Learning clinical networks from medical records based on information estimates in mixed-type data

Abstract

The precise diagnostics of complex diseases require to integrate a large amount of information from heterogeneous clinical and biomedical data, whose direct and indirect interdependences are notoriously difficult to assess. To this end, we propose an efficient computational approach to simultaneously compute and assess the significance of multivariate information between any combination of mixed-type (continuous/categorical) variables. The method is then used to uncover direct, indirect and possibly causal relationships between mixed-type data from medical records, by extending a recent machine learning method to reconstruct graphical models beyond simple categorical datasets. The method is shown to outperform existing tools on benchmark mixed-type datasets, before being applied to analyze the medical records of eldery patients with cognitive disorders from La Pitié-Salpêtrière Hospital, Paris. The resulting clinical network visually captures the global interdependences in these medical records and some facets of clinical diagnosis practice, without specific hypothesis nor prior knowledge on any clinically relevant information. In particular, it provides some physiological insights linking the consequence of cerebrovascular accidents to the atrophy of important brain structures associated to cognitive impairment.

Use of Machine Learning and Artificial Intelligence Methods in Geriatric Mental Health Research Involving Electronic Health Record or Administrative Claims Data: A Systematic Review

“…Support vector machine ( 15 , 31 ) and topic modeling ( 19 , 24 ) were used by two studies each. Finally, lasso ( 33 ), naïve Bayes ( 28 ), multivariate information-based inductive causation (MIIC) network learning algorithm ( 26 ), fuzzy c-means cluster analysis ( 23 ), conditional restricted Boltzmann machine ( 27 ), WEKA ( 15 ), and gradient boosted trees ( 33 ) were applied by one study each.…”

“…Major neurocognitive disorder, dementia, and Alzheimer's disease were the most common conditions among the articles included in our study and were reported in 11 studies ( 7 , 15 , 16 , 24 , 25 , 27 , 28 , 30 , 31 , 33 , 34 ). Among the other outcomes, a geriatric syndrome (falls, malnutrition, dementia, severe urinary control issues, absence of fecal control, visual impairment, walking difficulty, pressure ulcers, lack of social support, and weight loss) in 3 studies ( 18 , 21 , 22 ), delirium in 2 studies ( 20 , 32 ), mild cognitive impairment ( 33 ), cognitive disorder ( 26 ), multimorbidity pattern ( 23 ), mortality ( 29 ), hospital admission ( 17 ), and themes/topics mentioned in care providers' notes ( 19 ) were considered once. Outcomes were primarily defined using ICD diagnosis codes in AHD and EHR.…”

“…After removing duplicates, 363 titles and abstracts were screened for eligibility, and from there, 68 articles were selected for fulltext screening. After assessing full-texts, 21 articles were finally selected for the systematic review (7,(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34). The detailed study selection process is summarized in Figure 1.…”

Inferring Gene Networks in Bone Marrow Hematopoietic Stem Cell-Supporting Stromal Niche Populations

“…As analyses, such as WGCNA, are optimally suited to identify correlated networks, but do not identify direct gene-to-gene interactions or potential causal links, we used miic analysis (Sella et al, 2018;Verny et al, 2017), recently extended to analyze continuous or mixed-type (e.g., continuous-categorical) data (Cabeli et al, 2020). The miic algorithm identifies direct paths with high confidence, and, as such, reveals potential cause-and-effect links as well as latent regulators, which may not show up when looking only at gene expression data.…”

Estimating Conditional Mutual Information for Discrete-Continuous Mixtures using Multi-Dimensional Adaptive Histograms

“…While estimating (conditional) mutual information for purely discrete or continuous data is a well-studied problem [4,5,8,10,22], many real-world settings concern a mix of discrete and continuous random variables, such as age (in years) and height, or even random variables that can individually consist of a mixture of discrete and continuous components. Although several discretization-based approaches that can estimate MI for a mix of discrete and continuous random variables have recently emerged [2,17,31], so far only methods based on k-nearest neighbour (kNN) estimation were shown to work on mixed variables, which may consist of discrete-continuous mixture variables [7,19,24].…”