Prediction models for diagnosis and prognosis of covid-19: systematic review and critical appraisal

Early suPAR levels as a predictor of COVID-19 severity: A new tool for efficient patient triage

BACKGROUND

Following several waves of the COVID-19 pandemic, we are now facing a lower but persistent rate of SARS-CoV-2 infections, with seasonal resurgences often coinciding with other respiratory tract infections.

OBJECTIVE

We aimed to identify early clinico-biological variables predictive of an unfavorable outcome in patients with primary SARS-CoV-2 infection. We also evaluated the role of suPAR, an innovative biomarker, in predicting disease severity.

METHODS

We included 255 patients with PCR-confirmed primary SARS-CoV-2 infection and with a 30-day follow-up minimum. Blood samples were collected within the first 24 h of hospitalization to measure suPAR levels. Comprehensive data from medical records were analyzed to assess their predictive value in stratifying patients into seven severity groups, with groups 1 to 3 representing severe COVID-19 (death, intubation, ECMO, or non-invasive ventilation).

RESULTS

Early plasma suPAR levels were significantly associated with severe disease progression, as evidenced by ANOVA and logistic regression models, highlighting suPAR as a persistent predictive factor for unfavorable outcomes.

CONCLUSION

Our findings suggest that a single suPAR measurement, performed early after a positive PCR test for SARS-CoV-2, holds strong predictive value for patient outcomes. This biomarker, alongside pulse oximetry and CT scan results, could be instrumental in early patient triage during seasonal COVID-19 resurgences.

Predicting COVID-19 progression in hospitalized patients in Kurdistan Province using a multi-state model

Objectives

This study aimed to implement a multi-state risk prediction model to predict the progression of COVID-19 cases among hospitalized patients in Kurdistan province by analyzing hospital care data.

Methods

This retrospective analysis consisted of data from 17,286 patients admitted to hospitals with COVID-19 from March 23, 2019, to December 19, 2021, in various areas in the Kurdistan province. A multi-state prediction model was used to show that each transition is predicted by a different set of variables. These variables include underlying diseases (like diabetes, hypertension, etc.) and sociodemographic information (like sex and age). Model aims to predict the likelihood of recovery, the need for critical care intervention (e.g., transfer to isolation units or the ICU), or exits from the hospitalization course. We performed the statistical analysis using R software and the mstate package.

Results

Of the hospitalized patients studied, 5.6% died of the disease, 6.6% were admitted to ICUs, and 38.72% were treated in isolation units. Mortality rates in general wards, isolation units, and the ICU were 3.48%, 4.56%, and 26.6%, respectively. Significant predictors for ICU admission include age over 60 years (HR: 1.46, 95% CI 1.37-1.55), kidney-related conditions (HR: 2.19, 95% CI 1.65-2.91), cardiovascular diseases (HR: 1.68, 95% CI 1.46-1.94), lung disease (HR: 1.89,‏95% CI 1.43-2.05), and cancer (HR: 2.46,‏95% CI 1.77-3.41). The likelihood of in-hospital death is significantly increased by age over 60 years (HR: 2.40, 95% CI 2.09-2.76), diabetes (HR: 1.97, 95% CI 1.45-2.68), high blood pressure (HR: 2.30, 95% CI 1.78-2.97), and history of heart disease (HR: 3.01, 95% CI 2.29-3.95).

Conclusion

The model helps the provider and policymakers to make an informed decision depending on patient management and resource allocation within the health care systems.

A meta-analysis of the diagnostic test accuracy of artificial intelligence predicting emergency department dispositions

BACKGROUND

The rapid advancement of Artificial Intelligence (AI) has led to its widespread application across various domains, showing encouraging outcomes. Many studies have utilized AI to forecast emergency department (ED) disposition, aiming to forecast patient outcomes earlier and to allocate resources better; however, a dearth of comprehensive review literature exists to assess the objective performance standards of these predictive models using quantitative evaluations. This study aims to conduct a meta-analysis to assess the diagnostic accuracy of AI in predicting ED disposition, encompassing admission, critical care, and mortality.

METHODS

Multiple databases, including Scopus, Springer, ScienceDirect, PubMed, Wiley, Sage, and Google Scholar, were searched until December 31, 2023, to gather relevant literature. Risk of bias was assessed using the Prediction Model Risk of Bias Assessment Tool. Pooled estimates of sensitivity, specificity, and area under the receiver operating characteristic curve (AUROC) were calculated to evaluate AI's predictive performance. Sub-group analyses were performed to explore covariates affecting AI predictive model performance.

RESULTS

The study included 88 articles possessed with 117 AI models, among which 39, 45, and 33 models predicted admission, critical care, and mortality, respectively. The reported statistics for sensitivity, specificity, and AUROC represent pooled summary measures derived from the component studies included in this meta-analysis. AI's summary sensitivity, specificity, and AUROC for predicting admission were 0.81 (95% Confidence Interval [CI] 0.74-0.86), 0.87 (95% CI 0.81-0.91), and 0.87 (95% CI 0.84-0.93), respectively. For critical care, the values were 0.86 (95% CI 0.79-0.91), 0.89 (95% CI 0.83-0.93), and 0.93 (95% CI 0.89-0.95), respectively, and for mortality, they were 0.85 (95% CI 0.80-0.89), 0.94 (95% CI 0.90-0.96), and 0.93 (95% CI 0.89-0.96), respectively. Emergent sample characteristics and AI techniques showed evidence of significant covariates influencing the heterogeneity of AI predictive models for ED disposition.

CONCLUSIONS

The meta-analysis indicates promising performance of AI in predicting ED disposition, with certain potential for improvement, especially in sensitivity. Future research could explore advanced AI techniques such as ensemble learning and cross-validation with hyper-parameter tuning to enhance predictive model efficacy.

TRIAL REGISTRATION

This systematic review was not registered with PROSPERO or any other similar registry because the review was completed prior to the opportunity for registration, and PROSPERO currently does not accept registrations for reviews that are already completed. We are committed to transparency and have adhered to best practices in systematic review methodology throughout this study.

In-depth analysis of the risk factors for persistent severe acute respiratory syndrome coronavirus 2 infection and construction of predictive models: an exploratory research study

BACKGROUND

Persistent severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection differs from long coronavirus disease (COVID-19) (acute symptoms ≥ 12 weeks post-clearance). The Omicron BA.5 variant has a shorter median clearance time (10-14 days) than the Delta variant, suggesting that the traditional 20-day diagnostic threshold may delay interventions in high-risk populations. This study integrated multi-threshold analysis (14/20/30 days), whole-genome sequencing, and machine learning to investigate diagnostic thresholds for persistent SARS-CoV-2 infection and developed a generalizable risk prediction model.

METHODS

This retrospective study analyzed data from 1,216 patients with COVID-19 hospitalized at Aerospace Center Hospital between January 2021 and October 2024. We used whole-genome sequencing to genotype all COVID-19 cases and to identify major variants (such as Omicron BA. 5, Delta). The outcome, "persistent SARS-CoV-2 infection," was defined as viral nucleic acid positivity ≥ 14 days. Risk factors associated with persistent infection were identified through subgroup analysis with multiple logistic regression (adjusted for age, comorbidities, vaccination status, and virus strain) and machine learning models (70% training, 30% testing dataset).

RESULTS

Persistent SARS-CoV-2 infection was identified in 15.5% (188/1,216) of hospitalized COVID-19 patients. Key predictors included comorbidities-hypertension, diabetes, and active malignancy-and immune dysfunction, marked by reduced B-cell and CD4 + T-cell counts. Unvaccinated patients exhibited an 82% higher risk of persistent infection. Elevated inflammatory markers (C-reactive protein and interleukin-6) and bilateral lung infiltrates on computed tomography further distinguished persistent cases. The predictive model demonstrated strong discrimination with an area under the curve (AUC) of 0.847 (95% confidence interval: 0.815-0.879) and an AUC of 0.81 externally in external validation, underscoring its clinical utility for risk stratification.

CONCLUSIONS

Hypertension, diabetes, malignancy, immunosuppression (low B/CD4 + cells), and non-vaccination are independent risk factors for persistent SARS-CoV-2 infection. Integrating these factors into clinical risk stratification may optimize management of high-risk populations.

Rapid development of a registry to accelerate COVID-19 vaccine clinical trials

Response to the SARS-Cov-2 pandemic required the unprecedented, rapid activation of the COVID-19 Prevention Network (CoVPN) representing hundreds of sites conducting vaccine clinical trials. The CoVPN Volunteer Screening Registry (VSR) collected participant information, distributed qualified candidates across sites, and monitored enrollment progress. The system consisted of three web-based interfaces. The Volunteer Questionnaire flowed into a secure database. The Site Portal supported volunteer selection, analytics, and enrollment. The Administrative Portal enabled dynamic analytic reports by geography, clinical trial, and site, including volunteering rates over time. The VSR collected over 650,000 volunteers, serving a key role in the recruitment of diverse participants for multiple Phase 3 clinical trials. Over 47% of the 166,729 volunteers selected for screening represented prioritized groups. The success of the VSR demonstrates how digital tools can be rapidly yet safely integrated into an accelerated clinical trial setting. We summarize the development of the system and lessons learned for pandemic preparedness.

Assessing the risk of heart failure in type 2 diabetes: a prediction algorithm to sustain the evaluation of NT-proBNP in primary care

PURPOSE

Heart failure (HF) is a disease that leads to approximately 300,000 fatalities annually in Europe and 250,000 deaths each year in the United States. Type 2 Diabetes Mellitus (T2DM) is a significant risk factor for HF, and testing for N-terminal (NT)-pro hormone BNP (NT-proBNP) can aid in early detection of HF in T2DM patients. We therefore developed and validated the HFriskT2DM-HScore, an algorithm to predict the risk of HF in T2DM patients, so guiding NT-proBNP investigation in a primary care setting.

METHODS

Using a primary care database, we formed a cohort of patients aged ≥18 years diagnosed with T2DM between 2002 and 2022. A multivariate Cox model was adopted to assess the determinants associated with the occurrence of HF to combine them to form an individual score.

RESULTS

Within a cohort of 167,618 patients (52.3% males; mean age 64.4 (SD: 14.4); HF rate equal to 6.7 cases per 1000 person-years), we developed the HFriskT2DM-HScore. When it was applied to the validation sub-cohort we found an explained variation and discrimination value of 43% (95% CI: 42-44) and 81% (95% CI: 0.80-0.83), respectively. Calibration slope was equal to 0.93 (95% CI: 0.81-1.1; p = 0.3123).

CONCLUSION

The HFriskT2DM-HScore might be implemented as a decision support system for primary care to appropriately ease the prescription of NT-proBNP and early identification of HF.

An updated review of pulmonary radiological features of acute and chronic COVID-19

PURPOSE OF REVIEW

Significant progress has been made in our understanding of the acute and chronic clinical and radiological manifestations of coronavirus-19 (COVID-19). This article provides an updated review on pulmonary COVID-19, while highlighting the key imaging features that can identify and distinguish acute COVID-19 pneumonia and its chronic sequelae from other diseases.

RECENT FINDINGS

Acute COVID-19 pneumonia typically presents with manifestations of organizing pneumonia on computed tomography (CT). In cases of severe disease, patients clinically progress to acute respiratory distress syndrome, which manifests as diffuse alveolar damage on CT. The most common chronic imaging finding is ground-glass opacities, which commonly resolves, as well as subpleural bands and reticulation. Pulmonary fibrosis is an overall rare complication of COVID-19, with characteristic features, including architectural distortion, and traction bronchiectasis.

SUMMARY

Chest CT can be a helpful adjunct tool in both diagnosing and managing acute COVID-19 pneumonia and its chronic sequelae. It can identify high-risk cases and guide decision-making, particularly in cases of severe or complicated disease. Follow-up imaging can detect persistent lung abnormalities associated with long COVID and guide appropriate management.

Using machine learning involving diagnoses and medications as a risk prediction tool for post-acute sequelae of COVID-19 (PASC) in primary care

BACKGROUND

The aim of our study was to determine whether the application of machine learning could predict PASC by using diagnoses from primary care and prescribed medication 1 year prior to PASC diagnosis.

METHODS

This population-based case-control study included subjects aged 18-65 years from Sweden. Stochastic gradient boosting was used to develop a predictive model using diagnoses received in primary care, hospitalization due to acute COVID- 19, and prescribed medication. The variables with normalized relative influence (NRI) ≥ 1% showed were considered predictive. Odds ratios of marginal effects (ORME) were calculated.

RESULTS

The study included 47,568 PASC cases and controls. More females (n = 5113) than males (n = 2815) were diagnosed with PASC. Key predictive factors identified in both sexes included prior hospitalization due to acute COVID- 19 (NRI 16.1%, ORME 18.8 for females; NRI 41.7%, ORME 31.6 for males), malaise and fatigue (NRI 14.5%, ORME 4.6 for females; NRI 11.5%, ORME 7.9 for males), and post-viral and related fatigue syndromes (NRI 10.1%, ORME 21.1 for females; NRI 6.4%, ORME 28.4 for males).

CONCLUSIONS

Machine learning can predict PASC based on previous diagnoses and medications. Use of this AI method could support diagnostics of PASC in primary care and provide insight into PASC etiology.

A Systematic Review and Implementation Guidelines of Multimodal Foundation Models in Medical Imaging

Artificial Intelligence (AI) holds immense potential to transform healthcare, yet progress is often hindered by the reliance on large labeled datasets and unimodal data. Multimodal Foundation Models (FMs), particularly those leveraging Self-Supervised Learning (SSL) on multimodal data, offer a paradigm shift towards label-efficient, holistic patient modeling. However, the rapid emergence of these complex models has created a fragmented landscape. Here, we provide a systematic review of multimodal FMs for medical imaging applications. Through rigorous screening of 1,144 publications (2012-2024) and in-depth analysis of 48 studies, we establish a unified terminology and comprehensively assess the current state-of-the-art. Our review aggregates current knowledge, critically identifies key limitations and underexplored opportunities, and culminates in actionable guidelines for researchers, clinicians, developers, and policymakers. This work provides a crucial roadmap to navigate and accelerate the responsible development and clinical translation of next-generation multimodal AI in healthcare.

Dissecting clinical features of COVID-19 in a cohort of 21,312 acute care patients

BACKGROUND

Although, COVID-19 has resulted in over 7 million deaths globally, many questions still remain about the risk factors for disease severity and the effects of variants and vaccinations over the course of the pandemic. To address this gap, we conducted a retrospective analysis of electronic health records from COVID-19 patients over 2.5 years of the COVID-19 pandemic to identify associated clinical features.

METHODS

We analyze a retrospective cohort of 21,312 acute-care patients over a 2.5 year period and define six clinical trajectory groups (TGs) associated with demographics, diagnoses, vitals, labs, imaging, consultations, and medications.

RESULTS

We show that the proportion of mild patients increased over time, particularly during Omicron waves. Additionally, while mild and fatal patients had differences in age, age did not distinguish patients with severe versus critical disease. Furthermore, we find that both male sex and Hispanic/Latino ethnicity are associated with more severe/critical TGs. More severe patients also have a higher rate of neuropsychiatric diagnoses and consultations, along with an immunological signature of high neutrophils and immature granulocytes, and low lymphocytes and monocytes. Interestingly, low albumin is one of the best lab predictors of COVID-19 severity in association with higher malnutrition in severe/critical patients, raising concern of nutritional insufficiency influencing COVID-19 outcomes. Despite this, only a small fraction of severe/critical patients had nutritional labs checked (e.g. Vitamin D, thiamine, B vitamins) or received vitamin supplementation.

CONCLUSIONS

Our findings expand on clinical risk factors in COVID-19, and highlight the interaction between severity, nutritional status, and neuropsychiatric complications in acute care patients to enable identification of patients at risk for severe disease.

A Machine Learning Model for Predicting Intensive Care Unit Admission in Inpatients with COVID-19 Using Clinical Data and Laboratory Biomarkers

Background: Artificial intelligence tools can help improve the clinical management of patients with severe COVID-19. The aim of this study was to validate a machine learning model to predict admission to the Intensive Care Unit (ICU) in individuals with COVID-19. Methods: A total of 201 hospitalized patients with COVID-19 were included. Sociodemographic and clinical data as well as laboratory biomarker results were obtained from medical records and the clinical laboratory information system. Three machine learning models were generated, trained, and internally validated: logistic regression (LR), random forest (RF), and extreme gradient boosting (XGBoost). The models were evaluated for sensitivity (Sn), specificity (Sp), area under the curve (AUC), precision (P), SHapley Additive exPlanation (SHAP) values, and the clinical utility of predictive models using decision curve analysis (DCA). Results: The predictive model included the following variables: type 2 diabetes mellitus (T2DM), obesity, absolute neutrophil and basophil counts, the neutrophil-to-lymphocyte ratio (NLR), and D-dimer levels on the day of hospital admission. LR showed an Sn of 0.67, Sp of 0.65, AUC of 0.74, and P of 0.66. RF achieved an Sn of 0.87, Sp of 0.83, AUC of 0.96, and P of 0.85. XGBoost demonstrated an Sn of 0.87, Sp of 0.85, AUC of 0.95, and P of 0.86. Conclusions: Among the evaluated models, XGBoost showed robust predictive performance (Sn = 0.87, Sp = 0.85, AUC = 0.95, P = 0.86) and a favorable net clinical benefit in the decision curve analysis, confirming its suitability for predicting ICU admission in COVID-19 and aiding clinical decision-making.

A systematic review on the efficacy of artificial intelligence in geriatric healthcare: a critical analysis of current literature

OBJECTIVE

To carry out systematic analysis of existing literature on role of Artificial Intelligence in geriatric patient healthcare.

METHODS

A detailed online search was carried out using search phrases in reliable sources of information like Pubmed database, Embase database, Ovid database, Global Health database, PsycINFO, and Web of Science. Study specific information was gathered, including the organisation, year of publication, nation, setting, design of the research, information about population, size of study sample, group dynamics, eligibility and exclusion requirements, information about intervention, duration of exposure to the intervention , comparators, details of outcome measures, scheduling of evaluations, and consequences. After information gathering, the reviewers gathered to discuss any differences.

RESULTS

Thirty-one studies were finally selected for systemic review. Although there was some disagreement on the acceptance of AI-enhanced treatments in LTC settings, this review indicated that there was little consensus about the efficacy of those initiatives for older individuals. Social robots have been shown to increase social interaction and mood, but the data was more conflicting and less definitive for the other innovations and consequences. The majority of research evaluated a variety of results, which made it impossible to synthesise them in a meaningful way and prevented a meta-analysis. In addition, many studies have moderate to severe bias risks due to underpowered design CONCLUSION: It is challenging to determine whether AI supplemented technologies for geriatric patients are significantly beneficial. Although some encouraging findings were made, more study is required.

Clinical implementation of AI-based screening for risk for opioid use disorder in hospitalized adults

Adults with opioid use disorder (OUD) are at increased risk for opioid-related complications and repeated hospital admissions. Routine screening for patients at risk for an OUD to prevent complications is not standard practice in many hospitals, leading to missed opportunities for intervention. The adoption of electronic health records (EHRs) and advancements in artificial intelligence (AI) offer a scalable approach to systematically identify at-risk patients for evidence-based care. This pre-post quasi-experimental study evaluated whether an AI-driven OUD screener embedded in the EHR was non-inferior to usual care in identifying patients for addiction medicine consultations, aiming to provide a similarly effective but more scalable alternative to human-led ad hoc consultations. The AI screener used a convolutional neural network to analyze EHR notes in real time, identifying patients at risk and recommending consultations. The primary outcome was the proportion of patients who completed a consultation with an addiction medicine specialist, which included interventions such as outpatient treatment referral, management of complicated withdrawal, medication management for OUD and harm reduction services. The study period consisted of a 16-month pre-intervention phase followed by an 8-month post-intervention phase, during which the AI screener was implemented to support hospital providers in identifying patients for consultation. Consultations did not change between periods (1.35% versus 1.51%, P < 0.001 for non-inferiority). In secondary outcome analysis, the AI screener was associated with a reduction in 30-day readmissions (odds ratio: 0.53, 95% confidence interval: 0.30-0.91, P = 0.02) with an incremental cost of US$6,801 per readmission avoided, demonstrating its potential as a scalable, cost-effective solution for OUD care. ClinicalTrials.gov registration: NCT05745480 .

Analysis of six consecutive waves of ICU-admitted COVID-19 patients: key findings and insights from a Portuguese population

Identifying high-risk patients, particularly in intensive care units (ICUs), enhances treatment and reduces severe outcomes. Since the pandemic, numerous studies have examined COVID-19 patient profiles and factors linked to increased mortality. Despite six pandemic waves, to the best of our knowledge, there is no extensive comparative analysis of patients' characteristics across these waves in Portugal. Thus, we aimed to analyze the demographic and clinical features of 1041 COVID-19 patients admitted to an ICU and their relationship with the different SARS-Cov-2 variants in Portugal. Additionally, we conducted an in-depth examination of factors contributing to early and late mortality by analyzing clinical data and laboratory results from the first 72 h of ICU admission. Our findings revealed a notable decline in ICU admissions due to COVID-19, with the highest mortality rates observed during the second and third waves. Furthermore, immunization could have significantly contributed to the reduction in the median age of ICU-admitted patients and the severity of their conditions. The factors contributing to early and late mortality differed. Age, wave number, D-dimers, and procalcitonin were independently associated with the risk of early death. As a measure of discriminative power for the derived multivariable model, an AUC of 0.825 (p < 0.001; 95% CI, 0.719-0.931) was obtained. For late mortality, a model incorporating age, wave number, hematologic cancer, C-reactive protein, lactate dehydrogenase, and platelet counts resulted in an AUC of 0.795 (p < 0.001; 95% CI, 0.759-0.831). These findings underscore the importance of conducting comprehensive analyses across pandemic waves to better understand the dynamics of COVID-19.

Health diagnosis associated with COVID-19 death in the United States: A retrospective cohort study using electronic health records

BACKGROUND

The United States has experienced high surge in COVID-19 cases since the dawn of 2020. Identifying the types of diagnoses that pose a risk in leading COVID-19 death casualties will enable our community to obtain a better perspective in identifying the most vulnerable populations and enable these populations to implement better precautionary measures.

OBJECTIVE

To identify demographic factors and health diagnosis codes that pose a high or a low risk to COVID-19 death from individual health record data sourced from the United States.

METHODS

We used logistic regression models to analyze the top 500 health diagnosis codes and demographics that have been identified as being associated with COVID-19 death.

RESULTS

Among 223,286 patients tested positive at least once, 218,831 (98%) patients were alive and 4,455 (2%) patients died during the duration of the study period. Through our logistic regression analysis, four demographic characteristics of patients; age, gender, race and region, were deemed to be associated with COVID-19 mortality. Patients from the West region of the United States: Alaska, Arizona, California, Colorado, Hawaii, Idaho, Montana, Nevada, New Mexico, Oregon, Utah, Washington, and Wyoming had the highest odds ratio of COVID-19 mortality across the United States. In terms of diagnoses, Complications mainly related to pregnancy (Adjusted Odds Ratio, OR:2.95; 95% Confidence Interval, CI:1.4 - 6.23) hold the highest odds ratio in influencing COVID-19 death followed by Other diseases of the respiratory system (OR:2.0; CI:1.84 - 2.18), Renal failure (OR:1.76; CI:1.61 - 1.93), Influenza and pneumonia (OR:1.53; CI:1.41 - 1.67), Other bacterial diseases (OR:1.45; CI:1.31 - 1.61), Coagulation defects, purpura and other hemorrhagic conditions(OR:1.37; CI:1.22 - 1.54), Injuries to the head (OR:1.27; CI:1.1 - 1.46), Mood [affective] disorders (OR:1.24; CI:1.12 - 1.36), Aplastic and other anemias (OR:1.22; CI:1.12 - 1.34), Chronic obstructive pulmonary disease and allied conditions (OR:1.18; CI:1.06 - 1.32), Other forms of heart disease (OR:1.18; CI:1.09 - 1.28), Infections of the skin and subcutaneous tissue (OR: 1.15; CI:1.04 - 1.27), Diabetes mellitus (OR:1.14; CI:1.03 - 1.26), and Other diseases of the urinary system (OR:1.12; CI:1.03 - 1.21).

CONCLUSION

We found demographic factors and medical conditions, including some novel ones which are associated with COVID-19 death. These findings can be used for clinical and public awareness and for future research purposes.

Deep neural networks excel in COVID-19 disease severity prediction-a meta-regression analysis

COVID-19 is a disease in which early prognosis of severity is critical for desired patient outcomes and for the management of limited resources like intensive care unit beds and ventilation equipment. Many prognostic statistical tools have been developed for the prediction of disease severity, but it is still unclear which ones should be used in practice. We aim to guide clinicians in choosing the best available tools to make optimal decisions and assess their role in resource management and assess what can be learned from the COVID-19 scenario for development of prediction models in similar medical applications. Using the five major medical databases: MEDLINE (via PubMed), Embase, Cochrane Library (CENTRAL), Cochrane COVID-19 Study Register, and Scopus, we conducted a comprehensive systematic review of prediction tools between 2020 January and 2023 April for hospitalized COVID-19 patients. We identified both the relevant confounding factors of tool performance using the MetaForest algorithm and the best tools-comparing linear, machine learning, and deep learning methods-with mixed-effects meta-regression models. The risk of bias was evaluated using the PROBAST tool. Our systematic search identified eligible 27,312 studies, out of which 290 were eligible for data extraction, reporting on 430 independent evaluations of severity prediction tools with ~ 2.8 million patients. Neural Network-based tools have the highest performance with a pooled AUC of 0.893 (0.748-1.000), 0.752 (0.614-0.853) sensitivity, 0.914 (0.849-0.952) specificity, using clinical, laboratory, and imaging data. The relevant confounders of performance are the geographic region of patients, the rate of severe cases, and the use of C-Reactive Protein as input data. 88% of studies have a high risk of bias, mostly because of deficiencies in the data analysis. All investigated tools in use aid decision-making for COVID-19 severity prediction, but Machine Learning tools, specifically Neural Networks clearly outperform other methods, especially in cases when the basic characteristics of severe and non-severe patient groups are similar, and without the need for more data. When highly specific biomarkers are not available-such as in the case of COVID-19-practitioners should abandon general clinical severity scores and turn to disease specific Machine Learning tools.

External validation of the HEART, HEAR, and HET scores for prediction of major adverse cardiac events in adult patients with acute chest pain

BACKGROUND AND IMPORTANCE

In the cohort of patients presenting to the emergency department (ED) with acute chest pain differentiating between those at high risk of major adverse cardiac event (MACE), and those who can safely be discharged, remains a challenge. The history, ECG, age, risk factors, troponin (HEART) score, as well as several abridged versions [history, ECG, age, risk factors (HEAR), history, ECG, troponin (HET)]. are commonly used for this purpose. As with many clinical risk scores, they might be useful, but often lack proper validation. We aimed to externally validate the HEART, HEAR, and HET scores in the setting of a high-volume tertiary care ED in a healthcare system without gatekeeping functions and thus a low-risk population. We further aimed to compare the prognostic performance (discrimination and calibration) of the scores to each other.

DESIGN

External validation study.

SETTINGS AND PARTICIPANTS

On the basis of a-priori sample size calculations, we prospectively included consecutive adult patients presenting to the ED with acute chest pain.

OUTCOME MEASURES AND ANALYSIS

We assessed overall model performance, discrimination, and calibration of all scores, analyzed reclassification from the HEART score and performed decision curve analysis.

MAIN RESULTS

A total of 3273 patients were included, 383 (12%) suffered MACE within 30 days. Classification differed significantly between scores (HEART: 810; 25% low risk; HET: 55; 2%; HEAR: 195; 6%), as did overall performance (area under the curve: 0.85, 0.80, and 0.79, respectively; P < 0.001). HEART score misclassified 7/810 patients (0.9%; 95% confidence interval: 0.4-1.8%) with MACE as low risk, HET 2/55 (3.6%, 0.9-13.8%), and HEAR 0/195, whereas 2087 (72%), 2837 (98%), and 2695 (93%) patients without MACE were erroneously not classified as low risk.

CONCLUSION

The abridged scores fell short of their results in derivation studies, identifying only very few low-risk patients, and showing inferior model performance compared with the original HEART score. Instead of developing new scores, existing scores should be recalibrated to local population characteristics, as needed.

Machine Learning Based Multi-Parameter Modeling for Prediction of Post-Inflammatory Lung Changes

Objectives: Prediction of lung function deficits following pulmonary infection is challenging and suffers from inaccuracy. We sought to develop machine-learning models for prediction of post-inflammatory lung changes based on COVID-19 recovery data. Methods: In the prospective CovILD study (n = 420 longitudinal observations from n = 140 COVID-19 survivors), data on lung function testing (LFT), chest CT including severity scoring by a human radiologist and density measurement by artificial intelligence, demography, and persistent symptoms were collected. This information was used to develop models of numeric readouts and abnormalities of LFT with four machine learning algorithms (Random Forest, gradient boosted machines, neural network, and support vector machines). Results: Reduced DLCO (diffusion capacity for carbon monoxide <80% of reference) was found in 94 (22%) observations. Those observations were modeled with a cross-validated accuracy of 82-85%, AUC of 0.87-0.9, and Cohen's κ of 0.45-0.5. No reliable models could be established for FEV1 or FVC. For DLCO as a continuous variable, three machine learning algorithms yielded meaningful models with cross-validated mean absolute errors of 11.6-12.5% and R2 of 0.26-0.34. CT-derived features such as opacity, high opacity, and CT severity score were among the most influential predictors of DLCO impairment. Conclusions: Multi-parameter machine learning trained with demographic, clinical, and artificial intelligence chest CT data reliably and reproducibly predicts LFT deficits and outperforms single markers of lung pathology and human radiologist's assessment. It may improve diagnostic and foster personalized treatment.

Computational drug repurposing: approaches, evaluation of in silico resources and case studies

Repurposing of existing drugs for new indications has attracted substantial attention owing to its potential to accelerate drug development and reduce costs. Hundreds of computational resources such as databases and predictive platforms have been developed that can be applied for drug repurposing, making it challenging to select the right resource for a specific drug repurposing project. With the aim of helping to address this challenge, here we overview computational approaches to drug repurposing based on a comprehensive survey of available in silico resources using a purpose-built drug repurposing ontology that classifies the resources into hierarchical categories and provides application-specific information. We also present an expert evaluation of selected resources and three drug repurposing case studies implemented within the Horizon Europe REMEDi4ALL project to demonstrate the practical use of the resources. This comprehensive Review with expert evaluations and case studies provides guidelines and recommendations on the best use of various in silico resources for drug repurposing and establishes a basis for a sustainable and extendable drug repurposing web catalogue.

Comparing conventional and Bayesian workflows for clinical outcome prediction modelling with an exemplar cohort study of severe COVID-19 infection incorporating clinical biomarker test results

PURPOSE

Assessing risk factors and creating prediction models from real-world medical data is challenging, requiring numerous modelling decisions with clinical guidance. Logistic regression is a common model for such studies, for which we advocate the use of Bayesian methods that can jointly deliver probabilistic risk factor inference and prediction. As an exemplar, we compare Bayesian logistic regression with horseshoe priors and Projective Prediction variable selection with the established frequentist LASSO approach, to predict severe COVID-19 outcomes (death or ICU admittance) from demographic and laboratory biomarker data. Our study serves as guidance on data curation, variable selection, and performance assessment with cross-validation.

METHODS

Our source data is based on a retrospective observational cohort design with records from three National Health Service (NHS) Trusts in southwest England, UK. Models were fit to predict severe outcomes within 28 days after admission to hospital (or a positive PCR result if already admitted) using demographic data and the first result from 30 biomarker tests collected within 3 days after admission (or testing positive if already admitted).

RESULTS

Patients included hospitalized adults positive for COVID-19 from March to October 2020, 756 total patients: Mean age 71, 45% female, 31% (n=234) had a severe outcome, of whom 88% (n=206) died. Patients were split into training (n=534) and external validation groups (n=222). Using our Bayesian pipeline, we show a reduced variable model using Age, Urea, Prothrombin time (PT) C-reactive protein (CRP), and Neutrophil-Lymphocyte ratio (NLR) has better predictive performance (median external AUC: 0.71, 95% Quantile [0.7, 0.72]) relative to a GLM using all variables (external AUC: 0.67 [0.63, 0.71]).

CONCLUSION

Urea, PT, CRP, and NLR have been highlighted by other studies, and respectively suggest that hypovolemia, derangement of circulation via clotting, and inflammation are strong predictive risk factors of severity. This study provides guidance on conventional and Bayesian regression and prediction modelling with complex clinical data.

Predicting triage of pediatric patients in the emergency department using machine learning approach

BACKGROUND

The efficient performance of an Emergency Department (ED) relies heavily on an effective triage system that prioritizes patients based on the severity of their medical conditions. Traditional triage systems, including those using the Canadian Triage and Acuity Scale (CTAS), may involve subjective assessments by healthcare providers, leading to potential inconsistencies and delays in patient care.

OBJECTIVE

This study aimed to evaluate six Machine Learning (ML) models K-Nearest Neighbors (KNN), Support Vector Machine (SCM), Decision Tree (DT), Random Forest (RF), Gaussian Naïve Bayes (GNB), and Light GBM (Light Gradient Boosting Machine) for triage prediction in the King Abdulaziz University Hospital using the CTAS framework.

METHODOLOGY

We followed three essential phases: data collection (7125 records of ED patients), data exploration and processing, and the development of machine learning predictive models for ED triage at King Abdulaziz University Hospital.

RESULTS AND CONCLUSION

The overall predictive performance of CTAS was the highest using GNB = 0.984 accuracy. The CTAS-level model performance indicated that SVM, RF, and LGBM achieved the highest performance regarding the consistency of precision and recall values across all CTAS levels.

COVID-19 Readmission Is Highest Among Refugees in Denmark

Vulnerable groups, including certain immigrant populations, have faced higher COVID-19 incidence rates in several countries. This study addresses the gap in knowledge regarding disease severity and readmission odds among refugees, other immigrant groups, and native Danes. Using clinical data from 159 COVID-19-positive patients admitted to hospitals in the Central Denmark Region in 2020, this cross-sectional analysis compared clinical parameters at admission and 30-day readmission odds. The findings revealed no significant differences in clinical status upon admission between groups. Refugees (51.8%) and Others (41.7%) had fewer comorbidities than native Danes (61.2%). Native Danes were more frequently categorized with the highest Charlson Comorbidity Index (CCI) scores. Readmission prevalence was highest among Refugees (23.1%), followed by native Danes (17.0%) and Others (8.3%). After adjusting for age, sex, and CCI, Refugees had a readmission odds ratio (OR) of 1.88 (95% CI, 0.61-5.74) and Others had an OR of 0.61 (95% CI, 0.07-5.41) for readmission compared to native Danes, although this was not statistically significant. This study's significance lies in highlighting the distinct healthcare challenges faced by refugees during the pandemic. Its findings are beneficial for public health policymakers and healthcare professionals seeking to reduce readmission risks and improve COVID-19 outcomes for immigrant populations.

Overlooked and underpowered: a meta-research addressing sample size in radiomics prediction models for binary outcomes

OBJECTIVES

To investigate how studies determine the sample size when developing radiomics prediction models for binary outcomes, and whether the sample size meets the estimates obtained by using established criteria.

METHODS

We identified radiomics studies that were published from 01 January 2023 to 31 December 2023 in seven leading peer-reviewed radiological journals. We reviewed the sample size justification methods, and actual sample size used. We calculated and compared the actual sample size used to the estimates obtained by using three established criteria proposed by Riley et al. We investigated which characteristics factors were associated with the sufficient sample size that meets the estimates obtained by using established criteria proposed by Riley et al. RESULTS: We included 116 studies. Eleven out of one hundred sixteen studies justified the sample size, in which 6/11 performed a priori sample size calculation. The median (first and third quartile, Q1, Q3) of the total sample size is 223 (130, 463), and those of sample size for training are 150 (90, 288). The median (Q1, Q3) difference between total sample size and minimum sample size according to established criteria are -100 (-216, 183), and those differences between total sample size and a more restrictive approach based on established criteria are -268 (-427, -157). The presence of external testing and the specialty of the topic were associated with sufficient sample size.

CONCLUSION

Radiomics studies are often designed without sample size justification, whose sample size may be too small to avoid overfitting. Sample size justification is encouraged when developing a radiomics model.

KEY POINTS

Question Sample size justification is critical to help minimize overfitting in developing a radiomics model, but is overlooked and underpowered in radiomics research. Findings Few of the radiomics models justified, calculated, or reported their sample size, and most of them did not meet the recent formal sample size criteria. Clinical relevance Radiomics models are often designed without sample size justification. Consequently, many models are too small to avoid overfitting. It should be encouraged to justify, perform, and report the considerations on sample size when developing radiomics models.

Artificial intelligence in respiratory pandemics-ready for disease X? A scoping review

OBJECTIVES

This study aims to identify repeated previous shortcomings in medical imaging data collection, curation, and AI-based analysis during the early phase of respiratory pandemics. Based on the results, it seeks to highlight essential steps for improving future pandemic preparedness.

MATERIALS AND METHODS

We searched PubMed/MEDLINE, Scopus, and Cochrane Reviews for articles published from January 1, 2000, to December 31, 2021, using the terms "imaging" or "radiology" or "radiography" or "CT" or "x-ray" combined with "SARS," "MERS," "H1N1," or "COVID-19." WHO and CDC Databases were searched for case definitions.

RESULTS

Over the last 20 years, the world faced several international health emergencies caused by respiratory diseases such as SARS, MERS, H1N1, and COVID-19. During the same period, major technological advances enabled the analysis of vast amounts of imaging data and the continual development of artificial intelligence algorithms to support radiological diagnosis and prognosis. Timely availability of data proved critical, but so far, data collection attempts were initialized only as individual responses to each outbreak, leading to long delays and hampering unified guidelines and data-driven technology to support the management of pandemic outbreaks. Our findings highlight the multifaceted role of imaging in the early stages of SARS, MERS, H1N1, and COVID-19, and outline possible actions for advancing future pandemic preparedness.

CONCLUSIONS

Advancing international cooperation and action on these topics is essential to create a functional, effective, and rapid counteraction system to future respiratory pandemics exploiting state of the art imaging and artificial intelligence.

KEY POINTS

Question What has been the role of radiological data for diagnosis and prognosis in early respiratory pandemics and what challenges were present? Findings International cooperation is essential to developing an effective rapid response system for future respiratory pandemics using advanced imaging and artificial intelligence. Clinical relevance Strengthening global collaboration and leveraging cutting-edge imaging and artificial intelligence are crucial for developing rapid and effective response systems. This approach is essential for improving patient outcomes and managing future respiratory pandemics more effectively.

Severe outcomes of COVID-19 among adults with increased risk conditions: A population-based observational study

BACKGROUND

The individual risk of severe outcomes following COVID-19 is poorly understood in populations with prior immunity. The lack of contemporary estimates limits support of timely diagnosis and antiviral treatment for individuals most likely to benefit.

OBJECTIVE

To determine the risk of severe outcomes following COVID-19 within strata of comorbidities, including patients without documented infection.

DESIGN

Population-based cohort study utilizing electronic medical records and g methods to account for selection bias in the documentation of COVID-19 illnesses.

SETTING

A large health system in northeastern United States.

PATIENTS

Adults with increased risk conditions (90% vaccinated) and COVID-19 from June to December 2022.

MEASUREMENTS

Incidence of composite of inpatient admission within 14 days and death within 28 days of COVID-19 diagnosis.

RESULTS

An estimated 265,248 patients with at least one increased risk condition developed COVID-19, including 76,996 documented cases. Severe outcomes occurred in 3344 (1.3%) patients following COVID-19- 3147 (1.2%) hospitalizations and 376 (0.14%) deaths. In the absence of treatment, individuals with few increased risk conditions (MASS of 3 or less) accounted for 57% of infections and 0.7% developed severe outcomes. In contrast, 2.3% of patients with multiple increased risk conditions (MASS 4 or greater) or severe immunocompromise experienced severe outcomes, including 81% of deaths. The observed risk reduction with antiviral treatment was -0.1% (-0.2 to 0.02%), -0.6% (-0.9 to -0.4%), -1.3% (-2 to -1%), and -1.9% (-3 to -1%) for patients with MASS 3 or less, MASS 4 and 5, MASS 6 or greater, and severe immunocompromise, respectively.

LIMITATIONS

Estimated number COVID-19 cases cannot be directly verified.

CONCLUSIONS

Individuals with multiple medical conditions remain at substantial risk for severe outcomes of COVID-19 and benefit from treatment.

Continuous Monitoring Enables Dynamic Biomarkers to Assess Resilience in Acute COVID-19 Patients

Background/Objectives: The effective management of acute illnesses like COVID-19 requires tools to dynamically assess a patient's resilience to health stressors. This study evaluates novel dynamic biomarkers from continuous blood oxygen saturation (SpO2) monitoring, exploring their association with patient outcomes to support clinical decision making. Methods: We examined 200 hospital admissions from 181 adults diagnosed with COVID-19. Two dynamic biomarkers reflecting the homeostatic regulation efficiency of SpO2 were developed to assess their association with adverse hospital outcomes, specifically ICU admission or death, using binary logistic regressions. The resilience exponent α recorded the relative frequency of prolonged SpO2 declines, while O2 challenges quantified the dynamic response to changes in O2 supplementation. Results: Increased resilience exponent α corresponded to decreased odds of adverse outcomes (OR [95% CI] = 0.59 [0.37-0.93], p = 0.03). Larger SpO2 increases in response to O2 supplementation were associated with increased odds of adverse outcomes (OR [95% CI] = 1.40 [1.04-1.83], p = 0.03). Additionally, the number of O2 supplementation increases (OR [95% CI] = 2.91 [1.90-4.49]) and decreases (OR [95% CI] = 0.33 [0.20-0.55]) during hospitalization were independently linked to poorer and improved outcomes, respectively (both p < 0.001). Conclusions: The resilience exponent α and the O2 challenge response provide insights into the dynamic regulation of SpO2, reflecting physical resilience in COVID-19 patients. Continuous SpO2 monitoring in acute care settings could support more informed clinical decisions during patient management.

The Maastricht Intensive Care COVID Cohort: A Critical Appraisal of the Predefined Research Questions

IMPORTANCE

A review of the study processes and protocols afterward by the researchers themselves is scarce.

OBJECTIVES

The present study aimed to evaluate the study design and the process of data collection of the Maastricht Intensive Care COVID (MaastrICCht) cohort during the COVID-19 pandemic. This evaluation provides information about the quality of the predefined questions and contributes to transparency in science.

DESIGN, SETTING, AND PARTICIPANTS

Critical appraisal of studies using data from the MaastrICCht cohort.

MAIN OUTCOMES AND MEASURES

Evaluation of the process of study design and data collection during the COVID-19 pandemic, focusing on the research process and results.

RESULTS

From March 2020 to April 2023, all patients diagnosed with COVID-19 admitted to the ICU at Maastricht University Medical Center + (n = 544) were included in the MaastrICCht cohort. In total, 37 studies were carried out until April 2024. Fifteen studies addressed 11 of the 13 predetermined research questions, whereas 22 additional studies were performed based on the initial research questions described in the design. Furthermore, 10 studies were conducted with other researchers in national and international collaboration as a response to new arising questions based on evidence that appeared relevant during the pandemic.

CONCLUSIONS AND RELEVANCE

Our critical appraisal indicated that using a study protocol enabled many publications and (inter)national collaborations, although formulating pertinent research questions in the context of a novel disease appeared daunting. Despite this, most questions were successfully addressed, whereas few were resolved by other researchers or lost importance due to the expanding body of knowledge.

Blood neurofilament light chain and S100B as biomarkers of neurological involvement and functional prognosis in COVID-19: a multicenter study

BACKGROUND AND AIM

COVID-19 is associated with neurological complications, termed neuro-COVID, affecting patient outcomes. We aimed to evaluate the association between serum neurofilament light chain (NfL) and S100B biomarkers with the presence of neurological manifestations and functional prognosis in COVID-19 patients.

METHODS

A multicenter prospective cohort study was conducted in three hospitals in the Emilia-Romagna region, Italy, from March 2020 to April 2022. Hospitalized patients with PCR-confirmed COVID-19 were enrolled. Serum S100B and NfL levels were measured in the acute or subacute phase after admission. Diagnostic accuracy was assessed using receiver operating characteristic (ROC) analyses. Statistical analyses were performed to evaluate the association between biomarkers, clinical/laboratory variables, and prognosis, specifically focusing on worsening of the modified Rankin Scale (mRS) from admission to discharge.

RESULTS

A total of 279 patients (153 males, median age 76.7 years) were included. Among them, 69 (24.7%) developed neuro-COVID. Serum NfL levels were significantly higher in the neuro-COVID group (median 110 vs 68.3; p = 0.035) and correlated with severe encephalopathy and extracranial neurologic manifestations. The ROC analysis showed low accuracy in the discrimination between the two groups for both NfL and S100B. Key predictors of worsening mRS included mechanical ventilation (OR = 9.56, 95% CI = 1.67-54.75; p = 0.011), severe encephalopathy (OR = 5.10, 95% CI = 1.58-16.19; p = 0.006), and elevated S100B levels (OR = 2.62, 95% CI = 1.10-6.46; p = 0.037).

CONCLUSIONS

Serum NfL and S100B biomarkers were not accurate in discriminating neuro-COVID patients, however NfL levels were associated with severe and extracranial neuro-COVID, while S100B with functional outcomes, potentially informing clinical management.

The Short and Long of COVID-19: A Review of Acute and Chronic Radiologic Pulmonary Manifestations of SARS-2-CoV and Their Clinical Significance

Coronavirus disease 2019 (COVID-19) pneumonia has had catastrophic effects worldwide. Radiology, in particular computed tomography (CT) imaging, has proven to be valuable in the diagnosis, prognostication, and longitudinal assessment of those diagnosed with COVID-19 pneumonia. This article will review acute and chronic pulmonary radiologic manifestations of COVID-19 pneumonia with an emphasis on CT and also highlighting histopathology, relevant clinical details, and some notable challenges when interpreting the literature.

CLMS: Bridging domain gaps in medical imaging segmentation with source-free continual learning for robust knowledge transfer and adaptation

Deep learning shows promise for medical image segmentation but suffers performance declines when applied to diverse healthcare sites due to data discrepancies among the different sites. Translating deep learning models to new clinical environments is challenging, especially when the original source data used for training is unavailable due to privacy restrictions. Source-free domain adaptation (SFDA) aims to adapt models to new unlabeled target domains without requiring access to the original source data. However, existing SFDA methods face challenges such as error propagation, misalignment of visual and structural features, and inability to preserve source knowledge. This paper introduces Continual Learning Multi-Scale domain adaptation (CLMS), an end-to-end SFDA framework integrating multi-scale reconstruction, continual learning, and style alignment to bridge domain gaps across medical sites using only unlabeled target data or publicly available data. Compared to the current state-of-the-art methods, CLMS consistently and significantly achieved top performance for different tasks, including prostate MRI segmentation (improved Dice of 10.87 %), colonoscopy polyp segmentation (improved Dice of 17.73 %), and plus disease classification from retinal images (improved AUC of 11.19 %). Crucially, CLMS preserved source knowledge for all the tasks, avoiding catastrophic forgetting. CLMS demonstrates a promising solution for translating deep learning models to new clinical imaging domains towards safe, reliable deployment across diverse healthcare settings.

An Explainable Artificial Intelligence Text Classifier for Suicidality Prediction in Youth Crisis Text Line Users: Development and Validation Study

BACKGROUND

Suicide represents a critical public health concern, and machine learning (ML) models offer the potential for identifying at-risk individuals. Recent studies using benchmark datasets and real-world social media data have demonstrated the capability of pretrained large language models in predicting suicidal ideation and behaviors (SIB) in speech and text.

OBJECTIVE

This study aimed to (1) develop and implement ML methods for predicting SIBs in a real-world crisis helpline dataset, using transformer-based pretrained models as a foundation; (2) evaluate, cross-validate, and benchmark the model against traditional text classification approaches; and (3) train an explainable model to highlight relevant risk-associated features.

METHODS

We analyzed chat protocols from adolescents and young adults (aged 14-25 years) seeking assistance from a German crisis helpline. An ML model was developed using a transformer-based language model architecture with pretrained weights and long short-term memory layers. The model predicted suicidal ideation (SI) and advanced suicidal engagement (ASE), as indicated by composite Columbia-Suicide Severity Rating Scale scores. We compared model performance against a classical word-vector-based ML model. We subsequently computed discrimination, calibration, clinical utility, and explainability information using a Shapley Additive Explanations value-based post hoc estimation model.

RESULTS

The dataset comprised 1348 help-seeking encounters (1011 for training and 337 for testing). The transformer-based classifier achieved a macroaveraged area under the curve (AUC) receiver operating characteristic (ROC) of 0.89 (95% CI 0.81-0.91) and an overall accuracy of 0.79 (95% CI 0.73-0.99). This performance surpassed the word-vector-based baseline model (AUC-ROC=0.77, 95% CI 0.64-0.90; accuracy=0.61, 95% CI 0.61-0.80). The transformer model demonstrated excellent prediction for nonsuicidal sessions (AUC-ROC=0.96, 95% CI 0.96-0.99) and good prediction for SI and ASE, with AUC-ROCs of 0.85 (95% CI 0.97-0.86) and 0.87 (95% CI 0.81-0.88), respectively. The Brier Skill Score indicated a 44% improvement in classification performance over the baseline model. The Shapley Additive Explanations model identified language features predictive of SIBs, including self-reference, negation, expressions of low self-esteem, and absolutist language.

CONCLUSIONS

Neural networks using large language model-based transfer learning can accurately identify SI and ASE. The post hoc explainer model revealed language features associated with SI and ASE. Such models may potentially support clinical decision-making in suicide prevention services. Future research should explore multimodal input features and temporal aspects of suicide risk.

Extensive benchmarking of a method that estimates external model performance from limited statistical characteristics

Predictive model performance may deteriorate when applied to data sources that were not used for training, thus, external validation is a key step in successful model deployment. As access to patient-level external data sources is typically limited, we recently proposed a method that estimates external model performance using only external summary statistics. Here, we benchmark the proposed method on multiple tasks using five large heterogeneous US data sources, where each, in turn, plays the role of an internal source and the remaining-external. Results showed accurate estimations for all metrics: 95th error percentiles for the area under the receiver operating characteristics (discrimination), calibration-in-the-large (calibration), Brier and scaled Brier scores (overall accuracy) of 0.03, 0.08, 0.0002, and 0.07, respectively. These results demonstrate the feasibility of estimating the transportability of prediction models using an internal cohort and external statistics. It may become an important accelerator of model deployment.

Risk factors for disease progression and clinical outcomes in patients with COVID-19 in Taiwan: a nationwide population-based cohort study

BACKGROUND

Since 2021, COVID-19 has had a substantial impact on global health and continues to contribute to serious health outcomes. In Taiwan, most research has focused on hospitalized patients or mortality cases, leaving important gaps in understanding the broader effects of the disease and identifying individuals at high risk. This study aims to investigate the risk factors for disease progression through a nationwide population-based cohort study on COVID-19 in Taiwan.

METHODS

This study included 15,056 patients diagnosed with COVID-19 between January 1, 2021, and December 31, 2021, using the Taiwan National Health Insurance Research Database. Baseline and clinical characteristics were collected to verify the association with progression to severity outcomes, including hospital admission, intensive care unit (ICU) admission, invasive ventilatory support, fatal outcome, and the composite outcome of these four events. Patients were observed for 30 days for disease progression. Multivariable logistic regression models were used to calculate odd ratios and 95% confidence intervals (CIs) for each outcome, adjusting for age, sex, region, risk factors, and vaccination status.

RESULTS

Overall, 8,169 patients diagnosed during outpatient visits and 6,887 patients diagnosed during hospitalization were analyzed. Adjusting for age, sex, region, risk factors, and vaccination status, elderly patients had higher risks of hospital admission, ICU admission, invasive ventilatory support, fatal outcome, and composite outcome. Specifically, the risk of the fatal outcome was significantly higher for patients aged 75-84 (odds ratio: 6.11, 95% CI: 4.75-7.87) and those aged 85 years and older (12.70, 9.48-17.02). Patients with cardiovascular disease exhibited higher risks of hospital admission (1.60, 1.31-1.96), ICU admission (1.52, 1.31-1.78), invasive ventilatory support (1.57, 1.26-1.96), and fatal outcomes (1.26, 1.03-1.54) and the composite outcome (1.66, 1.20-1.54). Diabetes mellitus was identified as a significant risk factor for all clinical outcomes (hospital admission: 1.89, 1.53-2.35; ICU admission: 1.53, 1.30-1.79; invasive ventilatory support: 1.27, 1.01-1.60; the composite outcome: 1.45, 1.28-1.66), except for the fatal outcome.

CONCLUSIONS

This study indicated the impact of sex, age, and risk factors on the clinical outcomes of COVID-19 patients in Taiwan. Elderly patients and those with cardiovascular disease or diabetes mellitus had higher risks for severe outcomes, including hospitalization, ICU admission, invasive ventilatory support, and mortality. These findings can provide evidence for a better understanding of risk factors for disease progression and inform targeted intervention.

Role of digital technology in epidemic control: a scoping review on COVID-19 and Ebola

OBJECTIVE

To synthesise the role of digital technologies in epidemic control and prevention, focussing on Ebola and COVID-19.

DESIGN

A scoping review.

DATA SOURCES

A systematic search was done on PubMed, HINARI, Web of Science, Google Scholar and a direct Google search until 10 September 2024.

ELIGIBILITY CRITERIA

We included all qualitative and quantitative studies, conference papers or abstracts, anonymous reports, editorial reports and viewpoints published in English.

DATA EXTRACTION AND SYNTHESIS

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews checklist was used to select the included study. Data analysis was performed using Gale's framework thematic analysis method, resulting in the identification of key themes.

RESULTS

A total of 64 articles that examined the role of digital technology in the Ebola and COVID-19 pandemics were included in the final review. Five main themes emerged: digital epidemiological surveillance (using data visualisation tools and online sources for early disease detection), rapid case identification, community transmission prevention (via digital contact tracing and assessing interventions with mobility data), public education messages and clinical care. The identified barriers encompassed legal, ethical and privacy concerns, as well as organisational and workforce challenges.

CONCLUSION

Digital technologies have proven good for disease prevention and control during pandemics. While the adoption of these technologies has lagged in public health compared with other sectors, tools such as artificial intelligence, telehealth, wearable devices and data analytics offer significant potential to enhance epidemic responses. However, barriers to widespread implementation remain, and investments in digital infrastructure, training and strong data protection are needed to build trust among users. Future efforts should focus on integrating digital solutions into health systems, ensuring equitable access and addressing ethical concerns. As public health increasingly embraces digital innovations, collaboration among stakeholders will be crucial for effective pandemic preparedness and management.

Enhanced Multi-Model Deep Learning for Rapid and Precise Diagnosis of Pulmonary Diseases Using Chest X-Ray Imaging

Background: The global burden of respiratory diseases such as influenza, tuberculosis, and viral pneumonia necessitates rapid, accurate diagnostic tools to improve healthcare responses. Current methods, including RT-PCR and chest radiography, face limitations in accuracy, speed, accessibility, and cost-effectiveness, especially in resource-constrained settings, often delaying treatment and increasing transmission. Methods: This study introduces an Enhanced Multi-Model Deep Learning (EMDL) approach to address these challenges. EMDL integrates an ensemble of five pre-trained deep learning models (VGG-16, VGG-19, ResNet, AlexNet, and GoogleNet) with advanced image preprocessing (histogram equalization and contrast enhancement) and a novel multi-stage feature selection and optimization pipeline (PCA, SelectKBest, Binary Particle Swarm Optimization (BPSO), and Binary Grey Wolf Optimization (BGWO)). Results: Evaluated on two independent chest X-ray datasets, EMDL achieved high accuracy in the multiclass classification of influenza, pneumonia, and tuberculosis. The combined image enhancement and feature optimization strategies significantly improved diagnostic precision and model robustness. Conclusions: The EMDL framework provides a scalable and efficient solution for accurate and accessible pulmonary disease diagnosis, potentially improving treatment efficacy and patient outcomes, particularly in resource-limited settings.

Mitigating bias in AI mortality predictions for minority populations: a transfer learning approach

BACKGROUND

The COVID-19 pandemic has highlighted the crucial role of artificial intelligence (AI) in predicting mortality and guiding healthcare decisions. However, AI models may perpetuate or exacerbate existing health disparities due to demographic biases, particularly affecting racial and ethnic minorities. The objective of this study is to investigate the demographic biases in AI models predicting COVID-19 mortality and to assess the effectiveness of transfer learning in improving model fairness across diverse demographic groups.

METHODS

This retrospective cohort study used a population-based dataset of COVID-19 cases from the Centers for Disease Control and Prevention (CDC), spanning the years 2020-2024. The study analyzed AI model performance across different racial and ethnic groups and employed transfer learning techniques to improve model fairness by adapting pre-trained models to the specific demographic and clinical characteristics of the population.

RESULTS

Decision Tree (DT) and Random Forest (RF) models consistently showed improvements in accuracy, precision, and ROC-AUC scores for Non-Hispanic Black, Hispanic/Latino, and Asian populations. The most significant precision improvement was observed in the DT model for Hispanic/Latino individuals, which increased from 0.3805 to 0.5265. The precision for Asians or Pacific Islanders in the DT model increased from 0.4727 to 0.6071, and for Non-Hispanic Blacks, it rose from 0.5492 to 0.6657. Gradient Boosting Machines (GBM) produced mixed results, showing accuracy and precision improvements for Non-Hispanic Black and Asian groups, but declines for the Hispanic/Latino and American Indian groups, with the most significant decline in precision, which dropped from 0.4612 to 0.2406 in the American Indian group. Logistic Regression (LR) demonstrated minimal changes across all metrics and groups. For the Non-Hispanic American Indian group, most models showed limited benefits, with several performance metrics either remaining stable or declining.

CONCLUSIONS

This study demonstrates the potential of AI in predicting COVID-19 mortality while also underscoring the critical need to address demographic biases. The application of transfer learning significantly improved the predictive performance of models across various racial and ethnic groups, suggesting these techniques are effective in mitigating biases and promoting fairness in AI models.

A systematic review of machine learning applications in predicting opioid associated adverse events

Machine learning has increasingly been applied to predict opioid-related harms due to its ability to handle complex interactions and generating actionable predictions. This review evaluated the types and quality of ML methods in opioid safety research, identifying 44 studies using supervised ML through searches of Ovid MEDLINE, PubMed and SCOPUS databases. Commonly predicted outcomes included postoperative opioid use (n = 15, 34%) opioid overdose (n = 8, 18%), opioid use disorder (n = 8, 18%) and persistent opioid use (n = 5, 11%) with varying definitions. Most studies (96%) originated from North America, with only 7% reporting external validation. Model performance was moderate to strong, but calibration was often missing (41%). Transparent reporting of model development was often incomplete, with key aspects such as calibration, imbalance correction, and handling of missing data absent. Infrequent external validation limited the generalizability of current models. Addressing these aspects is critical for transparency, interpretability, and future implementation of the results.

Risk prediction tools for pressure injury occurrence: an umbrella review of systematic reviews reporting model development and validation methods

BACKGROUND

Pressure injuries (PIs) place a substantial burden on healthcare systems worldwide. Risk stratification of those who are at risk of developing PIs allows preventive interventions to be focused on patients who are at the highest risk. The considerable number of risk assessment scales and prediction models available underscores the need for a thorough evaluation of their development, validation, and clinical utility. Our objectives were to identify and describe available risk prediction tools for PI occurrence, their content and the development and validation methods used.

METHODS

The umbrella review was conducted according to Cochrane guidance. MEDLINE, Embase, CINAHL, EPISTEMONIKOS, Google Scholar, and reference lists were searched to identify relevant systematic reviews. The risk of bias was assessed using adapted AMSTAR-2 criteria. Results were described narratively. All included reviews contributed to building a comprehensive list of risk prediction tools.

RESULTS

We identified 32 eligible systematic reviews only seven of which described the development and validation of risk prediction tools for PI. Nineteen reviews assessed the prognostic accuracy of the tools and 11 assessed clinical effectiveness. Of the seven reviews reporting model development and validation, six included only machine learning models. Two reviews included external validations of models, although only one review reported any details on external validation methods or results. This was also the only review to report measures of both discrimination and calibration. Five reviews presented measures of discrimination, such as the area under the curve (AUC), sensitivities, specificities, F1 scores, and G-means. For the four reviews that assessed the risk of bias assessment using the PROBAST tool, all models but one were found to be at high or unclear risk of bias.

CONCLUSIONS

Available tools do not meet current standards for the development or reporting of risk prediction models. The majority of tools have not been externally validated. Standardised and rigorous approaches to risk prediction model development and validation are needed.

TRIAL REGISTRATION

The protocol was registered on the Open Science Framework ( https://osf.io/tepyk ).

Forecasting cardiovascular disease mortality using artificial neural networks in Sindh, Pakistan

Cardiovascular disease (CVD) is a leading cause of death and disability worldwide, and its incidence and prevalence are increasing in many countries. Modeling of CVD plays a crucial role in understanding the trend of CVD death cases, evaluating the effectiveness of interventions, and predicting future disease trends. This study aims to investigate the modeling and forecasting of CVD mortality, specifically in the Sindh province of Pakistan. The civil hospital in the Nawabshah area of Sindh province, Pakistan, provided the data set used in this study. It is a time series dataset with actual cardiovascular disease (CVD) mortality cases from 1999 to 2021 included. This study analyzes and forecasts the CVD deaths in the Sindh province of Pakistan using classical time series models, including Naïve, Holt-Winters, and Simple Exponential Smoothing (SES), which have been adopted and compared with a machine learning approach called the Artificial Neural Network Auto-Regressive (ANNAR) model. The performance of both the classical time series models and the ANNAR model has been evaluated using key performance indicators such as Root Mean Square Deviation Error, Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). After comparing the results, it was found that the ANNAR model outperformed all the selected models, demonstrating its effectiveness in predicting CVD mortality and quantifying future disease burden in the Sindh province of Pakistan. The study concludes that the ANNAR model is the best-selected model among the competing models for predicting CVD mortality in the Sindh province. This model provides valuable insights into the impact of interventions aimed at reducing CVD and can assist in formulating health policies and allocating economic resources. By accurately forecasting CVD mortality, policymakers can make informed decisions to address this public health issue effectively.

Is Generative AI Increasing the Risk for Technology-Mediated Trauma Among Vulnerable Populations?

The proliferation of Generative Artificial Intelligence (Generative AI) has led to an increased reliance on AI-generated content for designing and deploying digital health interventions. While generative AI has the potential to facilitate and automate healthcare, there are concerns that AI-generated content and AI-generated health advice could trigger, perpetuate, or exacerbate prior traumatic experiences among vulnerable populations. In this discussion article, I examined how generative-AI-powered digital health interventions could trigger, perpetuate, or exacerbate emotional trauma among vulnerable populations who rely on digital health interventions as complementary or alternative sources of seeking health services or information. I then proposed actionable strategies for mitigating AI-generated trauma in the context of digital health interventions. The arguments raised in this article are expected to shift the focus of AI practitioners against prioritizing dominant narratives in AI algorithms into seriously considering the needs of vulnerable minority groups who are at the greatest risk for trauma but are often invisible in AI data sets, AI algorithms, and their resultant technologies.

Epidemiological methods in transition: Minimizing biases in classical and digital approaches

Epidemiology and Public Health have increasingly relied on structured and unstructured data, collected inside and outside of typical health systems, to study, identify, and mitigate diseases at the population level. Focusing on infectious diseases, we review the state of Digital Epidemiology at the beginning of 2020 and how it changed after the COVID-19 pandemic, in both nature and breadth. We argue that Epidemiology's progressive use of data generated outside of clinical and public health systems creates several technical challenges, particularly in carrying specific biases that are almost impossible to correct for a priori. Using a statistical perspective, we discuss how a definition of Digital Epidemiology that emphasizes "data-type" instead of "data-source," may be more operationally useful, by clarifying key methodological differences and gaps. Therefore, we briefly describe some of the possible biases arising from varied collection methods and sources, and offer some recommendations to better explore the potential of Digital Epidemiology, particularly on how to help reduce inequity.

Retinal OCT biomarkers and their association with cognitive function-clinical and AI approaches

Retinal optical coherence tomography (OCT) biomarkers have the potential to serve as early, noninvasive, and cost-effective markers for identifying individuals at risk for cognitive impairments and neurodegenerative diseases. They may also aid in monitoring disease progression and evaluating the effectiveness of interventions targeting cognitive decline. The association between retinal OCT biomarkers and cognitive performance has been demonstrated in several studies, and their importance in cognitive assessment is increasingly being recognized. Machine learning (ML) is a branch of artificial intelligence (AI) with an exponential number of applications in the medical field, particularly its deep learning (DL) subset, which is widely used for the analysis of medical images. These techniques efficiently deal with novel biomarkers when their outcome for the applications of interest is unclear, e.g., for diagnosis, prognosis prediction, disease staging, or any other relevance to clinical practice. However, using AI-based tools for medical purposes must be approached with caution, despite the many efforts to address the black-box nature of such approaches, especially due to the general underperformance in datasets other than those used for their development. Retinal OCT biomarkers are promising as potential indicators for decline in cognitive function. The underlying mechanisms are currently being explored to gain deeper insights into this relationship linking retinal health and cognitive function. Insights from neurovascular coupling and retinal microvascular changes play an important role. Further research is needed to establish the validity and utility of retinal OCT biomarkers as early indicators of cognitive decline and neurodegenerative diseases in routine clinical practice. Retinal OCT biomarkers could then provide a new avenue for early detection, monitoring and intervention in cognitive impairment with the potential to improve patient care and outcomes.

The power of artificial intelligence for managing pandemics: A primer for public health professionals

Artificial intelligence (AI) applications are complex and rapidly evolving, and thus often poorly understood, but have potentially profound implications for public health. We offer a primer for public health professionals that explains some of the key concepts involved and examines how these applications might be used in the response to a future pandemic. They include early outbreak detection, predictive modelling, healthcare management, risk communication, and health surveillance. Artificial intelligence applications, especially predictive algorithms, have the ability to anticipate outbreaks by integrating diverse datasets such as social media, meteorological data, and mobile phone movement data. Artificial intelligence-powered tools can also optimise healthcare delivery by managing the allocation of resources and reducing healthcare workers' exposure to risks. In resource distribution, they can anticipate demand and optimise logistics, while AI-driven robots can minimise physical contact in healthcare settings. Artificial intelligence also shows promise in supporting public health decision-making by simulating the social and economic impacts of different policy interventions. These simulations help policymakers evaluate complex scenarios such as lockdowns and resource allocation. Additionally, it can enhance public health messaging, with AI-generated health communications shown to be more effective than human-generated messages in some cases. However, there are risks, such as privacy concerns, biases in models, and the potential for 'false confirmations', where AI reinforces incorrect decisions. Despite these challenges, we argue that AI will become increasingly important in public health crises, but only if integrated thoughtfully into existing systems and processes.

Diagnosis, Severity, and Prognosis from Potential Biomarkers of COVID-19 in Urine: A Review of Clinical and Omics Results

The COVID-19 pandemic, caused by the SARS-CoV-2 virus, has spurred an extraordinary scientific effort to better understand the disease's pathophysiology and develop diagnostic and prognostic tools to guide more precise and effective clinical management. Among the biological samples analyzed for biomarker identification, urine stands out due to its low risk of infection, non-invasive collection, and suitability for frequent, large-volume sampling. Integrating data from omics studies with standard biochemical analyses offers a deeper and more comprehensive understanding of COVID-19. This review aims to provide a detailed summary of studies published to date that have applied omics and clinical analyses on urine samples to identify potential biomarkers for COVID-19. In July 2024, an advanced search was conducted in Web of Science using the query: "covid* (Topic) AND urine (Topic) AND metabol* (Topic)". The search included results published up to 14 October 2024. The studies retrieved from this digital search were evaluated through a two-step screening process: first by reviewing titles and abstracts for eligibility, and then by retrieving and assessing the full texts of articles that met the specific criteria. The initial search retrieved 913 studies, of which 45 articles were ultimately included in this review. The most robust biomarkers identified include kynurenine, neopterin, total proteins, red blood cells, ACE2, citric acid, ketone bodies, hypoxanthine, amino acids, and glucose. The biological causes underlying these alterations reflect the multisystemic impact of COVID-19, highlighting key processes such as systemic inflammation, renal dysfunction, critical hypoxia, and metabolic stress.

X-Ray Image-Based Real-Time COVID-19 Diagnosis Using Deep Neural Networks (CXR-DNNs)

On 11 February 2020, the prevalent outbreak of COVID-19, a coronavirus illness, was declared a global pandemic. Since then, nearly seven million people have died and over 765 million confirmed cases of COVID-19 have been reported. The goal of this study is to develop a diagnostic tool for detecting COVID-19 infections more efficiently. Currently, the most widely used method is Reverse Transcription Polymerase Chain Reaction (RT-PCR), a clinical technique for infection identification. However, RT-PCR is expensive, has limited sensitivity, and requires specialized medical expertise. One of the major challenges in the rapid diagnosis of COVID-19 is the need for reliable imaging, particularly X-ray imaging. This work takes advantage of artificial intelligence (AI) techniques to enhance diagnostic accuracy by automating the detection of COVID-19 infections from chest X-ray (CXR) images. We obtained and analyzed CXR images from the Kaggle public database (4035 images in total), including cases of COVID-19, viral pneumonia, pulmonary opacity, and healthy controls. By integrating advanced techniques with transfer learning from pre-trained convolutional neural networks (CNNs), specifically InceptionV3, ResNet50, and Xception, we achieved an accuracy of 95%, significantly higher than the 85.5% achieved with ResNet50 alone. Additionally, our proposed method, CXR-DNNs, can accurately distinguish between three different types of chest X-ray images for the first time. This computer-assisted diagnostic tool has the potential to significantly enhance the speed and accuracy of COVID-19 diagnoses.

Validation of prognostic models predicting mortality or ICU admission in patients with COVID-19 in low- and middle-income countries: a global individual participant data meta-analysis

BACKGROUND

We evaluated the performance of prognostic models for predicting mortality or ICU admission in hospitalized patients with COVID-19 in the World Health Organization (WHO) Global Clinical Platform, a repository of individual-level clinical data of patients hospitalized with COVID-19, including in low- and middle-income countries (LMICs).

METHODS

We identified eligible multivariable prognostic models for predicting overall mortality and ICU admission during hospital stay in patients with confirmed or suspected COVID-19 from a living review of COVID-19 prediction models. These models were evaluated using data contributed to the WHO Global Clinical Platform for COVID-19 from nine LMICs (Burkina Faso, Cameroon, Democratic Republic of Congo, Guinea, India, Niger, Nigeria, Zambia, and Zimbabwe). Model performance was assessed in terms of discrimination and calibration.

RESULTS

Out of 144 eligible models, 140 were excluded due to a high risk of bias, predictors unavailable in LIMCs, or insufficient model description. Among 11,338 participants, the remaining models showed good discrimination for predicting in-hospital mortality (3 models), with areas under the curve (AUCs) ranging between 0.76 (95% CI 0.71-0.81) and 0.84 (95% CI 0.77-0.89). An AUC of 0.74 (95% CI 0.70-0.78) was found for predicting ICU admission risk (one model). All models showed signs of miscalibration and overfitting, with extensive heterogeneity between countries.

CONCLUSIONS

Among the available COVID-19 prognostic models, only a few could be validated on data collected from LMICs, mainly due to limited predictor availability. Despite their discriminative ability, selected models for mortality prediction or ICU admission showed varying and suboptimal calibration.

The OpenSAFELY Collaborative, Williamson EJ, Tazare J, Bhaskaran K, Walker AJ, McDonald HI, Tomlinson LA, Bacon S, Bates C, Curtis HJ, Forbes H, Minassian C, Morton CE, Nightingale E, Mehrkar A, Evans D, Nicholson BD, Leon D, Inglesby P, MacKenna B, Cockburn J, Davies NG, Hulme WJ, Morley J, Douglas IJ, Rentsch CT, Mathur R, Wong A, Schultze A, Croker R, Parry J, Hester F, Harper S, Perera R, Grieve R, Harrison D, Steyerberg E, Eggo RM, Diaz-Ordaz K, Keogh R, Evans SJ, Smeeth L, Goldacre B. Study protocol: Comparison of different risk prediction modelling approaches for COVID-19 related death using the OpenSAFELY platform

On March 11th 2020, the World Health Organization characterised COVID-19 as a pandemic. Responses to containing the spread of the virus have relied heavily on policies involving restricting contact between people. Evolving policies regarding shielding and individual choices about restricting social contact will rely heavily on perceived risk of poor outcomes from COVID-19. In order to make informed decisions, both individual and collective, good predictive models are required. For outcomes related to an infectious disease, the performance of any risk prediction model will depend heavily on the underlying prevalence of infection in the population of interest. Incorporating measures of how this changes over time may result in important improvements in prediction model performance. This protocol reports details of a planned study to explore the extent to which incorporating time-varying measures of infection burden over time improves the quality of risk prediction models for COVID-19 death in a large population of adult patients in England. To achieve this aim, we will compare the performance of different modelling approaches to risk prediction, including static cohort approaches typically used in chronic disease settings and landmarking approaches incorporating time-varying measures of infection prevalence and policy change, using COVID-19 related deaths data linked to longitudinal primary care electronic health records data within the OpenSAFELY secure analytics platform.

Artificial Intelligence in the Management of Patients with Respiratory Failure Requiring Mechanical Ventilation: A Scoping Review

Background: Mechanical ventilation (MV) is one of the most frequently used organ replacement modalities in the intensive care unit (ICU). Artificial intelligence (AI) presents substantial potential in optimizing mechanical ventilation management. The utility of AI in MV lies in its ability to harness extensive data from electronic monitoring systems, facilitating personalized care tailored to individual patient needs. This scoping review aimed to consolidate and evaluate the existing evidence for the application of AI in managing respiratory failure among patients necessitating MV. Methods: The literature search was conducted in PubMed, Scopus, and the Cochrane Library. Studies investigating the utilization of AI in patients undergoing MV, including observational and randomized controlled trials, were selected. Results: Overall, 152 articles were screened, and 37 were included in the analysis. We categorized the goals of AI in the included studies into the following groups: (1) prediction of requirement in MV; (2) prediction of outcomes in MV; (3) prediction of weaning from MV; (4) prediction of hypoxemia after extubation; (5) prediction models for MV-associated severe acute kidney injury; (6) identification of long-term outcomes after prolonged MV; (7) prediction of survival. Conclusions: AI has been studied in a wide variety of patients with respiratory failure requiring MV. Common applications of AI in MV included the assessment of the performance of ML for mortality prediction in patients with respiratory failure, prediction and identification of the most appropriate time for extubation, detection of patient-ventilator asynchrony, ineffective expiration, and the prediction of the severity of the respiratory failure.