Implementing Artificial Intelligence: Assessing the Cost and Benefits of Algorithmic Decision-Making in Critical Care

This article provides an overview of the most useful artificial intelligence algorithms developed in critical care, followed by a comprehensive outline of the benefits and limitations. We begin by describing how nurses and physicians might be aided by these new technologies. We then move to the possible changes in clinical guidelines with personalized medicine that will allow tailored therapies and probably will increase the quality of the care provided to patients. Finally, we describe how artificial intelligence models can unleash researchers' minds by proposing new strategies, by increasing the quality of clinical practice, and by questioning current knowledge and understanding.

Ursodeoxycholic Acid Does Not Improve COVID-19 Outcome in Hospitalized Patients

Ursodeoxycholic acid (UDCA) was demonstrated to reduce susceptibility to SARS-CoV-2 infection in vitro and improve infection course in chronic liver diseases. However, real-life evidence is lacking. We analyzed the impact of UDCA on COVID-19 outcomes in patients hospitalized in a tertiary center. Between January 2020 and January 2023, among 3847 patients consecutively hospitalized for COVID19, 57 (=UDCA group) were taking UDCA. The UDCA and the control groups (n = 3790) did not differ concerning comorbidities including diabetes mellitus type 2 (15.8% vs. 12.8%) and neoplasia (12.3% vs. 9.4%). Liver diseases and vaccination rate were more common in the UDCA group (14.0% vs. 2.5% and 54.4% vs. 30.2%, respectively). Overall mortality and CPAP treatment were 22.8 % and 15.7% in the UDCA, and 21.3% and 25.9% in the control group. Mortality was similar (p = 0.243), whereas UDCA was associated with a lower rate of CPAP treatment (OR = 0.76, p < 0.05). Treatment with UDCA was not an independent predictor of survival in patients hospitalized for COVID-19.

CT-based radiomics can identify physiological modifications of bone structure related to subjects' age and sex

PURPOSE

Radiomics of vertebral bone structure is a promising technique for identification of osteoporosis. We aimed at assessing the accuracy of machine learning in identifying physiological changes related to subjects' sex and age through analysis of radiomics features from CT images of lumbar vertebrae, and define its generalizability across different scanners.

MATERIALS AND METHODS

We annotated spherical volumes-of-interest (VOIs) in the center of the vertebral body for each lumbar vertebra in 233 subjects who had undergone lumbar CT for back pain on 3 different scanners, and we evaluated radiomics features from each VOI. Subjects with history of bone metabolism disorders, cancer, and vertebral fractures were excluded. We performed machine learning classification and regression models to identify subjects' sex and age respectively, and we computed a voting model which combined predictions.

RESULTS

The model was trained on 173 subjects and tested on an internal validation dataset of 60. Radiomics was able to identify subjects' sex within single CT scanner (ROC AUC: up to 0.9714), with lower performance on the combined dataset of the 3 scanners (ROC AUC: 0.5545). Higher consistency among different scanners was found in identification of subjects' age (R2 0.568 on all scanners, MAD 7.232 years), with highest results on a single CT scanner (R2 0.667, MAD 3.296 years).

CONCLUSION

Radiomics features are able to extract biometric data from lumbar trabecular bone, and determine bone modifications related to subjects' sex and age with great accuracy. However, acquisition from different CT scanners reduces the accuracy of the analysis.

REVersal of nEuromusculAr bLocking Agents in Patients Undergoing General Anaesthesia (REVEAL Study)

Background: Neuromuscular blocking agent (NMBA) monitoring and reversals are key to avoiding residual curarization and improving patient outcomes. Sugammadex is a NMBA reversal with favorable pharmacological properties. There is a lack of real-world data detailing how the diffusion of sugammadex affects anesthetic monitoring and practice. Methods: We conducted an electronic health record analysis study, including all adult surgical patients undergoing general anesthesia with orotracheal intubation, from January 2016 to December 2019, to describe changes and temporal trends of NMBAs and NMBA reversals administration. Results: From an initial population of 115,046 surgeries, we included 37,882 procedures, with 24,583 (64.9%) treated with spontaneous recovery from neuromuscular block and 13,299 (35.1%) with NMBA reversals. NMBA reversals use doubled over 4 years from 25.5% to 42.5%, mainly driven by sugammadex use, which increased from 17.8% to 38.3%. Rocuronium increased from 58.6% (2016) to 94.5% (2019). Factors associated with NMBA reversal use in the multivariable analysis were severe obesity (OR 3.33 for class II and OR 11.4 for class III obesity, p-value < 0.001), and high ASA score (OR 1.47 for ASA III). Among comorbidities, OSAS, asthma, and other respiratory diseases showed the strongest association with NMBA reversal administration. Conclusions: Unrestricted availability of sugammadex led to a considerable increase in pharmacological NMBA reversal, with rocuronium use also rising. More research is needed to determine how unrestricted and safer NMBA reversal affects anesthesia intraoperative monitoring and practice.

Artificial intelligence-based radiomics on computed tomography of lumbar spine in subjects with fragility vertebral fractures

PURPOSE

There is emerging evidence that radiomics analyses can improve detection of skeletal fragility. In this cross-sectional study, we evaluated radiomics features (RFs) on computed tomography (CT) images of the lumbar spine in subjects with or without fragility vertebral fractures (VFs).

METHODS

Two-hundred-forty consecutive individuals (mean age 60.4 ± 15.4, 130 males) were evaluated by radiomics analyses on opportunistic lumbar spine CT. VFs were diagnosed in 58 subjects by morphometric approach on CT or XR-ray spine (D4-L4) images. DXA measurement of bone mineral density (BMD) was performed on 17 subjects with VFs.

RESULTS

Twenty RFs were used to develop the machine learning model reaching 0.839 and 0.789 of AUROC in the train and test datasets, respectively. After correction for age, VFs were significantly associated with RFs obtained from non-fractured vertebrae indicating altered trabecular microarchitecture, such as low-gray level zone emphasis (LGLZE) [odds ratio (OR) 1.675, 95% confidence interval (CI) 1.215-2.310], gray level non-uniformity (GLN) (OR 1.403, 95% CI 1.023-1.924) and neighboring gray-tone difference matrix (NGTDM) contrast (OR 0.692, 95% CI 0.493-0.971). Noteworthy, no significant differences in LGLZE (p = 0.94), GLN (p = 0.40) and NGDTM contrast (p = 0.54) were found between fractured subjects with BMD T score < - 2.5 SD and those in whom VFs developed in absence of densitometric diagnosis of osteoporosis.

CONCLUSIONS

Artificial intelligence-based analyses on spine CT images identified RFs associated with fragility VFs. Future studies are needed to test the predictive value of RFs on opportunistic CT scans in identifying subjects with primary and secondary osteoporosis at high risk of fracture.

Outcome prediction during an ICU surge using a purely data-driven approach: A supervised machine learning case-study in critically ill patients from COVID-19 Lombardy outbreak

Purpose

COVID-19 disease frequently affects the lungs leading to bilateral viral pneumonia, progressing in some cases to severe respiratory failure requiring ICU admission and mechanical ventilation. Risk stratification at ICU admission is fundamental for resource allocation and decision making. We assessed performances of three machine learning approaches to predict mortality in COVID-19 patients admitted to ICU using early operative data from the Lombardy ICU Network.

Methods

This is a secondary analysis of prospectively collected data from Lombardy ICU network. A logistic regression, balanced logistic regression and random forest were built to predict survival on two datasets: dataset A included patient demographics, medications before admission and comorbidities, and dataset B included respiratory data the first day in ICU.

Results

Models were trained on 1484 patients on four outcomes (7/14/21/28 days) and reached the greatest predictive performance at 28 days (F1-score: 0.75 and AUC: 0.80). Age, number of comorbidities and male gender were strongly associated with mortality. On dataset B, mode of ventilatory assistance at ICU admission and fraction of inspired oxygen were associated with an increase in prediction performances.

Conclusions

Machine learning techniques might be useful in emergency phases to reach good predictive performances maintaining interpretability to gain knowledge on complex situations and enhance patient management and resources.

Artificial intelligence processing electronic health records to identify commonalities and comorbidities cluster at Immuno Center Humanitas

Background

Comorbidities are common in chronic inflammatory conditions, requiring multidisciplinary treatment approach. Understanding the link between a single disease and its comorbidities is important for appropriate treatment and management. We evaluate the ability of an NLP-based process for knowledge discovery to detect information about pathologies, patients' phenotype, doctors' prescriptions and commonalities in electronic medical records, by extracting information from free narrative text written by clinicians during medical visits, resulting in the extraction of valuable information and enriching real world evidence data from a multidisciplinary setting.

Methods

We collected clinical notes from the Allergy Department of Humanitas Research Hospital written in the last 3 years and used it to look for diseases that cluster together as comorbidities associated to the main pathology of our patients, and for the extent of prescription of systemic corticosteroids, thus evaluating the ability of NLP-based tools for knowledge discovery to extract structured information from free text.

Results

We found that the 3 most frequent comorbidities to appear in our clusters were asthma, rhinitis, and urticaria, and that 991 (of 2057) patients suffered from at least one of these comorbidities. The clusters which co-occur particularly often are oral allergy syndrome and urticaria (131 patients), angioedema and urticaria (105 patients), rhinitis and asthma (227 patients). With regards to systemic corticosteroid prescription volume by our clinicians, we found it was lower when compared to the therapy the patients followed before coming to our attention, with the exception of two diseases: Chronic obstructive pulmonary disease and Angioedema.

Conclusions

This analysis seems to be valid and is confirmed by the data from the literature. This means that NLP tools could have significant role in many other research fields of medicine, as it may help identify other important, and possibly previously neglected clusters of patients with comorbidities and commonalities. Another potential benefit of this approach lies in its potential ability to foster a multidisciplinary approach, using the same drugs to treat pathologies normally treated by physicians in different branches of medicine, thus saving resources and improving the pharmacological management of patients.

Early prediction of SARS-CoV-2 reproductive number from environmental, atmospheric and mobility data: A supervised machine learning approach

INTRODUCTION

SARS-CoV-2 was declared a pandemic by the WHO on March 11th, 2020. Public protective measures were enforced in every country to limit the diffusion of SARS-CoV-2. Its transmission, mainly by droplets, has been measured by the effective reproduction number (Rt) that counts the number of secondary cases caused in a population by an average infectious individual at time t. Current strategies to calculate Rt reflect the number of secondary cases after several days, due to a delay from symptoms onset to reporting. We propose a complementary Rt estimation using supervised machine learning techniques to predict short term variations with more timely results.

MATERIAL AND METHODS

Our primary goal was to predict Rt of the current day in the twelve provinces of Lombardy with the highest possible accuracy, and with no influence of the local testing strategies. We gathered data about mobility, weather, and pollution from different public sources as a proxy of human behavior and public health measures. We built four supervised machine learning algorithms with different strategies: the outcome variable was the daily median Rt values per province obtained from officially adopted algorithms.

RESULTS

Data from 243 days for every province were presented to our four models (from February 15th, 2020, to October 14th, 2020). Two models using differential calculation of Rt instead of the raw values showed the highest mean coefficient of determination (0.93 for both) and residuals reported the lowest mean error (-0.03 and 0.01) and standard deviation (0.13 for both) as well. The one with access to the value of Rt of the day before heavily relied on that feature for prediction, while the other one had more distributed weights.

DISCUSSION

The model that had not access to the Rt value of the previous day and used Rt differential value as outcome (FDRt) was considered the most robust according to the metrics. Its forecasts were able to predict the trend that Rt values would have developed over different weeks, but it was not particularly accurate in predicting the precise value of Rt. A correlation among mobility, atmospheric, features, pollution and Rt values is plausible, but further testing should be performed.

A 'Multiomic' Approach of Saliva Metabolomics, Microbiota, and Serum Biomarkers to Assess the Need of Hospitalization in Coronavirus Disease 2019

BACKGROUND AND AIMS

The SARS-CoV-2 pandemic has overwhelmed the treatment capacity of the health care systems during the highest viral diffusion rate. Patients reaching the emergency department had to be either hospitalized (inpatients) or discharged (outpatients). Still, the decision was taken based on the individual assessment of the actual clinical condition, without specific biomarkers to predict future improvement or deterioration, and discharged patients often returned to the hospital for aggravation of their condition. Here, we have developed a new combined approach of omics to identify factors that could distinguish coronavirus disease 19 (COVID-19) inpatients from outpatients.

METHODS

Saliva and blood samples were collected over the course of two observational cohort studies. By using machine learning approaches, we compared salivary metabolome of 50 COVID-19 patients with that of 270 healthy individuals having previously been exposed or not to SARS-CoV-2. We then correlated the salivary metabolites that allowed separating COVID-19 inpatients from outpatients with serum biomarkers and salivary microbiota taxa differentially represented in the two groups of patients.

RESULTS

We identified nine salivary metabolites that allowed assessing the need of hospitalization. When combined with serum biomarkers, just two salivary metabolites (myo-inositol and 2-pyrrolidineacetic acid) and one serum protein, chitinase 3-like-1 (CHI3L1), were sufficient to separate inpatients from outpatients completely and correlated with modulated microbiota taxa. In particular, we found Corynebacterium 1 to be overrepresented in inpatients, whereas Actinomycetaceae F0332, Candidatus Saccharimonas, and Haemophilus were all underrepresented in the hospitalized population.

CONCLUSION

This is a proof of concept that a combined omic analysis can be used to stratify patients independently from COVID-19.

An individualized algorithm to predict mortality in COVID-19 pneumonia: a machine learning based study

INTRODUCTION

Identifying SARS-CoV-2 patients at higher risk of mortality is crucial in the management of a pandemic. Artificial intelligence techniques allow one to analyze large amounts of data to find hidden patterns. We aimed to develop and validate a mortality score at admission for COVID-19 based on high-level machine learning.

MATERIAL AND METHODS

We conducted a retrospective cohort study on hospitalized adult COVID-19 patients between March and December 2020. The primary outcome was in-hospital mortality. A machine learning approach based on vital parameters, laboratory values and demographic features was applied to develop different models. Then, a feature importance analysis was performed to reduce the number of variables included in the model, to develop a risk score with good overall performance, that was finally evaluated in terms of discrimination and calibration capabilities. All results underwent cross-validation.

RESULTS

1,135 consecutive patients (median age 70 years, 64% male) were enrolled, 48 patients were excluded, and the cohort was randomly divided into training (760) and test (327) groups. During hospitalization, 251 (22%) patients died. After feature selection, the best performing classifier was random forest (AUC 0.88 ±0.03). Based on the relative importance of each variable, a pragmatic score was developed, showing good performances (AUC 0.85 ±0.025), and three levels were defined that correlated well with in-hospital mortality.

CONCLUSIONS

Machine learning techniques were applied in order to develop an accurate in-hospital mortality risk score for COVID-19 based on ten variables. The application of the proposed score has utility in clinical settings to guide the management and prognostication of COVID-19 patients.

Clinical Outcomes in the Second versus First Pandemic Wave in Italy: Impact of Hospital Changes and Reorganization

The region of Lombardy was the epicenter of the COVID-19 outbreak in Italy. Emergency Hospital 19 (EH19) was built in the Milan metropolitan area during the pandemic’s second wave as a facility of Humanitas Clinical and Research Center (HCRC). The present study aimed to assess whether the implementation of EH19 was effective in improving the quality of care of COVID-19 patients during the second wave compared with the first one. The demographics, mortality rate, and in-hospital length of stay (LOS) of two groups of patients were compared: the study group involved patients admitted at HCRC and managed in EH19 during the second pandemic wave, while the control group included patients managed exclusively at HCRC throughout the first wave. The study and control group included 903 (56.7%) and 690 (43.3%) patients, respectively. The study group was six years older on average and had more pre-existing comorbidities. EH19 was associated with a decrease in the intensive care unit admission rate (16.9% vs. 8.75%, p < 0.001), and an equal decrease in invasive oxygen therapy (3.8% vs. 0.23%, p < 0.001). Crude mortality was similar but overlap propensity score weighting revealed a trend toward a potential small decrease. The adjusted difference in LOS was not significant. The implementation of an additional COVID-19 hospital facility was effective in improving the overall quality of care of COVID-19 patients during the first wave of the pandemic when compared with the second. Further studies are necessary to validate the suggested approach.

The antibody response to SARS-CoV-2 infection persists over at least 8 months in symptomatic patients

Background

Persistence of antibodies to SARS-CoV-2 viral infection may depend on several factors and may be related to the severity of disease or to the different symptoms.

Methods

We evaluated the antibody response to SARS-CoV-2 in personnel from 9 healthcare facilities and an international medical school and its association with individuals’ characteristics and COVID-19 symptoms in an observational cohort study. We enrolled 4735 subjects (corresponding to 80% of all personnel) for three time points over a period of 8–10 months. For each participant, we determined the rate of antibody increase or decrease over time in relation to 93 features analyzed in univariate and multivariate analyses through a machine learning approach.

Results

Here we show in individuals positive for IgG (≥12 AU/mL) at the beginning of the study an increase [p = 0.0002] in antibody response in paucisymptomatic or symptomatic subjects, particularly with loss of taste or smell (anosmia/dysgeusia: OR 2.75, 95% CI 1.753 – 4.301), in a multivariate logistic regression analysis in the first three months. The antibody response persists for at least 8–10 months.

Conclusions

SARS-CoV-2 infection induces a long lasting antibody response that increases in the first months, particularly in individuals with anosmia/dysgeusia. This may be linked to the lingering of SARS-CoV-2 in the olfactory bulb.

Levi and Ubaldi et al. evaluate SARS-CoV-2 seroprevalence in a cohort of 4735 healthcare workers in northern Italy. In seropositive individuals, they show that antibodies are maintained over a period of 8 to 10 months and associate changes in antibody levels over this period with symptoms and specific subgroups of participants.

SARS-CoV-2 infection activates the body’s immune system to fight off infection. This immune response results in the production of proteins in the blood that target the virus called antibodies. The extent and duration of this antibody response may be associated with the type of symptoms the infected person is experiencing. Here, we analyzed SARS-CoV-2 antibody levels in individuals with asymptomatic, mild symptomatic (paucisymptomatic) and symptomatic disease in relation to the type of symptoms. We find that the antibody response is higher in people with symptoms and increases in the first three months, particularly in individuals with loss of smell or taste. In all people with SARS-CoV-2 antibodies at the start of the study, levels in the blood last for at least 8–10 months. Hence, SARS-CoV-2 infection is characterized by a long lasting antibody response which may protect from subsequent infections.

The effect of COVID-19 epidemic on vital signs in hospitalized patients: a pre-post heat-map study from a large teaching hospital

The Lombardy SARS-CoV-2 outbreak in February 2020 represented the beginning of COVID-19 epidemic in Italy. Hospitals were flooded by thousands of patients with bilateral pneumonia and severe respiratory, and vital sign derangements compared to the standard hospital population. We propose a new visual analysis technique using heat maps to describe the impact of COVID-19 epidemic on vital sign anomalies in hospitalized patients. We conducted an electronic health record study, including all confirmed COVID-19 patients hospitalized from February 21st, 2020 to April 21st, 2020 as cases, and all non-COVID-19 patients hospitalized in the same wards from January 1st, 2018 to December 31st, 2018. All data on temperature, peripheral oxygen saturation, respiratory rate, arterial blood pressure, and heart rate were retrieved. Derangement of vital signs was defined according to predefined thresholds. 470 COVID-19 patients and 9241 controls were included. Cases were older than controls, with a median age of 79 vs 76 years in non survivors (p =  < 0.002). Gender was not associated with mortality. Overall mortality in COVID-19 hospitalized patients was 18%, ranging from 1.4% in patients below 65 years to about 30% in patients over 65 years. Heat maps analysis demonstrated that COVID-19 patients had an increased frequency in episodes of compromised respiratory rate, acute desaturation, and fever. COVID-19 epidemic profoundly affected the incidence of severe derangements in vital signs in a large academic hospital. We validated heat maps as a method to analyze the clinical stability of hospitalized patients. This method may help to improve resource allocation according to patient characteristics.

Quantitative chest CT analysis in COVID-19 to predict the need for oxygenation support and intubation

Objective

Lombardy (Italy) was the epicentre of the COVID-19 pandemic in March 2020. The healthcare system suffered from a shortage of ICU beds and oxygenation support devices. In our Institution, most patients received chest CT at admission, only interpreted visually. Given the proven value of quantitative CT analysis (QCT) in the setting of ARDS, we tested QCT as an outcome predictor for COVID-19.

Methods

We performed a single-centre retrospective study on COVID-19 patients hospitalised from January 25, 2020, to April 28, 2020, who received CT at admission prompted by respiratory symptoms such as dyspnea or desaturation. QCT was performed using a semi-automated method (3D Slicer). Lungs were divided by Hounsfield unit intervals. Compromised lung (%CL) volume was the sum of poorly and non-aerated volumes (− 500, 100 HU). We collected patient’s clinical data including oxygenation support throughout hospitalisation.

Results

Two hundred twenty-two patients (163 males, median age 66, IQR 54–6) were included; 75% received oxygenation support (20% intubation rate). Compromised lung volume was the most accurate outcome predictor (logistic regression, p < 0.001). %CL values in the 6–23% range increased risk of oxygenation support; values above 23% were at risk for intubation. %CL showed a negative correlation with PaO₂/FiO₂ ratio (p < 0.001) and was a risk factor for in-hospital mortality (p < 0.001).

Conclusions

QCT provides new metrics of COVID-19. The compromised lung volume is accurate in predicting the need for oxygenation support and intubation and is a significant risk factor for in-hospital death. QCT may serve as a tool for the triaging process of COVID-19.

Key Points

• Quantitative computer-aided analysis of chest CT (QCT) provides new metrics of COVID-19.

• The compromised lung volume measured in the − 500, 100 HU interval predicts oxygenation support and intubation and is a risk factor for in-hospital death.

• Compromised lung values in the 6–23% range prompt oxygenation therapy; values above 23% increase the need for intubation.

The Role of Baroreflex Sensitivity in Acute Hypotensive Episodes Prediction in the Intensive Care Unit

A life threatening condition in Intensive Care Unit (ICU) is the Acute Hypotensive Episode (AHE). Patients experiencing an AHE may suffer from irreversible organ damage associated with increased mortality. Predicting the onset of AHE could be of pivotal importance to establish appropriate and timely interventions. We propose a method that, using waveforms widely acquired in ICU, like Arterial Blood Pressure (ABP) and Electrocardiogram (ECG), will extract features relative to the cardiac system to predict whether or not a patient will experience a hypotensive episode. Specifically, we want to assess if there are hidden patterns in the dynamics of baroreflex able to improve the prediction of AHEs. We will investigate the predictive power of features related to the baroreflex by performing classifications with and without them. Results are obtained using 17 classifiers belonging to different model families: classification trees, Support Vector Machines (SVMs), K-Nearest Neighbors (KNNs) replicated with different set of hyper-parameters and logistic regression. On average, the use of baroreflex features in the AHE prediction process increases the Area Under the Curve (AUC) by 10%.