Mapping of machine learning approaches for description, prediction, and causal inference in the social and health sciences.

Machine learning (ML) methodology used in the social and health sciences needs to fit the intended research purposes of description, prediction, or causal inference. This paper provides a comprehensive, systematic meta-mapping of research questions in the social and health sciences to appropriate ML approaches by incorporating the necessary requirements to statistical analysis in these disciplines. We map the established classification into description, prediction, counterfactual prediction, and causal structural learning to common research goals, such as estimating prevalence of adverse social or health outcomes, predicting the risk of an event, and identifying risk factors or causes of adverse outcomes, and explain common ML performance metrics. Such mapping may help to fully exploit the benefits of ML while considering domain-specific aspects relevant to the social and health sciences and hopefully contribute to the acceleration of the uptake of ML applications to advance both basic and applied social and health sciences research.

Prediction of AD dementia by biomarkers following the NIA-AA and IWG diagnostic criteria in MCI patients from three European memory clinics.

INTRODUCTION: Proposed diagnostic criteria (international working group and National Institute on Aging and Alzheimer's Association) for Alzheimer's disease (AD) include markers of amyloidosis (abnormal cerebrospinal fluid [CSF] amyloid beta [Aß]42) and neurodegeneration (hippocampal atrophy, temporo-parietal hypometabolism on [18F]-fluorodeoxyglucose-positron emission tomography (FDG-PET), and abnormal CSF tau). We aim to compare the accuracy of these biomarkers, individually and in combination, in predicting AD among mild cognitive impairment (MCI) patients. METHODS: In 73 MCI patients, followed to ascertain AD progression, markers were measured. Sensitivity and specificity, positive (LR+) and negative (LR-) likelihood ratios, and crude and adjusted hazard ratios were computed. RESULTS: Twenty-nine MCI patients progressed and 44 remained stable. Positivity to any marker achieved the lowest LR- (0.0), whereas the combination Aß42 plus FDG-PET achieved the highest LR+ (6.45). In a survival analysis, positivity to any marker was associated with 100% conversion rate, whereas negativity to all markers was associated with 100% stability. DISCUSSION: The best criteria combined amyloidosis and neurodegeneration biomarkers, whereas the individual biomarker with the best performance was FDG-PET.

Alzheimer Disease Biomarkers as Outcome Measures for Clinical Trials in MCI.

BACKGROUND: The aim of this study was to compare the performance and power of the best-established diagnostic biological markers as outcome measures for clinical trials in patients with mild cognitive impairment (MCI). METHODS: Magnetic resonance imaging, F-18 fluorodeoxyglucose positron emission tomography markers, and Alzheimer's Disease Assessment Scale-cognitive subscale were compared in terms of effect size and statistical power over different follow-up periods in 2 MCI groups, selected from Alzheimer's Disease Neuroimaging Initiative data set based on cerebrospinal fluid (abnormal cerebrospinal fluid Aß1-42 concentration-ABETA+) or magnetic resonance imaging evidence of Alzheimer disease (positivity to hippocampal atrophy-HIPPO+). Biomarkers progression was modeled through mixed effect models. Scaled slope was chosen as measure of effect size. Biomarkers power was estimated using simulation algorithms. RESULTS: Seventy-four ABETA+ and 51 HIPPO+ MCI patients were included in the study. Imaging biomarkers of neurodegeneration, especially MR measurements, showed highest performance. For all biomarkers and both MCI groups, power increased with increasing follow-up time, irrespective of biomarker assessment frequency. CONCLUSION: These findings provide information about biomarker enrichment and outcome measurements that could be employed to reduce MCI patient samples and treatment duration in future clinical trials.

A Predictive Study of Dirichlet Process Mixture Models for Curve Fitting.

This paper examines the use of Dirichlet process (DP) mixtures for curve fitting. An important modelling aspect in this setting is the choice between constant or covariate-dependent weights. By examining the problem of curve fitting from a predictive perspective, we show the advantages of using covariate-dependent weights. These advantages are a result of the incorporation of covariate proximity in the latent partition. However, closer examination of the partition yields further complications, which arise from the vast number of total partitions. To overcome this, we propose to modify the probability law of the random partition to strictly enforce the notion of covariate proximity, while still maintaining certain properties of the DP. This allows the distribution of the partition to depend on the covariate in a simple manner and greatly reduces the total number of possible partitions, resulting in improved curve fitting and faster computations. Numerical illustrations are presented.