, which allows for the detection of treatment-subgroup interactions, while accounting for the clustered structure of a dataset. The algorithm uses model-based recursive partitioning to detect treatment-subgroup interactions, and a GLMM to estimate the random-effects parameters. In a simulation study, GLMM trees show higher accuracy in recovering treatment-subgroup interactions, higher predictive accuracy, and lower type II error rates than linear-model-based recursive partitioning and mixed-effects regression trees. Also, GLMM trees show somewhat higher predictive accuracy than linear mixed-effects models with pre-specified interaction effects, on average. We illustrate the application of GLMM trees on an individual patient-level data metaanalysis on treatments for depression. We conclude that GLMM trees are a promising exploratory tool for the detection of treatment-subgroup interactions in clustered datasets.

For many years, psychiatrists have tried to understand factors involved in response to medications or psychotherapies, in order to personalize their treatment choices. There is now a broad and growing interest in the idea that we can develop models to personalize treatment decisions using new statistical approaches from the field of machine learning and applying them to larger volumes of data. In this pursuit, there has been a paradigm shift away from experimental studies to confirm or refute specific hypotheses towards a focus on the overall explanatory power of a predictive model when tested on new, unseen datasets. In this paper, we review key studies using machine learning to predict treatment outcomes in psychiatry, ranging from medications and psychotherapies to digital interventions and neurobiological treatments. Next, we focus on some new sources of data that are being used for the development of predictive models based on machine learning, such as electronic health records, smartphone and social media data, and on the potential utility of data from genetics, electrophysiology, neuroimaging and cognitive testing. Finally, we discuss how far the field has come towards implementing prediction tools in real‐world clinical practice. Relatively few retrospective studies to‐date include appropriate external validation procedures, and there are even fewer prospective studies testing the clinical feasibility and effectiveness of predictive models. Applications of machine learning in psychiatry face some of the same ethical challenges posed by these techniques in other areas of medicine or computer science, which we discuss here. In short, machine learning is a nascent but important approach to improve the effectiveness of mental health care, and several prospective clinical studies suggest that it may be working already.

The efficacy of treatments for depression is often measured by comparing observed total scores on self-report inventories, in both clinical practice and research. However, the occurrence of response shifts (changes in subjects' values, or their standards for measurement) may limit the validity of such comparisons. As most psychological treatments for depression are aimed at changing patients' values and frame of reference, response shifts are likely to occur over the course of such treatments. In this article, we tested whether response shifts occurred over the course of treatment in an influential randomized clinical trial. Using confirmatory factor analysis, measurement models underlying item scores on the Beck Depression Inventory (Beck & Beamesderfer, 1974) of the National Institute of Mental Health Treatment of Depression Collaborative Research Program (Elkin, Parloff, Hadley, & Autry, 1985) were analyzed. Compared with before treatment, after-treatment item scores appeared to overestimate depressive symptomatology, measurement errors were smaller, and correlations between constructs were stronger. These findings indicate a response shift, in the sense that participants seem to get better at assessing their level of depressive symptomatology. Comparing measurement models of patients receiving psychotherapy and medication suggested that the aforementioned effects were more apparent in the psychotherapy groups. Consequently, comparisons of observed total scores on self-report inventories may yield confounded measures of treatment efficacy.

BackgroundAlthough previous studies using non- elderly groups have assessed the factorial invariance of the Center for Epidemiological Studies Depression Scale (CES-D) across different groups with the same social-cultural backgrounds, few studies have tested the factorial invariance of the CES-D across two elderly groups from countries with different social cultures. The purposes of this study were to examine the factorial structure of the CES-D, and test its measurement invariance across two different national elderly populations.MethodsA total of 6806 elderly adults from China (n = 4903) and the Netherlands (n = 1903) were included in the final sample. The CES-D was assessed in both samples. Three strategies were used in the data analysis procedure. First, a confirmatory factor analysis (CFA) was carried out to determine the factor structures of the CES-D that best fitted the two samples. Second, the best fitting model was incorporated into a multi-group CFA model to test measurement invariance of the CES-D across the two population groups. Third, latent mean differences between the two groups were tested.ResultsThe results of confirmatory factor analysis (CFA) showed: 1) in both samples, Radloff's four-factor model resulted in a significantly better fit and the four dimensions (somatic complaints, depressed affect, positive affect, and interpersonal problems) of the CES-D seem to be the most informative in assessing depressive symptoms compared to the single-, three-, and the second-order factor models; and 2) the factorial structure was invariant across the populations under study. However, only partial scalar and uniqueness invariance of the CES-D items was supported. Latent means in the partial invariant model were lower for the Dutch sample, compared to the Chinese sample.ConclusionsOur findings provide evidence of a valid factorial structure of the CES-D that could be applied to elderly populations from both China and the Netherlands, producing a meaningful comparison of total scores between the two elderly groups. However, for some specific factors and items, caution is required when comparing the depressive symptoms between Chinese and Dutch elderly groups.

This study confirmed a link between depressive disorders and obesity that was influenced by lower social and physical activities among the depressed.