Early prediction of in-hospital mortality utilizing multivariate predictive modelling of electronic medical records and socio-determinants of health of the first day of hospitalization.

BACKGROUND: In France an average of 4% of hospitalized patients die during their hospital stay. To aid medical decision making and the attribution of resources, within a few days of admission the identification of patients at high risk of dying in hospital is essential. METHODS: We used de-identified routine patient data available in the first 2 days of hospitalization in a French University Hospital (between 2016 and 2018) to build models predicting in-hospital mortality (at ≥ 2 and ≤ 30 days after admission). We tested nine different machine learning algorithms with repeated 10-fold cross-validation. Models were trained with 283 variables including age, sex, socio-determinants of health, laboratory test results, procedures (Classification of Medical Acts), medications (Anatomical Therapeutic Chemical code), hospital department/unit and home address (urban, rural etc.). The models were evaluated using various performance metrics. The dataset contained 123,729 admissions, of which the outcome for 3542 was all-cause in-hospital mortality and 120,187 admissions (no death reported within 30 days) were controls. RESULTS: The support vector machine, logistic regression and Xgboost algorithms demonstrated high discrimination with a balanced accuracy of 0.81 (95%CI 0.80-0.82), 0.82 (95%CI 0.80-0.83) and 0.83 (95%CI 0.80-0.83) and AUC of 0.90 (95%CI 0.88-0.91), 0.90 (95%CI 0.89-0.91) and 0.90 (95%CI 0.89-0.91) respectively. The most predictive variables for in-hospital mortality in all three models were older age (greater risk), and admission with a confirmed appointment (reduced risk). CONCLUSION: We propose three highly discriminating machine-learning models that could improve clinical and organizational decision making for adult patients at hospital admission.

Dissociating the Influence of Affective Word Content and Cognitive Processing Demands on the Late Positive Potential.

The late positive potential (LPP) elicited by affective stimuli in the event-related brain potential (ERP) is often assumed to be a member of the P3 family. The present study addresses the relationship of the LPP to the classic P3b in a published data set, using a non-parametric permutation test for topographical comparisons, and residue iteration decomposition to assess the temporal features of the LPP and the P3b by decomposing the ERP into several component clusters according to their latency variability. The experiment orthogonally manipulated arousal and valence of words, which were either read or judged for lexicality. High-arousing and positive valenced words induced a larger LPP than low-arousing and negative valenced words, respectively, and the LDT elicited a larger P3b than reading. The experimental manipulation of arousal, valence, and task yielded main effects without any interactions on ERP amplitude in the LPP/P3b time range. The arousal and valence effects partially differed from the task effect in scalp topography; in addition, whereas the late positive component elicited by affective stimuli, defined as LPP, was stimulus-locked, the late positive component elicited by task demand, defined as P3b, was mainly latency-variable. Therefore LPP and P3b manifest different subcomponents.

Cohort profile for development of machine learning models to predict healthcare-related adverse events (Demeter): clinical objectives, data requirements for modelling and overview of data set for 2016-2018.

PURPOSE: In-hospital health-related adverse events (HAEs) are a major concern for hospitals worldwide. In high-income countries, approximately 1 in 10 patients experience HAEs associated with their hospital stay. Estimating the risk of an HAE at the individual patient level as accurately as possible is one of the first steps towards improving patient outcomes. Risk assessment can enable healthcare providers to target resources to patients in greatest need through adaptations in processes and procedures. Electronic health data facilitates the application of machine-learning methods for risk analysis. We aim, first to reveal correlations between HAE occurrence and patients' characteristics and/or the procedures they undergo during their hospitalisation, and second, to build models that allow the early identification of patients at an elevated risk of HAE. PARTICIPANTS: 143 865 adult patients hospitalised at Grenoble Alpes University Hospital (France) between 1 January 2016 and 31 December 2018. FINDINGS TO DATE: In this set-up phase of the project, we describe the preconditions for big data analysis using machine-learning methods. We present an overview of the retrospective de-identified multisource data for a 2-year period extracted from the hospital's Clinical Data Warehouse, along with social determinants of health data from the National Institute of Statistics and Economic Studies, to be used in machine learning (artificial intelligence) training and validation. No supplementary information or evaluation on the part of medical staff will be required by the information system for risk assessment. FUTURE PLANS: We are using this data set to develop predictive models for several general HAEs including secondary intensive care admission, prolonged hospital stay, 7-day and 30-day re-hospitalisation, nosocomial bacterial infection, hospital-acquired venous thromboembolism, and in-hospital mortality.

Mechanisms of face specificity - Differentiating speed and accuracy in face cognition by event-related potentials of central processing.

Given the crucial role of face recognition in social life, it is hardly surprising that cognitive processes specific for faces have been identified. In previous individual differences studies, the speed (measured in easy tasks) and accuracy (difficult tasks) of face cognition (FC, involving perception and recognition of faces) have been shown to form distinct abilities, going along with divergent factorial structures. This result has been replicated, but remained unexplained. To fill this gap, we first parameterized the sub-processes underlying speed vs. accuracy in easy and difficult memory tasks for faces and houses in a large sample. Then, we analyzed event-related potentials (ERPs) extracted from the EEG by using residue iteration decomposition (RIDE), yielding a central (C) component that is comparable to a purified P300. Structural equation modeling (SEM) was applied to estimate face specificity of C component latencies and amplitudes. If performance in easy tasks relies on purely general processes that are insensitive to stimulus content, there should be no specificity of individual differences in the latency recorded in easy tasks. However, in difficult tasks specificity was expected. Results indicated that, contrary to our predictions, specificity occurred in the C component latency of both speed-based and accuracy-based measures, but was stronger in accuracy. Further analyses suggested specific relationships between the face-related C latency and FC ability. Finally, we detected specificity in RTs of easy tasks when single tasks were modeled, but not when multiple tasks were jointly modeled. This suggests that the mechanisms leading to face specificity in performance speed are distinct across tasks.

Sex-specific relationships between face memory and the N170 component in event-related potentials.

At the group level, women consistently perform better in face memory tasks than men and also show earlier and larger N170 components of event-related brain potentials (ERP), considered to indicate perceptual structural encoding of faces. Here we investigated sex differences in the relationship between the N170 and face memory performance in 152 men and 141 women at group mean and individual differences levels. ERPs and performance were measured in separate tasks, avoiding statistical dependency between the two. We confirmed previous findings about superior face memory in women and a-sex-independent-negative relationship between N170 latency and face memory. However, whereas in men, better face memory was related to larger N170 components, face memory in women was unrelated with the amplitude or latency of the N170. These data provide solid evidence that individual differences in face memory within men are at least partially related to more intense structural face encoding.

Structural encoding processes contribute to individual differences in face and object cognition: Inferences from psychometric test performance and event-related brain potentials.

The enhanced N1 component in event-related potentials (ERP) to face stimuli, termed N170, is considered to indicate the structural encoding of faces. Previously, individual differences in the latency of the N170 have been related to face and object cognition abilities. By orthogonally manipulating content domain (faces vs objects) and task demands (easy/speed vs difficult/accuracy) in both psychometric and EEG tasks, we investigated the uniqueness of the processes underlying face cognition as compared with object cognition and the extent to which the N1/N170 component can explain individual differences in face and object cognition abilities. Data were recorded from N = 198 healthy young adults. Structural equation modeling (SEM) confirmed that the accuracies of face perception (FP) and memory are specific abilities above general object cognition; in contrast, the speed of face processing was not differentiable from the speed of object cognition. Although there was considerable domain-general variance in the N170 shared with the N1, there was significant face-specific variance in the N170. The brain-behavior relationship showed that faster face-specific processes for structural encoding of faces are associated with higher accuracy in both perceiving and memorizing faces. Moreover, in difficult task conditions, qualitatively different processes are additionally needed for recognizing face and object stimuli as compared with easy tasks. The difficulty-dependent variance components in the N170 amplitude were related with both face and object memory (OM) performance. We discuss implications for understanding individual differences in face cognition.