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.

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.

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.

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.

A type-2 fuzzy rule-based model for diagnosis of COVID-19

In this study, a type-2 fuzzy logic-based decision support system comprising clinical examination and blood test results that health professionals can use in addition to existing methods in the diagnosis of COVID-19 has been developed. The developed system consists of three fuzzy units. The first fuzzy unit produces COVID-19 positivity as a percentage according to the respiratory rate, loss of smell, and body temperature values, and the second fuzzy unit according to the C-reactive protein, lymphocyte, and D-dimer values obtained as a result of the blood tests. In the third fuzzy unit, the COVID-19 positivity risks according to the clinical examination and blood analysis results, which are the outputs of the first and second fuzzy units, are evaluated together and the result is obtained. As a result of the evaluation of the trials with 60 different scenarios by physicians, it has been revealed that the system can detect COVID-19 risk with 86.6% accuracy. [ FROM AUTHOR]

Identification of Drug-Related Problems and Investigation of Related Factors in Patients with COVID-19: An Observational Study

Objective: Clinical prognosis of coronavirus disease-19 (COVID-19) may be severe and unexpected. Patients may quickly progress to respiratory failure, infections, multiple organ dysfunction, and sepsis. The main objective of this study is to investigate the drug-related problems of patients with COVID-19 and related factors. Method: A prospective observational study was conducted on patients with COVID-19 between September 2020 and May 2021. Patients' demographics, comorbid diseases, prescribed medicines and laboratory findings were recorded. Drug-related problems (DRPs) were identified by a clinical pharmacist according to recent guidelines, UpToDate® clinical decision support system and evidence-based medicine. Results: The median age of 107 patients was 64 and 50.46% of them were male. The median number of comorbidities was 3 (2-4) per patient. The majority of the patients had at least one comorbidity (88.79%) other than COVID-19 and the most frequent comorbidities were hypertension, diabetes mellitus and coronary artery disease. The total number of DRPs was recorded as 201 and at least one DRP was seen in 75 out of 107 patients. The median number of DRPs was 2 (0-8). In multivariate model, number of comorbidities (odss ratio (OR)=1.952;95% confidence interval (CI)=1.07-3.54, p<0.05, number of medications (OR=1.344;95% CI=1.12-1.61, p<0.001), and serum potassium levels (OR=5.252;95% CI=1.57-17.56, p<0.001) were the factors related with DRP. Conclusion: This study highlights the DRPs and related factors in patients with COVID 19 in hospital settings. Considering unknown features of the infection and multiple medication use, DRPs are likely to occur. It would be beneficial to consider the related factors in order to reduce the number of the DRPs. (English) [ FROM AUTHOR]

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. © 2022 IEEE.

iApp: An Autonomous Inspection, Auscultation, Percussion, and Palpation Platform.

Disease symptoms often contain features that are not routinely recognized by patients but can be identified through indirect inspection or diagnosis by medical professionals. Telemedicine requires sufficient information for aiding doctors' diagnosis, and it has been primarily achieved by clinical decision support systems (CDSSs) utilizing visual information. However, additional medical diagnostic tools are needed for improving CDSSs. Moreover, since the COVID-19 pandemic, telemedicine has garnered increasing attention, and basic diagnostic tools (e.g., classical examination) have become the most important components of a comprehensive framework. This study proposes a conceptual system, iApp, that can collect and analyze quantified data based on an automatically performed inspection, auscultation, percussion, and palpation. The proposed iApp system consists of an auscultation sensor, camera for inspection, and custom-built hardware for automatic percussion and palpation. Experiments were designed to categorize the eight abdominal divisions of healthy subjects based on the system multi-modal data. A deep multi-modal learning model, yielding a single prediction from multi-modal inputs, was designed for learning distinctive features in eight abdominal divisions. The model's performance was evaluated in terms of the classification accuracy, sensitivity, positive predictive value, and F-measure, using epoch-wise and subject-wise methods. The results demonstrate that the iApp system can successfully categorize abdominal divisions, with the test accuracy of 89.46%. Through an automatic examination of the iApp system, this proof-of-concept study demonstrates a sophisticated classification by extracting distinct features of different abdominal divisions where different organs are located. In the future, we intend to capture the distinct features between normal and abnormal tissues while securing patient data and demonstrate the feasibility of a fully telediagnostic system that can support abnormality diagnosis.

Predict Pregnancy Outcomes in the COVID-19 Pandemic Using Electronic Health Records and Machine Learning Approach

COVID-19 negatively impacts maternal health. We use national electronic health records data and machine learning techniques to recognize risk factors that are predictive of negative maternal outcomes in the pandemic. The cohort has been built with 191,403 gestations. The findings of this study will help advance the clinical decision support system for preventing negative maternal outcomes and promoting maternal health. © 2022 IEEE.

On the Analysis of a Real Dataset of COVID-19 Patients in Alava

COVID-19 has been spread to many countries all over the world in a relatively short period, largely overwhelmed hospitals have been a direct consequence of the explosive increase of coronavirus cases. In this dire situation, the demand for the development of clinical decision support systems based on predictive algorithms has increased, since these predictive technologies may help to alleviate unmanageable stress on healthcare systems. We contribute to this effort by a comprehensive study over a real dataset of covid-19 patients from a local hospital. The collected dataset is representative of the local policies on data gathering implemented during the pandemic, showing high imabalance and large number of missing values. In this paper, we report a descriptive analysis of the data that points out the large disparity of data in terms of severity and age. Furthermore, we report the results of the principal component analysis (PCA) and Logistic Regression (LR) techniques to find out which variables are the most relevant and their respective weight. The results show that there are two very relevant variables for the detection of the most severe cases, yielding promissing results. One of our paper conclussions is a strong recommendation to the local authorities to improve the data gathering protocols. © 2022, Springer Nature Switzerland AG.

Mortality Prediction in ICU Patients Suffering from Stroke

Ischemic stroke is a medical emergency that requires hospitalization and occasionally, specialized care at the Intensive Care Unit. Mortality prediction in the ICUs has been a challenge for intensivists, since prompt identification could impact medical clinical practices and allow efficient allocation of health resources in the ICUs, which are extremely restricted, especially in the era of COVID-19 pandemic. Clinical decision support systems based on machine learning algorithms are taking advantage of the vast amount of information available in the ICUs and are becoming popular in the medical predictive analysis. This study aims to explore the feasibility of interpretable machine learning models to predict mortality in critically-ill patients suffering from stroke. To do so, a vast variety of clinical and laboratory information stored in the electronic health record, are pre-processed to allow taking into account the temporal characteristics of a patient's stay. An 8-hour sliding observation window was utilized. For the experimental evaluation we used the Medical Information Mart for Intensive Care Database (MIMIC-IV). Results indicate sufficient ability to predict mortality at the end of a given day during the patient's stay. Moreover, attribute evaluation highlights the important indicators to consider when following up with a patient. © 2022 ACM.

Human-centered design of clinical decision support for management of hypertension with chronic kidney disease.

BACKGROUND: Primary care providers face challenges in recognizing and controlling hypertension in patients with chronic kidney disease (CKD). Clinical decision support (CDS) has the potential to aid clinicians in identifying patients who could benefit from medication changes. This study designed an alert to control hypertension in CKD patients using an iterative human-centered design process. METHODS: In this study, we present a human-centered design process employing multiple methods for gathering user requirements and feedback on design and usability. Initially, we conducted contextual inquiry sessions to gather user requirements for the CDS. This was followed by group design sessions and one-on-one formative think-aloud sessions to validate requirements, obtain feedback on the design and layout, uncover usability issues, and validate changes. RESULTS: This study included 20 participants. The contextual inquiry produced 10 user requirements which influenced the initial alert design. The group design sessions revealed issues related to several themes, including recommendations and clinical content that did not match providers' expectations and extraneous information on the alerts that did not provide value. Findings from the individual think-aloud sessions revealed that participants disagreed with some recommended clinical actions, requested additional information, and had concerns about the placement in their workflow. Following each step, iterative changes were made to the alert content and design. DISCUSSION: This study showed that participation from users throughout the design process can lead to a better understanding of user requirements and optimal design, even within the constraints of an EHR alerting system. While raising awareness of design needs, it also revealed concerns related to workflow, understandability, and relevance. CONCLUSION: The human-centered design framework using multiple methods for CDS development informed the creation of an alert to assist in the treatment and recognition of hypertension in patients with CKD.

Promoting ß-lactam utilization through suppression of electronic medical record cross-allergy alerts.

PURPOSE: Current literature surrounding management of patients with reported ß-lactam allergies focuses on allergy delabeling. Standard clinical decision support tools have not been optimized to be compatible with the currently accepted cross-reaction rate of 1% to 2%. This potentially promotes use of non-ß-lactam antibiotics, which are often not first-line therapy and may carry increased risks. The impact of electronic medical record (EMR) clinical decision support tool optimization on utilization of ß-lactam antibiotics in ß-lactam-allergic patients was evaluated. METHODS: A retrospective pre-post ß-lactam cross-allergy EMR alert suppression quality improvement intervention cohort study of ß-lactam-allergic adult inpatients prescribed antibiotics was conducted. Preintervention baseline data were collected for an initial cohort admitted during September 2018. The intervention, in which clinical decision support rules were updated to display ß-lactam cross-sensitivity allergy alerts only for ß-lactam-allergic patients with documentation of organization-defined high-severity reactions of anaphylaxis, hives, and shortness of breath, was implemented August 20, 2019. The postintervention cohort included patients admitted during September 2019. RESULTS: A 91% increase in the percentage of ß-lactam-allergic patients who received a ß-lactam agent at any time during their admission was noted after the intervention (26.6% vs 51%, P < 0.001). Statistically significant decreases in prescribing of alternative antibiotic classes were seen for fluoroquinolones (decrease from 45.3% to 26%, P < 0.001), aminoglycosides (decrease from 9.4% to 2.9%, P = 0.002), and aztreonam (decrease from 30% to 16.7%, P < 0.001). CONCLUSION: EMR ß-lactam cross-allergy alert optimization consistent with current literature significantly improved the utilization of alternative ß-lactam subclasses, mostly through ß-lactam prescribing as initial therapy in ß-lactam-allergic patients.

Applying Bayesian Networks to help Physicians Diagnose Respiratory Diseases in the context of COVID-19 Pandemic

The differential diagnosis of respiratory diseases is usually a challenge for medical specialists in the first line of care, increased under the current COVID-19 pandemic. A Clinical Decision Support System -CDSS- is being developed using Bayesian Networks - BNs - to help physicians diagnose respiratory diseases, including those related to COVID-19. Network structure has been elicited from expert physicians, and network parameters (diseases prevalence, symptoms, findings, and lab results conditional probabilities) were extracted from relevant bibliography or currently standard global information sources. The CDSS is being tested using case studies taken from real situations, provided and validated by physicians. The resulting system demonstrates the suitability and flexibility of BNs for diagnosis support and healthcare training.

A Flexible Clinical Decision Support Model and its Application to COVID-19

The positive role of clinical decision support systems based on clinical guidelines in reducing medical errors and improving patient outcomes has been widely recognized. However, the knowledge in clinical guidelines is usually hard-coded into clinical decision support systems, making it difficult for these systems to adapt to the rapid changes of clinical guidelines. Knowledge being hard-coded into the system also means that the system is a black box, and doctors cannot understand the decision-making logic behind the system. These reasons make it difficult for clinical decision support systems to be applied on a large scale. This paper proposes a flexible clinical decision support model, which contains two key parts, namely the knowledge authoring environment and the knowledge execution environment. The transition of knowledge from hard-coded to flexible editing is illustrated in the COVID-19 case. This flexible method will be applied to more complex clinical problems in the future. © 2021 IEEE.

Pandemic-Related Impairment in the Monitoring of Patients With Hypertension and Diabetes and the Development of a Digital Solution for the Community Health Worker: Quasiexperimental and Implementation Study.

BACKGROUND: The restrictions imposed by the COVID-19 pandemic reduced health service access by patients with chronic diseases. The discontinuity of care is a cause of great concern, mainly in vulnerable regions. OBJECTIVE: This study aimed to assess the impact of the COVID-19 pandemic on people with hypertension and diabetes mellitus (DM) regarding the frequency of consultations and whether their disease was kept under control. The study also aimed to develop and implement a digital solution to improve monitoring at home. METHODS: This is a multimethodological study. A quasiexperimental evaluation assessed the impact of the pandemic on the frequency of consultations and control of patients with hypertension and DM in 34 primary health care centers in 10 municipalities. Then, an implementation study developed an app with a decision support system (DSS) for community health workers (CHWs) to identify and address at-risk patients with uncontrolled hypertension or DM. An expert panel assessment evaluated feasibility, usability, and utility of the software. RESULTS: Of 5070 patients, 4810 (94.87%) had hypertension, 1371 (27.04%) had DM, and 1111 (21.91%) had both diseases. There was a significant reduction in the weekly number of consultations (107, IQR 60.0-153.0 before vs 20.0, IQR 7.0-29.0 after social restriction; P<.001). Only 15.23% (772/5070) of all patients returned for a consultation during the pandemic. Individuals with hypertension had lower systolic (120.0, IQR 120.0-140.0 mm Hg) and diastolic (80.0, IQR 80.0-80.0 mm Hg) blood pressure than those who did not return (130.0, IQR 120.0-140.0 mm Hg and 80.0, IQR 80.0-90.0 mm Hg, respectively; P<.001). Also, those who returned had a higher proportion of controlled hypertension (64.3% vs 52.8%). For DM, there were no differences in glycohemoglobin levels. Concerning the DSS, the experts agreed that the CHWs can easily incorporate it into their routines and the app can identify patients at risk and improve treatment. CONCLUSIONS: The COVID-19 pandemic caused a significant drop in the number of consultations for patients with hypertension and DM in primary care. A DSS for CHW has proved to be feasible, useful, and easily incorporated into their routines.

Predicting hospital readmission risk in patients with COVID-19: A machine learning approach.

Introduction: The Coronavirus 2019 (COVID-19) epidemic stunned the health systems with severe scarcities in hospital resources. In this critical situation, decreasing COVID-19 readmissions could potentially sustain hospital capacity. This study aimed to select the most affecting features of COVID-19 readmission and compare the capability of Machine Learning (ML) algorithms to predict COVID-19 readmission based on the selected features. Material and methods: The data of 5791 hospitalized patients with COVID-19 were retrospectively recruited from a hospital registry system. The LASSO feature selection algorithm was used to select the most important features related to COVID-19 readmission. HistGradientBoosting classifier (HGB), Bagging classifier, Multi-Layered Perceptron (MLP), Support Vector Machine ((SVM) kernel = linear), SVM (kernel = RBF), and Extreme Gradient Boosting (XGBoost) classifiers were used for prediction. We evaluated the performance of ML algorithms with a 10-fold cross-validation method using six performance evaluation metrics. Results: Out of the 42 features, 14 were identified as the most relevant predictors. The XGBoost classifier outperformed the other six ML models with an average accuracy of 91.7%, specificity of 91.3%, the sensitivity of 91.6%, F-measure of 91.8%, and AUC of 0.91%. Conclusion: The experimental results prove that ML models can satisfactorily predict COVID-19 readmission. Besides considering the risk factors prioritized in this work, categorizing cases with a high risk of reinfection can make the patient triaging procedure and hospital resource utilization more effective.

The feasibility and ongoing use of electronic decision support to strengthen the implementation of IMCI in KwaZulu-Natal, South Africa.

BACKGROUND: Continued efforts are required to reduce preventable child deaths. User-friendly Integrated Management of Childhood Illness (IMCI) implementation tools and supervision systems are needed to strengthen the quality of child health services in South Africa. A 2018 pilot implementation of electronic IMCI case management algorithms in KwaZulu-Natal demonstrated good uptake and acceptance at primary care clinics. We aimed to investigate whether ongoing electronic IMCI implementation is feasible within the existing Department of Health infrastructure and resources. METHODS: In a mixed methods descriptive study, the electronic IMCI (eIMCI) implementation was extended to 22 health facilities in uMgungundlovu district from November 2019 to February 2021. Training, mentoring, supervision and IT support were provided by a dedicated project team. Programme use was tracked, quarterly assessments of the service delivery platform were undertaken and in-depth interviews were conducted with facility managers. RESULTS: From December 2019 - January 2021, 9 684 eIMCI records were completed across 20 facilities, with a median uptake of 29 records per clinic per month and a mean (range) proportion of child consultations using eIMCI of 15% (1-46%). The local COVID-19-related movement restrictions and epidemic peaks coincided with declines in the monthly eIMCI uptake. Substantial inter- and intra-facility variations in use were observed, with the use being positively associated with the allocation of an eIMCI trained nurse (p < 0.001) and the clinician workload (p = 0.032). CONCLUSION: The ongoing eIMCI uptake was sporadic and the implementation undermined by barriers such as low post-training deployment of nurses; poor capacity in the DoH for IT support; and COVID-19-related disruptions in service delivery. Scaling eIMCI in South Africa would rely on resolving these challenges.

Minimization of the Drug and Gene Interactions in Polypharmacy Therapies Augmented with COVID-19 Medications.

Medications Dexamethasone, Remdesivir or Colchicine, used to treat COVID-19 patients, have significant interactions with other medications and the human genome. The study presented in this paper investigates how to use the Personalized Medicine Therapy Optimization Method (PM-TOM) to minimize these interactions in polypharmacy therapies of COVID-19 patients. We applied PM-TOM on the EMR database of Harvard Personal Genome Project (PGP), drug database DrugBank and Comprehensive Toxicogenomics Database (CTD) to analyze polypharmacy therapies augmented with these medications. The main finding is that these COVID-19 medications significantly increase the drug and gene interactions in partially optimized (or unoptimized) therapies, which is not the case in the fully optimized ones. For example, the test results show that in polypharmacy treatments for patients having between 3 and 8 conditions, the average number of drug and gene interactions in partially optimized therapies ranges from 3 to 18 after adding Remdesivir, 4.3 to 20 Colchicine, and 4.7 to 23 Dexamethasone. On the other hand, these interactions in fully optimized therapies range only 0.6 to 5.2, 1.2 to 7, and 2.7 to 11, respectively. These results suggest that polypharmacy therapies should be carefully examined before adding these medications. This recommendation applies to all other situations when polypharmacy patients may conduct new serious conditions, such as COVID-19, requiring additional medications with a high number of drug and gene interactions.

ALFABETO: Supporting COVID-19 hospital admissions with Bayesian Networks

The ongoing pandemics of coronavirus disease has accelerated the implementation of machine learning methods (ML) to support clinical decisions. Within this context, we present the ALFABETO project, whose aim is to aid clinicians during COVID-19 patients hospital admission through the application of ML approaches exploiting clinical and chest x-ray features. Yet, non linear ML classifiers are often perceived as not easily interpretable by users, thus hampering trust in ML predictions. Moreover, these ML models, such as Neural Networks or Random Forest, are not able to include pre-exisisting knowledge about a specific domain and are not designed to find causal relationships between variables. For these reasons, we wanted to investigate if Bayesian Networks were able to properly describe the hospital admission decision process. Bayesian Networks are probabilistic graphical models representing a set of variables and their conditional dependencies. The network structure was derived both from existing medical knowledge and from patients data collected during the first wave of the pandemic. While being explainable, we show that the Bayesian network has similar performance when compared to a less explainable ML model and that was able to generalize well across COVID-19 waves. © 2021 Copyright for this paper by its authors.