AI and Dynamic Prediction of Deterioration in Covid-19

During the COVID-19 pandemic it became evident that outcome prediction of patients is crucial for triaging, when resources are limited and enable early start or increase of available therapeutic support. COVID-19 demographic risk factors for severe disease and death were rapidly established, including age and sex. Common Clinical Decision Support Systems (CDSS) and Early Warning Systems (EWS) have been used to triage based on demographics, vital signs and laboratory results. However, all of these have limitations, such as dependency of laboratory investigations or set threshold values, were derived from more or less specific cohort studies. Instead, individual illness dynamics and patterns of recovery might be essential characteristics in understanding the critical course of illness.The pandemic has been a game changer for data, and the concept of real-time massive health data has emerged as one of the important tools in battling the pandemic. We here describe the advantages and limitations of established risk scoring systems and show how artificial intelligence applied on dynamic vital parameter changes, may help to predict critical illness, adverse events and death in patients hospitalized with COVID-19.Machine learning assisted dynamic analysis can improve and give patient-specific prediction in Clinical Decision Support systems that have the potential of reducing both morbidity and mortality. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2022.

Artificial intelligence and machine learning in clinical trial design and application

In 2020, COVID-19 changed how health care was approached both in the United States and globally. In the early phases, the vast majority of energy and attention was devoted to containing the pandemic and treating the infected. Toward the end of 2020, that attention expanded to vaccinating people across the globe. What was not being considered at the time were challenges to future health and clinical trials that power new treatments for COVID-19 and non-COVID-19 treatments. © 2023 Elsevier Inc. All rights reserved.

PDSS: A Pharmacological Decision Support System for Diabetics Patients with COVID-19.

With the advent of SARS-CoV-2, several studies have shown that there is a higher mortality rate in patients with diabetes and, in some cases, it is one of the side effects of overcoming the disease. However, there is no clinical decision support tool or specific treatment protocols for these patients. To tackle this issue, in this paper we present a Pharmacological Decision Support System (PDSS) providing intelligent decision support for COVID-19 diabetic patient treatment selection, based on an analysis of risk factors with data from electronic medical records using Cox regression. The goal of the system is to create real world evidence including the ability to continuously learn to improve clinical practice and outcomes of diabetic patients with COVID-19.

Development and validation of multivariable prediction models of serological response to SARS-CoV-2 vaccination in kidney transplant recipients.

Repeated vaccination against SARS-CoV-2 increases serological response in kidney transplant recipients (KTR) with high interindividual variability. No decision support tool exists to predict SARS-CoV-2 vaccination response to third or fourth vaccination in KTR. We developed, internally and externally validated five different multivariable prediction models of serological response after the third and fourth vaccine dose against SARS-CoV-2 in previously seronegative, COVID-19-naïve KTR. Using 20 candidate predictor variables, we applied statistical and machine learning approaches including logistic regression (LR), least absolute shrinkage and selection operator (LASSO)-regularized LR, random forest, and gradient boosted regression trees. For development and internal validation, data from 590 vaccinations were used. External validation was performed in four independent, international validation cohorts comprising 191, 184, 254, and 323 vaccinations, respectively. LASSO-regularized LR performed on the whole development dataset yielded a 20- and 10-variable model, respectively. External validation showed AUC-ROC of 0.840, 0.741, 0.816, and 0.783 for the sparser 10-variable model, yielding an overall performance 0.812. A 10-variable LASSO-regularized LR model predicts vaccination response in KTR with good overall accuracy. Implemented as an online tool, it can guide decisions whether to modulate immunosuppressive therapy before additional active vaccination, or to perform passive immunization to improve protection against COVID-19 in previously seronegative, COVID-19-naïve KTR.

Improving guideline-concordant thromboprophylaxis prescribing for children admitted to hospital with COVID-19.

BACKGROUND: The incidence of venous thrombo-embolism (VTE) in hospitalized children has increased by 130%-200% over the last two decades. Given this increase, many centers utilize electronic clinical decision support (CDS) to prognosticate VTE risk and recommend prophylaxis. SARS-CoV-2 infection (COVID-19) is a risk factor for VTE; however, CDS developed before the COVID-19 pandemic may not accurately prognosticate VTE risk in children with COVID-19. This study's objective was to identify areas to improve thromboprophylaxis recommendations for children with COVID-19. METHODS: Inpatients with a positive COVID-19 test at admission were identified at a quaternary-care pediatric center between March 1, 2020 and January 20, 2022. The results of the institution's automated CDS thromboprophylaxis recommendations were compared to institutional COVID-19 thromboprophylaxis guidelines and to the actual thromboprophylaxis received. CDS optimization was performed to improve adherence to COVID-19 thromboprophylaxis recommendations. RESULTS: Of the 329 patients included in this study, 106 (28.2%) were prescribed pharmaco-prophylaxis, 167 (50.8%) were identified by the institutional COVID-19 guidelines as requiring pharmaco-prophylaxis, and 45 (13.2%) were identified by the CDS as needing pharmaco-prophylaxis. On univariate analysis, only age 12 years or more was associated with recipient of appropriate prophylaxis (OR 1.78, 95% CI: 1.13-2.82, p = .013). Five patients developed VTEs; three had symptoms at presentation, two were identified as high risk for VTE by both the automated and best practice assessments but were not prescribed pharmaco-prophylaxis. CONCLUSION: Automated thromboprophylaxis recommendations developed prior to the COVID-19 pandemic may not identify all COVID-19 patients needing pharmaco-prophylaxis. Existing CDS tools need to be updated to reflect COVID-19-specific risk factors for VTEs.

Federated Learning for Healthcare Domain - Pipeline, Applications and Challenges

Federated learning is the process of developing machine learning models over datasets distributed across data centers such as hospitals, clinical research labs, and mobile devices while preventing data leakage. This survey examines previous research and studies on federated learning in the healthcare sector across a range of use cases and applications. Our survey shows what challenges, methods, and applications a practitioner should be aware of in the topic of federated learning. This paper aims to lay out existing research and list the possibilities of federated learning for healthcare industries.© 2022 Copyright held by the owner/author(s).

Predicting COVID-19 Severity Integrating RNA-Seq Data Using Machine Learning Techniques

A fundamental challenge in the fight against COVID-19 is the development of reliable and accurate tools to predict disease progression in a patient. This information can be extremely useful in distinguishing hospitalized patients at higher risk for needing UCI from patients with low severity. How SARS-CoV-2 infection will evolve is still unclear. Method(s): A novel pipeline was developed that can integrate RNA-Seq data from different databases to obtain a genetic biomarker COVID-19 severity index using an artificial intelligence algorithm. Our pipeline ensures robustness through multiple cross-validation processes in different steps. Result(s): CD93, RPS24, PSCA, and CD300E were identified as COVID-19 severity gene signatures. Furthermore, using the obtained gene signature, an effective multi-class classifier capable of discrimi-nating between control, outpatient, inpatient, and ICU COVID-19 patients was optimized, achieving an accuracy of 97.5%. Conclusion(s): In summary, during this research, a new intelligent pipeline was implemented to develop a specific gene signature that can detect the severity of patients suffering COVID-19. Our approach to clinical decision support systems achieved excellent results, even when processing unseen samples. Our system can be of great clinical utility for the strategy of planning, organizing and managing human and material resources, as well as for automatically classifying the severity of patients affected by COVID-19.Copyright © 2023 Bentham Science Publishers.

PiMS: A Pre-ML Labelling Tool

Machine Learning (ML) techniques in clinical decision support systems are scarce due to the limited availability of clinically validated and labelled training data sets. We present a framework to (1) enable quality controls at data submission toward ML appropriate data, (2) provide in-situ algorithm assessments, and (3) prepare dataframes for ML training and robust stochastic analysis. We developed and evaluated PiMS (Pandemic Intervention and Monitoring Systems): a remote monitoring solution for patients that are Covid-positive. The system was trialled at two hospitals in Melbourne, Australia (Alfred Health and Monash Health) involving 109 patients and 15 clinicians.

Decision support systems in healthcare: systematic review, meta- analysis and prediction, with example of COVID-19

Objective: The objective of this study was to provide an overview of Decision Support Systems (DSS) applied in healthcare used for diagnosis, monitoring, prediction and recommendation in medicine. Methods: We conducted a systematic review using PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) guidelines of articles published until September 2022 from PubMed, Cochrane, Scopus and web of science databases. We used KH coder to analyze included research. Then we categorized decision support systems based on their types and medical applications. Results: The search strategy provided a total of 1605 articles in the studied period. Of these, 231 articles were included in this qualitative review. This research was classified into 4 categories based on the DSS type used in healthcare: Alert Systems, Monitoring Systems, Recommendation Systems and Prediction Systems. Under each category, domain applications were specified according to the disease the system was applied to. Conclusion: In this systematic review, we collected CDSS studies that use ML techniques to provide insights into different CDSS types. We highlighted the importance of ML to support physicians in clinical decision-making and improving healthcare according to their purposes.

Ensemble Image Explainable AI (XAI) Algorithm for Severe Community-Acquired Pneumonia and COVID-19 Respiratory Infections

Since the onset of the COVID-19 pandemic in 2019, many clinical prognostic scoring tools have been proposed or developed to aid clinicians in the disposition and severity assessment of pneumonia. However, there is limited work that focuses on explaining techniques that are best suited for clinicians in their decision making. In this article, we present a new image explainability method named ensemble AI explainability (XAI), which is based on the SHAP and Grad-CAM++ methods. It provides a visual explanation for a deep learning prognostic model that predicts the mortality risk of community-acquired pneumonia and COVID-19 respiratory infected patients. In addition, we surveyed the existing literature and compiled prevailing quantitative and qualitative metrics to systematically review the efficacy of ensemble XAI, and to make comparisons with several state-of-the-art explainability methods (LIME, SHAP, saliency map, Grad-CAM, Grad-CAM++). Our quantitative experimental results have shown that ensemble XAI has a comparable absence impact (decision impact: 0.72, confident impact: 0.24). Our qualitative experiment, in which a panel of three radiologists were involved to evaluate the degree of concordance and trust in the algorithms, has showed that ensemble XAI has localization effectiveness (mean set accordance precision: 0.52, mean set accordance recall: 0.57, mean set F1: 0.50, mean set IOU: 0.36) and is the most trusted method by the panel of radiologists (mean vote: 70.2%). Finally, the deep learning interpretation dashboard used for the radiologist panel voting will be made available to the community. Our code is available at https://github.com/IHIS-HealthInsights/Interpretation-Methods-Voting-dashboard. © 2020 IEEE.

A decision support system for selecting the most suitable machine learning in healthcare using user parameters and requirements

The application of machine learning in the medical field is still limited. The main reason behind the lack of use is the unavailability of an easy-to-use machine learning system that targets non-technical users. The objective of this paper is to propose an automated machine learning system to aid non-technical users. The proposed system provides the user with simple choices to provide suggestions to the system. The system uses the combination of the user's choices and performance evaluation to select the most suited model from available options. In this study, we employed the system on a Parkinson's disease dataset. The templates for support vector machine and random forest algorithms are provided to the system. Support vector machines and random forests were able to produce 80% and 75% accuracy, respectively. The system used performance parameters of the system and user choices to select the most suited models for each test case. The support vector machine was selected as the most suited model in three test cases, while random forest was selected as the most suited for one test case. The test cases also showed that the weighted time parameter impacted the results heavily.Copyright © 2022 The Author(s)

Getting to knowledge-enhanced health and healthcare

In this chapter, we describe advances in technology capabilities and in the healthcare ecosystem that are driving breakthrough innovations in clinical decision support (CDS). New and enhanced capabilities include the rise of precision medicine, patient/user engagement, aggregation of data for population health, onset of an "app culture,” artificial intelligence, and interoperability and standards development and adoption. Advances in healthcare delivery include new demands, stimuli, and incentives for CDS brought about by the factors such as broad adoption of electronic health record (EHR) systems, national programs such as the US Meaningful Use EHR certification, value-driven healthcare delivery and financing models, quality monitoring and reporting, as well as remarkable sociotechnical shifts during the COVID-19 pandemic. This chapter discusses the impact of these changes on the CDS landscape along with unique possibilities for CDS moving forward. © 2023 Elsevier Inc. All rights reserved.

Population health management

Population health management (PHM) is a systematic approach that uses information technology and digital health tools to improve health and healthcare at the population-level. PHM programs identify individuals who could benefit from a set of PHM interventions;implement computable logic to stratify patients according to risk;and implement protocol-based logic to assign individuals within each stratum to specific interventions. PHM is a promising approach to help achieve the Quintuple Aim of healthcare: (i) improving population health through population-level interventions;(ii) enhancing the care experience by shifting healthcare from the clinic to the patient's home;(iii) reducing costs by focusing on health promotion and prevention;(iv) improving the work life of the health care workforce by reducing clinic workload;and (v) advancing health equity by maximizing reach through a combination of digital and human-based patient outreach interventions. This chapter discusses the components of a technical infrastructure to support PHM, including data sources (registries, electronic health records), data analytics tools, patient outreach and engagement tools, and patient tracking dashboards. We also describe real-world examples of PHM programs focused on chronic disease management, genetic testing for hereditary cancers, colorectal cancer screening, COVID-19 testing and vaccination, and tobacco cessation. PHM is expected to experience substantial growth with novel digital health technologies, such as sensors, phone apps, conversational agents, and virtual reality;artificial intelligence;and new data sources. © 2023 Elsevier Inc. All rights reserved.

Implementing patient decision aids into general practice clinical decision support systems: Feasibility study in cardiovascular disease prevention.

Objective: Patient decision aids (DA) facilitate shared decision making, but implementation remains a challenge. This study tested the feasibility of integrating a cardiovascular disease (CVD) prevention DA into general practice software. Methods: We developed a desktop computer application (app) to auto-populate a CVD prevention DA from general practice medical records. 4 practices received monthly practice reports from July-Nov 2021, and 2 practices use the app with limited engagement. CVD risk assessment data and app use were monitored. Results: The proportion of eligible patients with complete CVD risk assessment data ranged from 59 to 94%. Monthly app use ranged from 0 to 285 sessions by 13 individual practice staff including GPs and nurses, with staff using the app an average of 67 sessions during the study period. High users in the 5-month study period continued to use the app for 10 months. Low use was attributed to reduced staff capacity during COVID-19 and technical issues. Conclusion: High users sustained interest in the app, but additional strategies are required for low users. The study will inform implementation plans for new guidelines. Innovation: This study showed it is feasible to integrate patient decision aids with Australian general practice software, despite the challenges of COVID-19 at the time of the study.

Lessons learned from developing a COVID-19 algorithm governance framework in Aotearoa New Zealand

Aotearoa New Zealand's response to the COVID-19 pandemic has included the use of algorithms that could aid decision making. Te Pokapū Hātepe o Aotearoa, the New Zealand Algorithm Hub, was established to evaluate and host COVID-19 related models and algorithms, and provide a central and secure infrastructure to support the country's pandemic response. A critical aspect of the Hub was the formation of an appropriate governance group to ensure that algorithms being deployed underwent cross-disciplinary scrutiny prior to being made available for quick and safe implementation. This framework necessarily canvassed a broad range of perspectives, including from data science, clinical, Māori, consumer, ethical, public health, privacy, legal and governmental perspectives. To our knowledge, this is the first implementation of national algorithm governance of this type, building upon broad local and global discussion of guidelines in recent years. This paper describes the experiences and lessons learned through this process from the perspective of governance group members, emphasising the role of robust governance processes in building a high-trust platform that enables rapid translation of algorithms from research to practice.

Deep learning-based hybrid clinical decision support system algorithm for COVID-19 diagnosis via PCR graphics and Thorax CT images, preliminary data

Introduction: The limited sensitivity of microbiological testing, challenges in radiological differential diagnosis, and expectations of quick and accurate diagnosis required developing clinical decision support systems (CDSS). We propose a new deep learning-based hybrid CDSS that combines the advantageous aspects of thorax computed tomography(CT) and reverse transcriptase-polymerase chain reaction(PCR) to overcome the weakness of each one. Method(s): We retrospectively constructed a database that contains CT images of healthy subjects and patients with COVID-19 pneumonia(CP), bacterial/viral pneumonia(BVP), interstitial lung diseases(ILD), and PCR data of patients who were tested positive and negative for SARS-CoV-2. A new 3D-convolutional neural network (3D-CNN) and long short-term memory network(LSTM) based CDSS is developed to perform accurate and robust detection of COVID19 using CT images and PCR data. Result(s): Performance results of the proposed models (Fig1) provide highly reliable diagnosis of COVID-19 with 93.2% and 99.7% AUC for CT and PCR data, respectively. Conclusion(s): Proposed CDSS with state-of-the-art deep learning methods provides similar performance compared to both radiologists in CT evaluation and microbiologists in PCR evaluation and can be safely used. We plan to develop a hybrid CDSS algorithm further, combining laboratory data with CT and PCR models.

Automated Monitoring of Adherence to Evidenced-Based Clinical Guideline Recommendations: Design and Implementation Study.

BACKGROUND: Clinical practice guidelines are systematically developed statements intended to optimize patient care. However, a gapless implementation of guideline recommendations requires health care personnel not only to be aware of the recommendations and to support their content but also to recognize every situation in which they are applicable. To not miss situations in which recommendations should be applied, computerized clinical decision support can be provided through a system that allows an automated monitoring of adherence to clinical guideline recommendations in individual patients. OBJECTIVE: This study aims to collect and analyze the requirements for a system that allows the monitoring of adherence to evidence-based clinical guideline recommendations in individual patients and, based on these requirements, to design and implement a software prototype that integrates guideline recommendations with individual patient data, and to demonstrate the prototype's utility in treatment recommendations. METHODS: We performed a work process analysis with experienced intensive care clinicians to develop a conceptual model of how to support guideline adherence monitoring in clinical routine and identified which steps in the model could be supported electronically. We then identified the core requirements of a software system to support recommendation adherence monitoring in a consensus-based requirements analysis within the loosely structured focus group work of key stakeholders (clinicians, guideline developers, health data engineers, and software developers). On the basis of these requirements, we designed and implemented a modular system architecture. To demonstrate its utility, we applied the prototype to monitor adherence to a COVID-19 treatment recommendation using clinical data from a large European university hospital. RESULTS: We designed a system that integrates guideline recommendations with real-time clinical data to evaluate individual guideline recommendation adherence and developed a functional prototype. The needs analysis with clinical staff resulted in a flowchart describing the work process of how adherence to recommendations should be monitored. Four core requirements were identified: the ability to decide whether a recommendation is applicable and implemented for a specific patient, the ability to integrate clinical data from different data formats and data structures, the ability to display raw patient data, and the use of a Fast Healthcare Interoperability Resources-based format for the representation of clinical practice guidelines to provide an interoperable, standards-based guideline recommendation exchange format. CONCLUSIONS: Our system has advantages in terms of individual patient treatment and quality management in hospitals. However, further studies are needed to measure its impact on patient outcomes and evaluate its resource effectiveness in different clinical settings. We specified a modular software architecture that allows experts from different fields to work independently and focus on their area of expertise. We have released the source code of our system under an open-source license and invite for collaborative further development of the system.

De-implementing opioid prescribing in a dental group practice: Lessons learned.

BACKGROUND: Drug overdose has become a leading cause of accidental death in the United States. Between 2000 and 2015, the rate of deaths from drug overdoses increased 137%, including a 200% increase in the rate of overdose deaths involving opioids (including opioid pain relievers and heroin). Unnecessary opioid prescribing is one of the factors driving this epidemic. OBJECTIVES: The primary objective of this paper is to share lessons learned while conducting a randomized trial to de-implement opioids for post-extraction pain management utilizing clinical decision support (CDS) with and without patient education. The lessons learned from conducting this trial in a real-world setting can be applied to future dissemination and implementation oral health research. METHODS: The sources informing lessons learned were generated from qualitative interviews conducted with 20 of the forty-nine dental providers involved in the study following the implementation phase of the trial. Ongoing policy, social and environmental factors were tracked throughout the study. RESULTS: Dental providers in the trial identified the impact of training that involved health professionals sharing information about the personal impact of pain and opioid use. Additionally, they found utility in being presented with a dashboard detailing their prescribing patterns related to other dentists. For the 30 general dentists with access to the CDS, use of its portal varied widely, with most using it 10%-49% of the time related to extractions. CONCLUSIONS: In the context of a downward trend in opioid prescribing and considering the influence of the COVID pandemic during the trial, dental providers indicated benefit in training about negative personal impacts of prescribing opioids, and personally relevant feedback about their prescribing patterns. Only modest use of the CDS was realized. Implementation of this trial was impacted by governmental and health system policies and the COVID pandemic, prompt the consideration of implications regarding continuing ways to limit opioid prescribing among dental providers.

Assessing the effects of data drift on the performance of machine learning models used in clinical sepsis prediction.

BACKGROUND: Data drift can negatively impact the performance of machine learning algorithms (MLAs) that were trained on historical data. As such, MLAs should be continuously monitored and tuned to overcome the systematic changes that occur in the distribution of data. In this paper, we study the extent of data drift and provide insights about its characteristics for sepsis onset prediction. This study will help elucidate the nature of data drift for prediction of sepsis and similar diseases. This may aid with the development of more effective patient monitoring systems that can stratify risk for dynamic disease states in hospitals. METHODS: We devise a series of simulations that measure the effects of data drift in patients with sepsis, using electronic health records (EHR). We simulate multiple scenarios in which data drift may occur, namely the change in the distribution of the predictor variables (covariate shift), the change in the statistical relationship between the predictors and the target (concept shift), and the occurrence of a major healthcare event (major event) such as the COVID-19 pandemic. We measure the impact of data drift on model performances, identify the circumstances that necessitate model retraining, and compare the effects of different retraining methodologies and model architecture on the outcomes. We present the results for two different MLAs, eXtreme Gradient Boosting (XGB) and Recurrent Neural Network (RNN). RESULTS: Our results show that the properly retrained XGB models outperform the baseline models in all simulation scenarios, hence signifying the existence of data drift. In the major event scenario, the area under the receiver operating characteristic curve (AUROC) at the end of the simulation period is 0.811 for the baseline XGB model and 0.868 for the retrained XGB model. In the covariate shift scenario, the AUROC at the end of the simulation period for the baseline and retrained XGB models is 0.853 and 0.874 respectively. In the concept shift scenario and under the mixed labeling method, the retrained XGB models perform worse than the baseline model for most simulation steps. However, under the full relabeling method, the AUROC at the end of the simulation period for the baseline and retrained XGB models is 0.852 and 0.877 respectively. The results for the RNN models were mixed, suggesting that retraining based on a fixed network architecture may be inadequate for an RNN. We also present the results in the form of other performance metrics such as the ratio of observed to expected probabilities (calibration) and the normalized rate of positive predictive values (PPV) by prevalence, referred to as lift, at a sensitivity of 0.8. CONCLUSION: Our simulations reveal that retraining periods of a couple of months or using several thousand patients are likely to be adequate to monitor machine learning models that predict sepsis. This indicates that a machine learning system for sepsis prediction will probably need less infrastructure for performance monitoring and retraining compared to other applications in which data drift is more frequent and continuous. Our results also show that in the event of a concept shift, a full overhaul of the sepsis prediction model may be necessary because it indicates a discrete change in the definition of sepsis labels, and mixing the labels for the sake of incremental training may not produce the desired results.