Triage and monitoring of COVID-19 patients in intensive care using unsupervised machine learning.

BACKGROUND: We designed an algorithm to assess COVID-19 patients severity and dynamic intubation needs and predict their length of stay using the breathing frequency (BF) and oxygen saturation (SpO2) signals. METHODS: We recorded the BF and SpO2 signals for confirmed COVID-19 patients admitted to the ICU of a teaching hospital during both the first and subsequent outbreaks of the pandemic in France. An unsupervised machine-learning algorithm (the Gaussian mixture model) was applied to the patients' data for clustering. The algorithm's robustness was ensured by comparing its results against actual intubation rates. We predicted intubation rates using the algorithm every hour, thus conducting a severity evaluation. We designed a S24 severity score that represented the patient's severity over the previous 24 h; the validity of MS24, the maximum S24 score, was checked against rates of intubation risk and prolonged ICU stay. RESULTS: Our sample included 279 patients. . The unsupervised clustering had an accuracy rate of 87.8% for intubation recognition (AUC = 0.94, True Positive Rate 86.5%, true Negative Rate 90.9%). The S24 score of intubated patients was significantly higher than that of non-intubated patients at 48 h before intubation. The MS24 score allowed for the distinguishing between three severity levels with an increased risk of intubation: green (3.4%), orange (37%), and red (77%). A MS24 score over 40 was highly predictive of an ICU stay greater than 5 days at an accuracy rate of 81.0% (AUC = 0.87). CONCLUSIONS: Our algorithm uses simple signals and seems to efficiently visualize the patients' respiratory situations, meaning that it has the potential to assist staffs' in decision-making. Additionally, real-time computation is easy to implement.

Spine Surgical Procedures during Coronavirus Disease 2019 Pandemic: Is It Still Possible to Take Care of Patients? Results of an Observational Study in the First Month of Confinement.

STUDY DESIGN: Observational study. PURPOSE: The actual sanitary crisis led to a massive mobilization of the sanitary system toward intensive care units and management of coronavirus disease 2019 (COVID-19) patients. However, some patients still require spinal interventions. The present study aimed to assess the impact of the COVID-19 pandemic on spine surgical in a moderate COVID-19 cluster region. OVERVIEW OF LITERATURE: Previous studies have reported screening and management of patients with spinal conditions during the COVID-19 pandemic. However, to date, knowledge, no observational study on spine surgeries during the pandemic has not been reported. METHODS: Between March 17, 2020 and April 17, 2020, information on spine surgical activity was prospectively collected at our institution. This surgical activity related to the first month of confinement in France was compared to the activity during the same period in 2019 to evaluate the impact of the COVID-19 pandemic on surgical activities. RESULTS: In order to reduce the contamination rate of patients and medical staff during hospitalization, the spine department was completely reorganized. Non-urgent elective spine surgeries were cancelled. When considering the global amount of surgeries procedures during the first month of confinement, a decrease of almost 50% was observed in the number of surgical procedures. During the study period, 62 patients were eligible for spine surgery. The numbers of patients managed for tumor and infectious cases were stable, while a considerable reduction was observed in the number of trauma and degenerative cases. During the follow-up period, two patients were tested as COVID+ during the postoperative course, and no cases of medical or paramedical staff contamination were reported using polymerase chain reaction-testing. CONCLUSIONS: During the COVID-19 pandemic, it is possible to maintain spine surgical activity. Each surgical procedure must be discussed and organized with all the caregivers involved. Indications for surgery must be in line with the scientific guidelines and adapted to each healthcare facility.

Multidrug-resistant Pseudomonas aeruginosa and mortality in mechanically ventilated ICU patients.

BACKGROUND: The link between bacterial resistance and prognosis remains controversial. Predominant pathogen causing ventilator-associated pneumonia (VAP) is Pseudomonas aeruginosa (Pa), which has increasingly become multidrug resistant (MDR). The aim of this study was to evaluate the relationship between MDR VAP Pa episodes and 30-day mortality. METHODS: From a longitudinal prospective French multicenter database (2010-2016), Pa VAP onset and physiological data were recorded. MDR was defined as non-susceptibility to at least 1 agent in 3 or more antimicrobial categories. To analyze if MDR episodes were associated with greater in-hospital 30-day mortality, we performed a multivariate survival analysis using the multivariate nonlinear frailty model. RESULTS: A total of 230 patients presented 286 Pa VAP. A maximum of 3 episodes per patient was observed; 73 episodes were MDR and 213 were susceptible. In the multivariate model, factors independently associated with 30-day mortality included hospitalization in the 6 months preceding the first episode (hazard ratio [HR], 2.31; 95% confidence interval [CI], 1.50-3.60; P = .0002), chronic renal failure (HR, 2.34; 95% CI, 1.15-4.77; P = .0196), and Pa VAP recurrence (HR, 2.29; 95% CI, 1.79-4.87; P = .032). Finally, MDR Pa VAP was not associated with death (HR, 0.87; 95% CI; 0.52-1.45; P = .59). CONCLUSIONS: This study did not identify a relationship between the resistance profile of Pseudomonas aeruginosa and mortality.

Age-Related Changes in Segmental Body Composition by Ethnicity and History of Weight Change across the Adult Lifespan.

This study assessed age-related changes in body composition (specifically in trunk fat and appendicular lean masses), with consideration of body mass index (BMI) at age 20 years (BMI reference age, "BMIref"), ethnicity and lifetime weight change history. A cross-sectional dual-energy X-ray absorptiometry-based dataset was extracted from the U.S. National Health and Nutrition Examination Survey (NHANES) 1999-2004. Only European-American and African-American subjects were used (2705 men, 2527 women). For each gender and ethnicity, 6 analytic cases were considered, based on three BMIref categories (normal, overweight and obese, being 22, 27 and 30 kg/m², respectively) and two weight contexts (stable weight or weight gain across the lifespan). A nonparametric model was developed to investigate age-related changes in body composition. Then, parametric modelling was developed for assessing BMIref- and ethnicity-specific effects during aging. In the stable weight, both genders' and ethnicities' trunk fat (TF) increased gradually; body fat (BF) remained stable until 40 years and increased thereafter; trunk lean (TL) remained stable, but appendicular lean (APL) and body lean (BL) declined from 20 years. In the weight gain context, TF and BF increased at a constant rate, while APL, TL and BL increased until 40-50 years, and then declined slightly. Compared with European-American subjects of both genders, African-American subjects had lower TF and BF masses. Ethnic differences in body composition were quantified and found to remain constant across the lifespan.

MetaGenSense: A web-application for analysis and exploration of high throughput sequencing metagenomic data.

The detection and characterization of emerging infectious agents has been a continuing public health concern. High Throughput Sequencing (HTS) or Next-Generation Sequencing (NGS) technologies have proven to be promising approaches for efficient and unbiased detection of pathogens in complex biological samples, providing access to comprehensive analyses. As NGS approaches typically yield millions of putatively representative reads per sample, efficient data management and visualization resources have become mandatory. Most usually, those resources are implemented through a dedicated Laboratory Information Management System (LIMS), solely to provide perspective regarding the available information. We developed an easily deployable web-interface, facilitating management and bioinformatics analysis of metagenomics data-samples. It was engineered to run associated and dedicated Galaxy workflows for the detection and eventually classification of pathogens. The web application allows easy interaction with existing Galaxy metagenomic workflows, facilitates the organization, exploration and aggregation of the most relevant sample-specific sequences among millions of genomic sequences, allowing them to determine their relative abundance, and associate them to the most closely related organism or pathogen. The user-friendly Django-Based interface, associates the users' input data and its metadata through a bio-IT provided set of resources (a Galaxy instance, and both sufficient storage and grid computing power). Galaxy is used to handle and analyze the user's input data from loading, indexing, mapping, assembly and DB-searches. Interaction between our application and Galaxy is ensured by the BioBlend library, which gives API-based access to Galaxy's main features. Metadata about samples, runs, as well as the workflow results are stored in the LIMS. For metagenomic classification and exploration purposes, we show, as a proof of concept, that integration of intuitive exploratory tools, like Krona for representation of taxonomic classification, can be achieved very easily. In the trend of Galaxy, the interface enables the sharing of scientific results to fellow team members.

A multivariate model for predicting segmental body composition.

The aims of the present study were to propose a multivariate model for predicting simultaneously body, trunk and appendicular fat and lean masses from easily measured variables and to compare its predictive capacity with that of the available univariate models that predict body fat percentage (BF%). The dual-energy X-ray absorptiometry (DXA) dataset (52% men and 48% women) with White, Black and Hispanic ethnicities (1999-2004, National Health and Nutrition Examination Survey) was randomly divided into three sub-datasets: a training dataset (TRD), a test dataset (TED); a validation dataset (VAD), comprising 3835, 1917 and 1917 subjects. For each sex, several multivariate prediction models were fitted from the TRD using age, weight, height and possibly waist circumference. The most accurate model was selected from the TED and then applied to the VAD and a French DXA dataset (French DB) (526 men and 529 women) to assess the prediction accuracy in comparison with that of five published univariate models, for which adjusted formulas were re-estimated using the TRD. Waist circumference was found to improve the prediction accuracy, especially in men. For BF%, the standard error of prediction (SEP) values were 3.26 (3.75) % for men and 3.47 (3.95)% for women in the VAD (French DB), as good as those of the adjusted univariate models. Moreover, the SEP values for the prediction of body and appendicular lean masses ranged from 1.39 to 2.75 kg for both the sexes. The prediction accuracy was best for age < 65 years, BMI < 30 kg/m2 and the Hispanic ethnicity. The application of our multivariate model to large populations could be useful to address various public health issues.

Infectivity of GI and GII noroviruses established from oyster related outbreaks.

Noroviruses (NoVs) are the major cause of acute epidemic gastroenteritis in industrialized countries. Outbreak strains are predominantly genogroup II (GII) NoV, but genogroup I (GI) strains are regularly found in oyster related outbreaks. The prototype Norwalk virus (GI), has been shown to have high infectivity in a human challenge study. Whether other NoVs are equally infectious via natural exposure remains to be established. Human susceptibility to NoV is partly determined by the secretor status (Se+/-). Data from five published oyster related outbreaks were analyzed in a Bayesian framework. Infectivity estimates where high and consistent with NV(GI) infectivity, for both GII and GI strains. The median and CI95 probability of infection and illness, in Se+ subjects, associated with exposure to a mean of one single NoV genome copy were around 0.29[0.015-0.61] for GI and 0.4[0.04-0.61] for GII, and for illness 0.13[0.007-0.39] for GI and 0.18[0.017-0.42] for GII. Se- subjects were strongly protected against infection. The high infectivity estimates for Norwalk virus GI and GII, makes NoVs critical target for food safety regulations.

Bayesian development of a dose-response model for Aspergillus fumigatus and invasive aspergillosis.

Invasive aspergillosis (IA) is a major cause of mortality in immunocompromized hosts, most often consecutive to the inhalation of spores of Aspergillus. However, the relationship between Aspergillus concentration in the air and probability of IA is not quantitatively known. In this study, this relationship was examined in a murine model of IA. Immunosuppressed Balb/c mice were exposed for 60 minutes at day 0 to an aerosol of A. fumigatus spores (Af293 strain). At day 10, IA was assessed in mice by quantitative culture of the lungs and galactomannan dosage. Fifteen separate nebulizations with varying spore concentrations were performed. Rates of IA ranged from 0% to 100% according to spore concentrations. The dose-response relationship between probability of infection and spore exposure was approximated using the exponential model and the more flexible beta-Poisson model. Prior distributions of the parameters of the models were proposed then updated with data in a Bayesian framework. Both models yielded close median dose-responses of the posterior distributions for the main parameter of the model, but with different dispersions, either when the exposure dose was the concentration in the nebulized suspension or was the estimated quantity of spores inhaled by a mouse during the experiment. The median quantity of inhaled spores that infected 50% of mice was estimated at 1.8 × 10(4) and 3.2 × 10(4) viable spores in the exponential and beta-Poisson models, respectively. This study provides dose-response parameters for quantitative assessment of the relationship between airborne exposure to the reference A. fumigatus strain and probability of IA in immunocompromized hosts.

A meta-analysis accounting for sources of variability to estimate heat resistance reference parameters of bacteria using hierarchical Bayesian modeling: Estimation of D at 121.1 °C and pH 7, zT and zpH of Geobacillus stearothermophilus.

Predicting microbial survival requires reference parameters for each micro-organism of concern. When data are abundant and publicly available, a meta-analysis is a useful approach for assessment of these parameters, which can be performed with hierarchical Bayesian modeling. Geobacillus stearothermophilus is a major agent of microbial spoilage of canned foods and is therefore a persistent problem in the food industry. The thermal inactivation parameters of G. stearothermophilus (D(ref), i.e.the decimal reduction time D at the reference temperature 121.1°C and pH 7.0, z(T) and z(pH)) were estimated from a large set of 430 D values mainly collected from scientific literature. Between-study variability hypotheses on the inactivation parameters D(ref), z(T) and z(pH) were explored, using three different hierarchical Bayesian models. Parameter estimations were made using Bayesian inference and the models were compared with a graphical and a Bayesian criterion. Results show the necessity to account for random effects associated with between-study variability. Assuming variability on D(ref), z(T) and z(pH), the resulting distributions for D(ref), z(T) and z(pH) led to a mean of 3.3 min for D(ref) (95% Credible Interval CI=[0.8; 9.6]), to a mean of 9.1°C for z(T) (CI=[5.4; 13.1]) and to a mean of 4.3 pH units for z(pH) (CI=[2.9; 6.3]), in the range pH 3 to pH 7.5. Results are also given separating variability and uncertainty in these distributions, as well as adjusted parametric distributions to facilitate further use of these results in aqueous canned foods such as canned vegetables.

Hierarchical Bayesian models to assess between- and within-batch variability of pathogen contamination in food.

Assessing within-batch and between-batch variability is of major interest for risk assessors and risk managers in the context of microbiological contamination of food. For example, the ratio between the within-batch variability and the between-batch variability has a large impact on the results of a sampling plan. Here, we designed hierarchical Bayesian models to represent such variability. Compatible priors were built mathematically to obtain sound model comparisons. A numeric criterion is proposed to assess the contamination structure comparing the ability of the models to replicate grouped data at the batch level using a posterior predictive loss approach. Models were applied to two case studies: contamination by Listeria monocytogenes of pork breast used to produce diced bacon and contamination by the same microorganism on cold smoked salmon at the end of the process. In the first case study, a contamination structure clearly exists and is located at the batch level, that is, between batches variability is relatively strong, whereas in the second a structure also exists but is less marked.

Fat-free mass predictions through a Bayesian Network enable body composition comparisons in various populations.

The respective contribution of fat-free mass (FFM) and fat mass to body weight (Wgt) is a relevant indicator of risk for major public health issues. In an earlier study, a Bayesian Network (BN) was designed to predict FFM from a DXA database (1999-2004 NHANES, n = 10,402) with easily accessible variables [sex, age, Wgt, and height (Hgt)]. The objective of the present study was to assess the robustness of these BN predictions in different population contexts (age, BMI, ethnicity, etc.) when covariables were stochastically deduced from population-based distributions. BN covariables were adjusted to 82 published distributions for age, Wgt, and Hgt from 16 studies assessing body composition. Anthropometric adjustments required a surrogate database (n = 23,411) to get the missing correlation between published Wgt and Hgt distributions. Published BMI distributions and their predicted BN counterparts were correlated (R(2) = 0.99; P < 0.001). Predicted FFM distributions were closely adjusted to their published counterparts for both sexes between 20 and 79 y old, with some discrepancies for Asian populations. In addition, BN predictions revealed a very good agreement between FFM assessed in different population contexts. The mean difference between published FFM values (61.1 ± 3.44 and 42.7 ± 3.32 kg for men and women, respectively) and BN predictions (61.6 ± 3.11 and 42.4 ± 2.76 kg for men and women, respectively) was <1% when FFM was assessed by DXA; the difference rose to 3.6% when FFM was assessed by bioelectric impedance analysis or by densitometry methods. These results suggest that it is possible, within certain anthropometric limitations, to use BN predictions as a complementary body composition analysis for large populations.

A Bayesian evidence synthesis for estimating campylobacteriosis prevalence.

Stakeholders making decisions in public health and world trade need improved estimations of the burden-of-illness of foodborne infectious diseases. In this article, we propose a Bayesian meta-analysis or more precisely a Bayesian evidence synthesis to assess the burden-of-illness of campylobacteriosis in France. Using this case study, we investigate campylobacteriosis prevalence, as well as the probabilities of different events that guide the disease pathway, by (i) employing a Bayesian approach on French and foreign human studies (from active surveillance systems, laboratory surveys, physician surveys, epidemiological surveys, and so on) through the chain of events that occur during an episode of illness and (ii) including expert knowledge about this chain of events. We split the target population using an exhaustive and exclusive partition based on health status and the level of disease investigation. We assume an approximate multinomial model over this population partition. Thereby, each observed data set related to the partition brings information on the parameters of the multinomial model, improving burden-of-illness parameter estimates that can be deduced from the parameters of the basic multinomial model. This multinomial model serves as a core model to perform a Bayesian evidence synthesis. Expert knowledge is introduced by way of pseudo-data. The result is a global estimation of the burden-of-illness parameters with their accompanying uncertainty.

Probabilistic modeling of the fate of Listeria monocytogenes in diced bacon during the manufacturing process.

To assess the impact of the manufacturing process on the fate of Listeria monocytogenes, we built a generic probabilistic model intended to simulate the successive steps in the process. Contamination evolution was modeled in the appropriate units (breasts, dice, and then packaging units through the successive steps in the process). To calibrate the model, parameter values were estimated from industrial data, from the literature, and based on expert opinion. By means of simulations, the model was explored using a baseline calibration and alternative scenarios, in order to assess the impact of changes in the process and of accidental events. The results are reported as contamination distributions and as the probability that the product will be acceptable with regards to the European regulatory safety criterion. Our results are consistent with data provided by industrial partners and highlight that tumbling is a key step for the distribution of the contamination at the end of the process. Process chain models could provide an important added value for risk assessment models that basically consider only the outputs of the process in their risk mitigation strategies. Moreover, a model calibrated to correspond to a specific plant could be used to optimize surveillance.

Body composition predicted with a Bayesian network from simple variables.

The relative contributions of fat-free mass (FFM) and fat mass (FM) to body weight are key indicators for several major public health issues. Predictive models could offer new insights into body composition analysis. A non-parametric equation derived from a probabilistic Bayesian network (BN) was established by including sex, age, body weight and height. We hypothesised that it would be possible to assess the body composition of any subject from easily accessible covariables by selecting an adjusted FFM value within a reference dual-energy X-ray absorptiometry (DXA) measurement database (1999-2004 National Health and Nutrition Examination Survey (NHANES), n 10 402). FM was directly calculated as body weight minus FFM. A French DXA database (n 1140) was used (1) to adjust the model parameters (n 380) and (2) to cross-validate the model responses (n 760). French subjects were significantly different from American NHANES subjects with respect to age, weight and FM. Despite this different population context, BN prediction was highly reliable. Correlations between BN predictions and DXA measurements were significant for FFM (R2 0·94, P < 0·001, standard error of prediction (SEP) 2·82 kg) and the percentage of FM (FM%) (R2 0·81, P < 0·001, SEP 3·73 %). Two previously published linear models were applied to the subjects of the French database and compared with BN predictions. BN predictions were more accurate for both FFM and FM than those obtained from linear models. In addition, BN prediction generated stochastic variability in the FM% expressed in terms of BMI. The use of such predictions in large populations could be of interest for many public health issues.

Spread of Q fever within dairy cattle herds: key parameters inferred using a Bayesian approach.

Q fever is a worldwide zoonosis caused by Coxiella burnetii. Although ruminants are recognized as the most important source of human infection, no previous studies have focused on assessing the characteristics of the bacterial spread within a cattle herd and no epidemic model has been proposed in this context. We assess the key epidemiological parameters from field data in a Bayesian framework that takes into account the available knowledge, missing data and the uncertainty of the observation process owing to the imperfection of diagnostic tests. We propose an original individual-based Markovian model in discrete time describing the evolution of the infection for each animal. Markov chain Monte Carlo methodology is used to estimate parameters of interest from data consisting of individual health states of 217 cows of five chronically infected dairy herds sampled every week for a four-week period. Outputs are the posterior distributions of the probabilities of transition between health states and of the environmental bacterial load. Our findings show that some herds are characterized by a very low infection risk while others have a mild infection risk and a non-negligible intermittent shedding probability. Moreover, the antibody status seems to be a key point in the bacterial spread (shedders with antibodies shed for a longer period of time than shedders without antibodies). In addition to the biological insights, these estimates also provide information for calibrating simulation models to assess control strategies for C. burnetii infection.

Quantitative risk assessment from farm to fork and beyond: a global Bayesian approach concerning food-borne diseases.

A novel approach to the quantitative assessment of food-borne risks is proposed. The basic idea is to use Bayesian techniques in two distinct steps: first by constructing a stochastic core model via a Bayesian network based on expert knowledge, and second, using the data available to improve this knowledge. Unlike the Monte Carlo simulation approach as commonly used in quantitative assessment of food-borne risks where data sets are used independently in each module, our consistent procedure incorporates information conveyed by data throughout the chain. It allows "back-calculation" in the food chain model, together with the use of data obtained "downstream" in the food chain. Moreover, the expert knowledge is introduced more simply and consistently than with classical statistical methods. Other advantages of this approach include the clear framework of an iterative learning process, considerable flexibility enabling the use of heterogeneous data, and a justified method to explore the effects of variability and uncertainty. As an illustration, we present an estimation of the probability of contracting a campylobacteriosis as a result of broiler contamination, from the standpoint of quantitative risk assessment. Although the model thus constructed is oversimplified, it clarifies the principles and properties of the method proposed, which demonstrates its ability to deal with quite complex situations and provides a useful basis for further discussions with different experts in the food chain.

Quantitative risk assessment of Listeria monocytogenes in French cold-smoked salmon: I. Quantitative exposure assessment.

A quantitative assessment of the exposure to Listeria monocytogenes from cold-smoked salmon (CSS) consumption in France is developed. The general framework is a second-order (or two-dimensional) Monte Carlo simulation, which characterizes the uncertainty and variability of the exposure estimate. The model takes into account the competitive bacterial growth between L. monocytogenes and the background competitive flora from the end of the production line to the consumer phase. An original algorithm is proposed to integrate this growth in conditions of varying temperature. As part of a more general project led by the French Food Safety Agency (Afssa), specific data were acquired and modeled for this quantitative exposure assessment model, particularly time-temperature profiles, prevalence data, and contamination-level data. The sensitivity analysis points out the main influence of the mean temperature in household refrigerators and the prevalence of contaminated CSS on the exposure level. The outputs of this model can be used as inputs for further risk assessment.