Pre-hospital glycemia as a biomarker for in-hospital all-cause mortality in diabetic patients - a pilot study.

BACKGROUND: Type 2 Diabetes Mellitus (T2DM) presents a significant healthcare challenge, with considerable economic ramifications. While blood glucose management and long-term metabolic target setting for home care and outpatient treatment follow established procedures, the approach for short-term targets during hospitalization varies due to a lack of clinical consensus. Our study aims to elucidate the impact of pre-hospitalization and intra-hospitalization glycemic indexes on in-hospital survival rates in individuals with T2DM, addressing this notable gap in the current literature. METHODS: In this pilot study involving 120 hospitalized diabetic patients, we used advanced machine learning and classical statistical methods to identify variables for predicting hospitalization outcomes. We first developed a 30-day mortality risk classifier leveraging AdaBoost-FAS, a state-of-the-art ensemble machine learning method for tabular data. We then analyzed the feature relevance to identify the key predictive variables among the glycemic and routine clinical variables the model bases its predictions on. Next, we conducted detailed statistical analyses to shed light on the relationship between such variables and mortality risk. Finally, based on such analyses, we introduced a novel index, the ratio of intra-hospital glycemic variability to pre-hospitalization glycemic mean, to better characterize and stratify the diabetic population. RESULTS: Our findings underscore the importance of personalized approaches to glycemic management during hospitalization. The introduced index, alongside advanced predictive modeling, provides valuable insights for optimizing patient care. In particular, together with in-hospital glycemic variability, it is able to discriminate between patients with higher and lower mortality rates, highlighting the importance of tightly controlling not only pre-hospital but also in-hospital glycemic levels. CONCLUSIONS: Despite the pilot nature and modest sample size, this study marks the beginning of exploration into personalized glycemic control for hospitalized patients with T2DM. Pre-hospital blood glucose levels and related variables derived from it can serve as biomarkers for all-cause mortality during hospitalization.

Rapid Assessment of COVID-19 Mortality Risk with GASS Classifiers.

Risk prediction models are fundamental to effectively triage incoming COVID-19 patients. However, current triaging methods often have poor predictive performance, are based on variables that are expensive to measure, and often lead to hard-to-interpret decisions. We introduce two new classification methods that can predict COVID-19 mortality risk from the automatic analysis of routine clinical variables with high accuracy and interpretability. SVM22-GASS and Clinical-GASS classifiers leverage machine learning methods and clinical expertise, respectively. Both were developed using a derivation cohort of 499 patients from the first wave of the pandemic and were validated with an independent validation cohort of 250 patients from the second pandemic phase. The Clinical-GASS classifier is a threshold-based classifier that leverages the General Assessment of SARS-CoV-2 Severity (GASS) score, a COVID-19-specific clinical score that recently showed its effectiveness in predicting the COVID-19 mortality risk. The SVM22-GASS model is a binary classifier that non-linearly processes clinical data using a Support Vector Machine (SVM). In this study, we show that SMV22-GASS was able to predict the mortality risk of the validation cohort with an AUC of 0.87 and an accuracy of 0.88, better than most scores previously developed. Similarly, the Clinical-GASS classifier predicted the mortality risk of the validation cohort with an AUC of 0.77 and an accuracy of 0.78, on par with other established and emerging machine-learning-based methods. Our results demonstrate the feasibility of accurate COVID-19 mortality risk prediction using only routine clinical variables, readily collected in the early stages of hospital admission.

Toward a unifying framework for the modeling and identification of motor primitives.

A large body of evidence suggests that human and animal movements, despite their apparent complexity and flexibility, are remarkably structured. Quantitative analyses of various classes of motor behaviors consistently identify spatial and temporal features that are invariant across movements. Such invariant features have been observed at different levels of organization in the motor system, including the electromyographic, kinematic, and kinetic levels, and are thought to reflect fixed modules-named motor primitives-that the brain uses to simplify the construction of movement. However, motor primitives across space, time, and organization levels are often described with ad-hoc mathematical models that tend to be domain-specific. This, in turn, generates the need to use model-specific algorithms for the identification of both the motor primitives and additional model parameters. The lack of a comprehensive framework complicates the comparison and interpretation of the results obtained across different domains and studies. In this work, we take the first steps toward addressing these issues, by introducing a unifying framework for the modeling and identification of qualitatively different classes of motor primitives. Specifically, we show that a single model, the anechoic mixture model, subsumes many popular classes of motor primitive models. Moreover, we exploit the flexibility of the anechoic mixture model to develop a new class of identification algorithms based on the Fourier-based Anechoic Demixing Algorithm (FADA). We validate our framework by identifying eight qualitatively different classes of motor primitives from both simulated and experimental data. We show that, compared to established model-specific algorithms for the identification of motor primitives, our flexible framework reaches overall comparable and sometimes superior reconstruction performance. The identification framework is publicly released as a MATLAB toolbox (FADA-T, https://tinyurl.com/compsens) to facilitate the identification and comparison of different motor primitive models.

A Dynamical Generative Model of Social Interactions.

The ability to make accurate social inferences makes humans able to navigate and act in their social environment effortlessly. Converging evidence shows that motion is one of the most informative cues in shaping the perception of social interactions. However, the scarcity of parameterized generative models for the generation of highly-controlled stimuli has slowed down both the identification of the most critical motion features and the understanding of the computational mechanisms underlying their extraction and processing from rich visual inputs. In this work, we introduce a novel generative model for the automatic generation of an arbitrarily large number of videos of socially interacting agents for comprehensive studies of social perception. The proposed framework, validated with three psychophysical experiments, allows generating as many as 15 distinct interaction classes. The model builds on classical dynamical system models of biological navigation and is able to generate visual stimuli that are parametrically controlled and representative of a heterogeneous set of social interaction classes. The proposed method represents thus an important tool for experiments aimed at unveiling the computational mechanisms mediating the perception of social interactions. The ability to generate highly-controlled stimuli makes the model valuable not only to conduct behavioral and neuroimaging studies, but also to develop and validate neural models of social inference, and machine vision systems for the automatic recognition of social interactions. In fact, contrasting human and model responses to a heterogeneous set of highly-controlled stimuli can help to identify critical computational steps in the processing of social interaction stimuli.