To develop and evaluate machine learning (ML) models that infer preoperative cognitive function from intraoperative electroencephalography (EEG). This was a retrospective ML study that used a training dataset derived from the MINDDS study (306 patients, USA), and an external testing dataset from the Electroencephalographic Biomarker to Predict Acute Post-Operatory Cognitive Dysfunction study (92 patients, Chile). Both contained patients older than 60 years undergoing either cardiac (training dataset) or non-cardiac (testing dataset) surgery under general anesthesia. Preoperative cognitive function was assessed using the Montreal Cognitive Assessment (MoCA) in both cohorts. Four types of ML models were used: logistic regression with L2 penalty, random forest, gradient boosting tree, and extreme gradient boosting. Models were evaluated in terms of weighted root mean square error (WRMSE) and monotonic correlations towards actual MoCA scores (Spearman's rho). A logistic regression model with L2 regularization performed best in the training dataset (WRMSE 2.82 [2.60 - 3.03 95%CI], Spearman's rho 0.18 [0.06 - 0.29], p 0.0015). This performance mostly generalized to the test dataset (WRMSE 2.72 [2.51 - 2.94], Spearman's rho 0.14 [-0.05 - 0.31], p 0.18). This study shows that ML models trained on intraoperative EEG can effectively infer preoperative cognitive function in older patients, with generalizability across distinct populations and relatively low error (<3 MoCA points). However, the correlations were weak, indicating limited ability to capture consistent monotonic relationships. Incorporating this approach into perioperative care could enable early detection and mitigation of neurocognitive disorders, improving surgical outcomes through tailored interventions. Further refinement and validation are required before clinical implementation.

Introduction: Postoperative delirium is common and associated with poor postoperative outcomes. However, the predictive power of intraoperative electroencephalogram (EEG) features for postoperative delirium has not yet been well studied. Methods: Intraoperative EEG data from 261 patients who underwent major cardiac surgery were analyzed. Cases were identified using the Confusion Assessment Method. Predictive analytics for delirium outcome were performed using (1) only clinical data, (2) only EEG data, and (3) a combined list of important features from the first two stages. Results: Eleven percentage of participants experienced postoperative delirium. The patients were generally older and had lower physical and cognitive function. EEG models were found to be highly specific but less sensitive in identifying delirium cases. The combined EEG-clinical model performed comparably to the clinical-only model (AUC = 80%) but outperformed the EEG-only model (AUC = 56%). After adjusting for clinical covariates, only interhemispheric mutual information remained significantly associated with delirium (OR = 2.29, p = 0.03), with a positive correlation with delirium severity (ρ = 0.18, P ≤ 0.01). Conclusions: This study enhances our understanding of delirium neurophysiology by emphasizing the role of intraoperative EEG as a marker of brain vulnerability. Although EEG may not constitute a standalone biomarker of delirium, it holds promise for delirium risk stratification.

BACKGROUND:The developing neonatal brain displays different electroencephalographic (EEG) responses to GABAergic anaesthetics than adults. Evidence suggests the importance of isoelectric-like activity patterns. However, markers of hypnotic depth are currently lacking for this population. OBJECTIVE:To explore potential EEG markers of propofol-induced hypnosis in sedated critically ill term neonates. DESIGN:Observational exploratory cohort study. PATIENTS:Twenty critically ill term neonates (postmenstrual age 37 to 44 weeks) undergoing intensive care and requiring anaesthesia for noncardiac surgery. Patients with perinatal asphyxia, neurological pathology, brain malformations and metabolic or haemodynamic instability were excluded. INTERVENTIONS:Frontal EEG (Sedline) was recorded before induction and during a 20-min continuous rate propofol infusion. MAIN OUTCOME MEASURES:Depending on peak amplitude, segmented EEG signals (1 s epochs) were classified as either isoelectric (<10 μV), low-voltage 10 to 25 μV), or high-voltage (>25 μV). Propofol effects were evaluated in terms of time occupancy and spectral properties within these EEG states. Correlations between clinical variables and EEG states were explored. RESULTS:The EEGs of 17 neonates were analysed. Most showed periods of low-voltage (16/17, 94%) and isoelectric states (2/17, 70.5%) before anaesthesia. The time spent in these EEG states increased significantly during propofol infusion; 17/17 (100%), P < 0.001 and 16/17 (94.1%), P = 0.016, respectively. Propofol increased the mean [95% confidence interval (CI)] time spent in the isoelectric state per patient: 12.4 (3.3 to 21.5)% versus 28.6 (14.4 to 42.8)%, P < 0.002. A reduced spectral power was observed across all frequency bands during low-voltage states (all P < 0.026). Gestational age was negatively correlated with time in the isoelectric state; rho, 95% CI, -0.539 (-0.11 to -0.87), P = 0.031. CONCLUSION:Our results show that isoelectric periods are common before anaesthesia in our studied population and more frequent in patients born at earlier gestational ages. The data suggest that propofol anaesthesia increases isoelectric EEG states while also reducing the spectral power, specifically during low-voltage EEG states. Potentially, both of these EEG changes could be biomarkers of neonatal hypnosis depth in this particular critically ill subpopulation. TRIAL REGISTRATION:ClinicalTrials.gov identifier: NCT04904965.

Background Electroencephalogram oscillations during general anesthesia may change as a function of cognitive and physical health. This study aimed to characterize associations between anesthesia-induced oscillations and postoperative outcomes in cardiac surgery patients over 60 years. Methods This was a prespecified secondary data analysis from the Minimizing Intensive Care Unit Dysfunction with Dexmedetomidine-induced Sleep (MINDDS) study. Participants were admitted from home for elective cardiac surgery with cardiopulmonary bypass. The primary outcome was postoperative delirium obtained using the Confusion Assessment Method. Secondary outcomes were non-home discharge and 30-day readmission. The exposure of interest was alpha power measured during the maintenance phase of isoflurane-general anesthesia. Confounding cognitive and physical health variables were collected. Results Of 394 participants in the MINDDS study, 302 had analyzable electroencephalograms. The incidence of postoperative delirium was 11.1 %. Odds of postoperative delirium decreased by 14 % for every decibel increase in alpha power (OR 0.86, 95 % CI: 0.78 to 0.95; P = 0.004). This finding was not significant in adjusted analysis (ORadj 0.92, 95 % CI: 0.81 to 1.03; P = 0.154). Non-home discharge setting findings were not associated with alpha power. The odds of 30-day readmission decreased by 20 % for every decibel increase in alpha power (ORadj 0.80, 95 % CI: 0.71 to 0.91; P < 0.001). Findings were conserved in exploratory and sensitivity analyses. Conclusions In this study anesthesia-induced oscillations were associated with postoperative outcomes; however, these were not independently associated with delirium or discharge disposition after considering preoperative cognitive and physical health. These oscillations were robustly associated with 30-day readmission however, which may help anesthesiologists identify high-risk patients, offering benefits beyond the operating room.Clinical trial registration: Registration Number: NCT02856594

BACKGROUND:Frail elderly patients have a higher risk of postoperative morbidity and mortality. Prehabilitation is a potential intervention for optimizing postoperative outcomes in frail patients. We studied the impact of a prehabilitation program on length of stay (LOS) in frail elderly patients undergoing elective surgery. METHODS:An RCT study was conducted. Frail patients scheduled for elective surgery were randomized to receive either pre-surgical conditioning protocol (PCP) or standard preoperative care. PCP included nursing, anesthetic, and geriatric assessment, nutritional intervention, and physical training for 4-weeks preoperatively. A nurse followed both groups until discharge criteria were met. The primary outcome was postoperative LOS. Secondary outcomes were nutritional status, preoperative frailty status (frailty phenotype-FP) after PCP, and postoperative complications up to three months categorized according to the Clavien-Dindo Classification. Means and medians between the control and intervention groups were compared, with statistical significance set at α=5%. RESULTS:Thirty-four patients were to intervention and Thirty-seven to the control group. In the intervention group, adherence to prehabilitation was 90%. The median LOS after surgery was three days in both groups, without finding statistically significant differences between groups (P=0.754), although there was a trend towards lower LOS in the urologic surgery subgroup. We found a significant reduction in frailty status after PCP (FPpre=2.4±0.5 and FPpost=1.7±0.5, P<0.001). Nutritional status significantly improved in frail patients after prehabilitation (MNAbasal=9.0±2.5 and MNApost=10.6±2.6), P=0.028. The intervention group had less severe postoperative complications, which were not statistically significant. CONCLUSIONS:The PCP conducted both in-person and online, for older frail patients undergoing elective colorectal and urological surgery was not associated with shorter LOS. However, frailty status significantly improved after completing PCP.

BACKGROUND:The Nociception Level Index has shown benefits in estimating the nociception/antinociception balance in adults, but there is limited evidence in the pediatric population. Evaluating the index performance in children might provide valuable insights to guide opioid administration. AIMS:To evaluate the Nociception Level Index ability to identify a standardized nociceptive stimulus and the analgesic effect of a fentanyl bolus. Additionally, to characterize the pharmacokinetic/pharmacodynamic relationship of fentanyl with the Nociception Level Index response during sevoflurane anesthesia. METHODS:Nineteen children, 5.3 (4.1-6.7) years, scheduled for lower abdominal or urological surgery, were studied. After sevoflurane anesthesia and caudal block, a tetanic stimulus (50 Hz, 60 mA, 5 s) was performed in the forearm. Following the administration of fentanyl 2 μg/kg intravenous bolus, three similar consecutive tetanic stimuli were performed at 5-, 15-, and 30-min post-fentanyl administration. Changes in the Nociception Level Index, heart rate, mean arterial pressure, and bispectral index were compared in response to the tetanic stimuli. Fentanyl plasma concentrations and the Nociception Level Index data were used to elaborate a pharmacokinetic/pharmacodynamic model using a sequential modeling approach in NONMEM®. RESULTS:After the first tetanic stimulus, both the Nociception Level Index and the heart rate increased compared to baseline (8 ± 7 vs. 19 ± 10; mean difference (CI95) -12(-18--6) and 100 ± 10 vs. 102 ± 10; -2(-4--0.1)) and decrease following fentanyl administration (19 ± 10 vs. 8 ± 8; 12 (5-18) and 102 ± 10 vs. 91 ± 11; 11 (7-16)). In subsequent tetanic stimuli, heart rate remained unchanged, while the Nociception Level Index progressively increased within 15 min to values similar to those before fentanyl. An allometric weight-scaled, 3-compartment model best characterized the pharmacokinetic profile of fentanyl. The pharmacokinetic/pharmacodynamic modeling analysis revealed hysteresis between fentanyl plasma concentrations and the Nociception Level Index response, characterized by plasma effect-site equilibration half-time of 1.69 (0.4-2.9) min. The estimated fentanyl C50 was 1.93 (0.73-4.2) ng/mL. CONCLUSION:The Nociception Level Index showed superior capability compared to traditional hemodynamic variables in discriminating different nociception-antinociception levels during varying fentanyl concentrations in children under sevoflurane anesthesia.

Cities in the global south face dire climate impacts. It is in socioeconomically marginalized urban communities of the global south that the effects of climate change are felt most deeply. Santiago de Chile, a major mid-latitude Andean city of 7.7 million inhabitants, is already undergoing the so-called “climate penalty” as rising temperatures worsen the effects of endemic ground-level ozone pollution. As many cities in the global south, Santiago is highly segregated along socioeconomic lines, which offers an opportunity for studying the effects of concurrent heatwaves and ozone episodes on distinct zones of affluence and deprivation. Here, we combine existing datasets of social indicators and climate-sensitive health risks with weather and air quality observations to study the response to compound heat-ozone extremes of different socioeconomic strata. Attributable to spatial variations in the ground-level ozone burden (heavier for wealthy communities), we found that the mortality response to extreme heat (and the associated further ozone pollution) is stronger in affluent dwellers, regardless of comorbidities and lack of access to health care affecting disadvantaged population. These unexpected findings underline the need of a site-specific hazard assessment and a community-based risk management.

IntroductionThe CALL score is a predictive tool for respiratory failure progression in COVID-19. Whether the CALL score is useful to predict short- and medium-term mortality in an unvaccinated population is unknown.Materials and methodsThis is a prospective cohort study in unvaccinated inpatients with a COVID-19 pneumonia diagnosis upon hospital admission. Patients were followed up for mortality at 28 days, 3, 6, and 12 months. Associations between CALL score and mortality were analyzed using logistic regression. The prediction performance was evaluated using the area under a receiver operating characteristic curve (AUROC).ResultsA total of 592 patients were included. On average, the CALL score was 9.25 (±2). Higher CALL scores were associated with increased mortality at 28 days [univariate: odds ratio (OR) 1.58 (95% CI, 1.34–1.88), p < 0.001; multivariate: OR 1.54 (95% CI, 1.26–1.87), p < 0.001] and 12 months [univariate OR 1.63 (95% CI, 1.38–1.93), p < 0.001; multivariate OR 1.63 (95% CI, 1.35–1.97), p < 0.001]. The prediction performance was good for both univariate [AUROC 0.739 (0.687–0.791) at 28 days and 0.869 (0.828–0.91) at 12 months] and multivariate models [AUROC 0.752 (0.704–0.8) at 28 days and 0.862 (0.82–0.905) at 12 months].ConclusionThe CALL score exhibits a good predictive capacity for short- and medium-term mortality in an unvaccinated population.

Cognitive decline is common among older individuals, and although the underlying brain mechanisms are not entirely understood, researchers have suggested using EEG frontal alpha activity during general anaesthesia as a potential biomarker for cognitive decline. This is because frontal alpha activity associated with GABAergic general anaesthetics has been linked to cognitive function. However, oscillatory-specific alpha power has also been linked with chronological age. We hypothesize that cognitive function mediates the association between chronological age and (oscillatory-specific) alpha power. We analysed data from 380 participants (aged over 60) with baseline screening assessments and intraoperative EEG. We utilized the telephonic Montreal Cognitive Assessment to assess cognitive function. We computed total band power, oscillatory-specific alpha power, and aperiodics to measure anaesthesia-induced alpha activity. To test our mediation hypotheses, we employed structural equation modelling. Pairwise correlations between age, cognitive function and alpha activity were significant. Cognitive function mediated the association between age and classical alpha power [age -> cognitive function -> classical alpha; beta = -0.0168 (95% confidence interval: -0.0313 to -0.00521); P = 0.0016] as well as the association between age and oscillatory-specific alpha power [age -> cognitive function -> oscillatory-specific alpha power; beta = -0.00711 (95% confidence interval: -0.0154 to -0.000842); P = 0.028]. However, cognitive function did not mediate the association between age and aperiodic activity (1/f slope, P = 0.43; offset, P = 0.0996). This study is expected to provide valuable insights for anaesthesiologists, enabling them to make informed inferences about a patient's age and cognitive function from an analysis of anaesthetic-induced EEG signals in the operating room. To ensure generalizability, further studies across different populations are needed. Interpreting low intraoperative EEG alpha activity can be challenging due to its associations with both age and cognitive function. However, Boncompte et al. have skilfully addressed this issue by showing that cognitive function plays a significant role in mediating the relationship between age and the oscillatory component of alpha activity. Graphical Abstract

• Frailty was common in elderly patients undergoing non cardiac surgery . • Electroencephalogram alpha-band power does not predict preoperative frailty above patients' age. • Frailty predictions by machine learning algorithms, were not improved by the addition of electroencephalogram features. • Frailty might be different to the concept of brain vulnerability, accounting for a possible brain-body dissociation.

Background Improving anesthesia administration for elderly population is of particular importance because they undergo considerably more surgical procedures and are at the most risk of suffering from anesthesia-related complications. Intraoperative brain monitors electroencephalogram (EEG) have proved useful in the general population, however, in elderly subjects this is contentious. Probably because these monitors do not account for the natural differences in EEG signals between young and older patients. In this study we attempted to systematically characterize the age-dependence of different EEG measures of anesthesia hypnosis. Methods We recorded EEG from 30 patients with a wide age range (19–99 years old) and analyzed four different proposed indexes of depth of hypnosis before, during and after loss of behavioral response due to slow propofol infusion during anesthetic induction. We analyzed Bispectral Index (BIS), Alpha Power and two entropy-related EEG measures, Lempel-Ziv complexity (LZc), and permutation entropy (PE) using mixed-effect analysis of variances (ANOVAs). We evaluated their possible age biases and their trajectories during propofol induction. Results All measures were dependent on anesthesia stages. BIS, LZc, and PE presented lower values at increasing anesthetic dosage. Inversely, Alpha Power increased with increasing propofol at low doses, however this relation was reversed at greater effect-site propofol concentrations. Significant group differences between elderly patients (>65 years) and young patients were observed for BIS, Alpha Power, and LZc, but not for PE. Conclusion BIS, Alpha Power, and LZc show important age-related biases during slow propofol induction. These should be considered when interpreting and designing EEG monitors for clinical settings. Interestingly, PE did not present significant age differences, which makes it a promising candidate as an age-independent measure of hypnotic depth to be used in future monitor development.

The emergence of artificial intelligence and machine learning in medicine determines that healthcare professionals should understand generalities of their methodologies.This narrative review consists of two parts.The first consists of an exploration of the main methods used to model in machine learning, described in a simple way by medical and mathematical authors, with the purpose to bring this methodology healthcare workers.Here we will describe the basic structure of a machine learning algorithm (input information, task to execute, output result, optimization, and adjustment), its main classifications (supervised, unsupervised and by reinforcement) and the main modeling methods used.We will review regression and then explore decision trees, support vector machines, principal component analysis, clustering, K-means, hierarchical clustering, deep learning, and convolutional neural networks.In this way, we hope to bring this methodology closer to healthcare personnel to increase the interpretability of the published work in medicine that use these methodologies.