Clinical Decision Support Research Articles

Data-driven Machine Learning Models for Risk Stratification and Prediction of Emergence Delirium in Pediatric Patients Underwent Tonsillectomy/Adenotonsillectomy.

In the pediatric surgical population, Emergence Delirium (ED) poses a significant challenge. This study aims to develop and validate machine learning (ML) models to identify key features associated with ED and predict its occurrence in children undergoing tonsillectomy or adenotonsillectomy. The analysis involved data cleaning, exploratory data analysis (EDA), supervised predictive modeling, and unsupervised learning on a medical dataset (n = 423). After preliminary data cleaning, EDA encompassed plotting histograms, boxplots, pairplots, and correlation heatmaps to understand variable distributions and relationships. Four predictive models were trained including logistic regression (LR), random forest (RF), Support Vector Machine (SVM), and Gradient Boosting (XGBoost). The models were evaluated and compared using Receiver Operating Characteristic (ROC) Area Under the Curve (AUC), precision, recall, and feature importance. The RF model showed better performance and was used for the test (AUC-ROC 0.96, precision 1.00, and recall 0.92 on the validation set). K-means clustering was applied to find groups within the data. Elbow method and silhouette scores were used to determine the optimal number of clusters. The formed clusters were analyzed by aggregating features to understand the characteristics of each cluster. EDA revealed significant positive correlations between age, weight, American Society of Anesthesiologists (ASA) health score, and surgery duration with the risk of developing ED. Among the ML models, RF achieved the highest performance. Key predictive variables, based on the model's feature importance, included delirium screening scales, extubation time, and time to regain consciousness. Unsupervised K-means clustering identified 2-3 optimal clusters, which represented distinct patient subgroups: younger, healthier, low-risk individuals (cluster 0), and older patients with increasing chronic disease burden, higher delirium screening scores, and consequently higher post-operative delirium risk (clusters 1 and 2). ML techniques are valuable tools for extracting insights and making accurate predictions from healthcare data. High-performing algorithm-based models can be implemented for clinical decision support systems, facilitating early identification and intervention for ED in pediatric patients. By investigating various variables, it is possible to assess risk and implement preventive measures effectively. Furthermore, unsupervised clustering reveals distinct patient subgroups, enabling personalized perioperative management strategies and enhancing overall patient care.

Adoption, acceptance, and use of a decision support tool to promote timely investigations for cancer in primary care.

The complexities of diagnosing cancer in general practice has driven the development of quality improvement (QI) interventions, including clinical decision support (CDS) and auditing tools. Future Health Today (FHT) is a novel QI tool, consisting of CDS at the point-of-care, practice population-level auditing, recall, and the monitoring of QI activities. Explore the acceptability and usability of the FHT cancer module, which flags patients with abnormal test results that may be indicative of undiagnosed cancer. Interviews were conducted with general practitioners (GPs) and general practice nurses (GPNs), from practices participating in a randomized trial evaluating the appropriate follow-up of patients. Clinical Performance Feedback Intervention Theory (CP-FIT) was used to analyse and interpret the data. The majority of practices reported not using the auditing and QI components of the tool, only the CDS which was delivered at the point-of-care. The tool was used primarily by GPs; GPNs did not perceive the clinical recommendations to be within their role. For the CDS, facilitators for use included a good workflow fit, ease of use, low time cost, importance, and perceived knowledge gain. Barriers for use of the CDS included accuracy, competing priorities, and the patient population. The CDS aligned with the clinical workflow of GPs, was considered non-disruptive to the consultation and easy to implement into usual care. By applying the CP-FIT theory, we were able to demonstrate the key drivers for GPs using the tool, and what limited the use by GPNs.

Design and development of a machine-learning-driven opioid overdose risk prediction tool integrated in electronic health records in primary care settings

BackgroundIntegrating advanced machine-learning (ML) algorithms into clinical practice is challenging and requires interdisciplinary collaboration to develop transparent, interpretable, and ethically sound clinical decision support (CDS) tools. We aimed to design a ML-driven CDS tool to predict opioid overdose risk and gather feedback for its integration into the University of Florida Health (UFHealth) electronic health record (EHR) system.MethodsWe used user-centered design methods to integrate the ML algorithm into the EHR system. The backend and UI design sub-teams collaborated closely, both informed by user feedback sessions. We conducted seven user feedback sessions with five UF Health primary care physicians (PCPs) to explore aspects of CDS tools, including workflow, risk display, and risk mitigation strategies. After customizing the tool based on PCPs’ feedback, we held two rounds of one-on-one usability testing sessions with 8 additional PCPs to gather feedback on prototype alerts. These sessions informed iterative UI design and backend processes, including alert frequency and reappearance circumstances.ResultsThe backend process development identified needs and requirements from our team, information technology, UFHealth, and PCPs. Thirteen PCPs (male = 62%, White = 85%) participated across 7 user feedback sessions and 8 usability testing sessions. During the user feedback sessions, PCPs (n = 5) identified flaws such as the term “high risk” of overdose potentially leading to unintended consequences (e.g., immediate addiction services referrals), offered suggestions, and expressed trust in the tool. In the first usability testing session, PCPs (n = 4) emphasized the need for natural risk presentation (e.g., 1 in 200) and suggested displaying the alert multiple times yearly for at-risk patients. Another 4 PCPs in the second usability testing session valued the UFHealth-specific alert for managing new or unfamiliar patients, expressed concerns about PCPs’ workload when prescribing to high-risk patients, and recommended incorporating the details page into training sessions to enhance usability.ConclusionsThe final backend process for our CDS alert aligns with PCP needs and UFHealth standards. Integrating feedback from PCPs in the early development phase of our ML-driven CDS tool helped identify barriers and facilitators in the CDS integration process. This collaborative approach yielded a refined prototype aimed at minimizing unintended consequences and enhancing usability.

Potential for an Electronic Clinical Decision Support Tool to Support Appropriate Antibiotic Use for Pediatric Diarrhea Among Village Doctors in Bangladesh.

Potential for an Electronic Clinical Decision Support Tool to Support Appropriate Antibiotic Use for Pediatric Diarrhea Among Village Doctors in Bangladesh.

Applying the Non-Adoption, Abandonment, Scale-up, Spread, and Sustainability Framework Across Implementation Stages to Identify Key Strategies to Facilitate Clinical Decision Support System Integration Within a Large Metropolitan Health Service: Interview and Focus Group Study.

Computerized clinical decision support systems (CDSSs) enhance patient care through real-time, evidence-based guidance for health care professionals. Despite this, the effective implementation of these systems for health services presents multifaceted challenges, leading to inappropriate use and abandonment over the course of time. Using the Non-Adoption, Abandonment, Scale-Up, Spread, and Sustainability (NASSS) framework, this qualitative study examined CDSS adoption in a metropolitan health service, identifying determinants across implementation stages to optimize CDSS integration into health care practice. This study aims to identify the theory-informed (NASSS) determinants, which included multiple CDSS interventions across a 2-year period, both at the health-service level and at the individual hospital setting, that either facilitate or hinder the application of CDSSs within a metropolitan health service. In addition, this study aimed to map these determinants onto specific stages of the implementation process, thereby developing a system-level understanding of CDSS application across implementation stages. Participants involved in various stages of the implementation process were recruited (N=30). Participants took part in interviews and focus groups. We used a hybrid inductive-deductive qualitative content analysis and a framework mapping approach to categorize findings into barriers, enablers, or neutral determinants aligned to NASSS framework domains. These determinants were also mapped to implementation stages using the Active Implementation Framework stages approach. Participants comprised clinical adopters (14/30, 47%), organizational champions (5/30, 16%), and those with roles in organizational clinical informatics (5/30, 16%). Most determinants were mapped to the organization level, technology, and adopter subdomains. However, the study findings also demonstrated a relative lack of long-term implementation planning. Consequently, determinants were not uniformly distributed across the stages of implementation, with 61.1% (77/126) identified in the exploration stage, 30.9% (39/126) in the full implementation stage, and 4.7% (6/126) in the installation stages. Stakeholders engaged in more preimplementation and full-scale implementation activities, with fewer cycles of monitoring and iteration activities identified. These findings addressed a substantial knowledge gap in the literature using systems thinking principles to identify the interdependent dynamics of CDSS implementation. A lack of sustained implementation strategies (ie, training and longer-term, adopter-level championing) weakened the sociotechnical network between developers and adopters, leading to communication barriers. More rigorous implementation planning, encompassing all 4 implementation stages, may, in a way, help in addressing the barriers identified and enhancing enablers.

Multiclass classification of Autism Spectrum Disorder, attention deficit hyperactivity disorder, and typically developed individuals using fMRI functional connectivity analysis.

Neurodevelopmental conditions, such as Autism Spectrum Disorder (ASD) and Attention Deficit Hyperactivity Disorder (ADHD), present unique challenges due to overlapping symptoms, making an accurate diagnosis and targeted intervention difficult. Our study employs advanced machine learning techniques to analyze functional magnetic resonance imaging (fMRI) data from individuals with ASD, ADHD, and typically developed (TD) controls, totaling 120 subjects in the study. Leveraging multiclass classification (ML) algorithms, we achieve superior accuracy in distinguishing between ASD, ADHD, and TD groups, surpassing existing benchmarks with an area under the ROC curve near 98%. Our analysis reveals distinct neural signatures associated with ASD and ADHD: individuals with ADHD exhibit altered connectivity patterns of regions involved in attention and impulse control, whereas those with ASD show disruptions in brain regions critical for social and cognitive functions. The observed connectivity patterns, on which the ML classification rests, agree with established diagnostic approaches based on clinical symptoms. Furthermore, complex network analyses highlight differences in brain network integration and segregation among the three groups. Our findings pave the way for refined, ML-enhanced diagnostics in accordance with established practices, offering a promising avenue for developing trustworthy clinical decision-support systems.

Physiotherapy in the digital age: A narrative review of the paradigm shift driven by the integration of artificial intelligence and machine learning

Abstract: A variety of physical impairments and functional restrictions are assessed and treated in the practice of physiotherapy. Subjective measures, rater variability, and restricted access to high-quality care are some of the unavoidable problems that contemporary physical therapy practice approaches must overcome. In light of these challenges, cutting-edge technologies such as artificial intelligence (AI) and machine learning (ML) are demonstrating remarkable efficacy in tackling these issues head-on. The focus of this review is to explore how the integration of AI and ML might change physical therapy practice and education in the age of digital communication. It delves into the challenges accompanying this integration and considers future prospects in this domain. A literature search was conducted using data base PubMed, Google Scholar, Web of Science, and Scopus with keywords such as ‘physiotherapy’, ‘artificial intelligence’, and ‘machine learning’, limited to English articles from 2014 to 2024. Results were imported into reference management software, duplicates removed, and relevant articles were screened and assessed for inclusion, with reasons for exclusion documented. Emerging technologies like AI and ML use algorithms to examine patient data and make automatic decisions, enhancing areas such as virtual reality therapy (VR), tele-rehabilitation, clinical decision support, individualized rehabilitation and physical function evaluation. These advance technologies optimize physical therapy and improve outcomes, but further research is needed to address obstacles like bias and data privacy to ensure responsible implementation. AI and ML can revolutionize physical therapy by improving therapy precision, patient monitoring, optimization and individualized therapy plan. However, it’s crucial for physiotherapists to balance technological advancements with compassionate, patient-centred approach.

Reinforcement Learning to Optimize Ventilator Settings for Patients on Invasive Mechanical Ventilation: Retrospective Study.

One of the significant changes in intensive care medicine over the past 2 decades is the acknowledgment that improper mechanical ventilation settings substantially contribute to pulmonary injury in critically ill patients. Artificial intelligence (AI) solutions can optimize mechanical ventilation settings in intensive care units (ICUs) and improve patient outcomes. Specifically, machine learning algorithms can be trained on large datasets of patient information and mechanical ventilation settings. These algorithms can then predict patient responses to different ventilation strategies and suggest personalized ventilation settings for individual patients. In this study, we aimed to design and evaluate an AI solution that could tailor an optimal ventilator strategy for each critically ill patient who requires mechanical ventilation. We proposed a reinforcement learning-based AI solution using observational data from multiple ICUs in the United States. The primary outcome was hospital mortality. Secondary outcomes were the proportion of optimal oxygen saturation and the proportion of optimal mean arterial blood pressure. We trained our AI agent to recommend low, medium, and high levels of 3 ventilator settings-positive end-expiratory pressure, fraction of inspired oxygen, and ideal body weight-adjusted tidal volume-according to patients' health conditions. We defined a policy as rules guiding ventilator setting changes given specific clinical scenarios. Off-policy evaluation metrics were applied to evaluate the AI policy. We studied 21,595 and 5105 patients' ICU stays from the e-Intensive Care Unit Collaborative Research (eICU) and Medical Information Mart for Intensive Care IV (MIMIC-IV) databases, respectively. Using the learned AI policy, we estimated the hospital mortality rate (eICU 12.1%, SD 3.1%; MIMIC-IV 29.1%, SD 0.9%), the proportion of optimal oxygen saturation (eICU 58.7%, SD 4.7%; MIMIC-IV 49%, SD 1%), and the proportion of optimal mean arterial blood pressure (eICU 31.1%, SD 4.5%; MIMIC-IV 41.2%, SD 1%). Based on multiple quantitative and qualitative evaluation metrics, our proposed AI solution outperformed observed clinical practice. Our study found that customizing ventilation settings for individual patients led to lower estimated hospital mortality rates compared to actual rates. This highlights the potential effectiveness of using reinforcement learning methodology to develop AI models that analyze complex clinical data for optimizing treatment parameters. Additionally, our findings suggest the integration of this model into a clinical decision support system for refining ventilation settings, supporting the need for prospective validation trials.

Enhancing the accuracy and effectiveness of diagnosis of spontaneous bacterial peritonitis in cirrhotic patients: A machine learning approach utilizing clinical and laboratory data

PurposeSpontaneous bacterial peritonitis (SBP) is a bacterial infection of ascitic fluid that develops naturally, without being triggered by any surgical conditions or procedures, and is a common complication of cirrhosis. With a potential mortality rate of 40 ​%, accurate diagnosis and prompt initiation of appropriate antibiotic therapy are crucial for optimizing patient outcomes and preventing life-threatening complications. This study aimed to expand the use of computational models to improve the diagnostic accuracy of SBP in cirrhotic patients by incorporating a broader range of data, including clinical variables and laboratory values. Patients and methodsWe employed 5 machine learning classification methods - Decision Tree, Support Vector Machine, Naive Bayes, K-Nearest Neighbor, and Random Forest, utilizing a variety of demographic, clinical, and laboratory features and biomarkers. ResultsAscitic fluid markers, including white blood cell (WBC) count, lactate dehydrogenase (LDH), total protein, and polymorphonuclear cells (PMN), significantly differentiated between SBP and non-SBP patients. The Random Forest model demonstrated the highest overall accuracy at 86 ​%, while the Naive Bayes model achieved the highest sensitivity at 72 ​%. Utilizing 10 key features instead of the full feature set improved model performance, notably enhancing specificity and accuracy. ConclusionOur analysis highlights the potential of machine learning to enhance the accuracy of SBP diagnosis in cirrhotic patients. Integrating these models into clinical workflows could substantially improve patient outcomes. To achieve this, ongoing multidisciplinary research is crucial. Ensuring model interpretability, continuous monitoring, and rigorous validation will be essential for the successful implementation of real-time clinical decision support systems.

Competing and Noncompeting Risk Models for Predicting Kidney Allograft Failure.

Prognostic models are becoming increasingly relevant in clinical trials as potential surrogate endpoints, and for patient management as clinical decision support tools. However, the impact of competing risks on model performance remains poorly investigated. We aimed to carefully assess the performance of competing risk and noncompeting risk models in the context of kidney transplantation, where allograft failure and death with a functioning graft are two competing outcomes. We included 11,046 kidney transplant recipients enrolled in 10 countries. We developed prediction models for long-term kidney graft failure prediction, without accounting (i.e., censoring) and accounting for the competing risk of death with a functioning graft, using Cox, Fine-Gray, and cause-specific Cox regression models. To this aim, we followed a detailed and transparent analytical framework for competing and noncompeting risk modelling, and carefully assessed the models' development, stability, discrimination, calibration, overall fit, clinical utility, and generalizability in external validation cohorts and subpopulations. More than 15 metrics were used to provide an exhaustive assessment of model performance. Among 11,046 recipients in the derivation and validation cohorts, 1,497 (14%) lost their graft and 1,003 (9%) died with a functioning graft after a median follow-up post-risk evaluation of 4.7 years (IQR 2.7-7.0). The cumulative incidence of graft loss was similarly estimated by Kaplan-Meier and Aalen-Johansen methods (17% versus 16% in the derivation cohort). Cox and competing risk models showed similar and stable risk estimates for predicting long-term graft failure (average mean absolute prediction error of 0.0140, 0.0138 and 0.0135 for Cox, Fine-Gray, and cause-specific Cox models, respectively). Discrimination and overall fit were comparable in the validation cohorts, with concordance index ranging from 0.76 to 0.87. Across various subpopulations and clinical scenarios, the models performed well and similarly, although in some high-risk groups (such as donors over 65 years old), the findings suggest a trend towards moderately improved calibration when using a competing risk approach. Competing and noncompeting risk models performed similarly in predicting long-term kidney graft failure.

Personalised decision support in the management of patients with musculoskeletal pain in primary physiotherapy care: a cluster randomised controlled trial (the SupportPrim project).

We developed the SupportPrim PT clinical decision support system (CDSS) using the artificial intelligence method case-based reasoning to support personalised musculoskeletal pain management. The aim of this study was to evaluate the effectiveness of the CDSS for patients in physiotherapy practice. A cluster randomised controlled trial was conducted in primary care in Norway. We randomised 44 physiotherapists to (1) use the CDSS alongside usual care or (2) usual care alone. The CDSS provided personalised treatment recommendations based on a case base of 105 patients with positive outcomes. During the trial, the case-based reasoning system did not have an active learning capability; therefore, the case base size remained the same throughout the study. We included 724 patients presenting with neck, shoulder, back, hip, knee, or complex pain (CDSS; n = 358, usual care; n = 366). Primary outcomes were assessed with multilevel logistic regression using self-reported Global Perceived Effect (GPE) and Patient-Specific Functional Scale (PSFS). At 12 weeks, 165/298 (55.4%) patients in the intervention group and 176/321 (54.8%) in the control group reported improvement in GPE (odds ratio, 1.18; confidence interval, 0.50-2.78). For PSFS, 173/290 (59.7%) patients in the intervention group and 218/310 (70.3%) in the control group reported clinically important improvement in function (odds ratio, 0.41; confidence interval, 0.20-0.85). No significant between-group differences were found for GPE. For PSFS, there was a significant difference favouring the control group, but this was less than the prespecified difference of 15%. We identified several study limitations and recommend further investigation into artificial intelligence applications for managing musculoskeletal pain.

Comprehensive Approach to TB Diagnosis Powered by Machine Learning in Endemic Countries

Abstract Introduction/Objective Tuberculosis (TB), a leading cause of mortality worldwide, is primarily diagnosed through methods that face numerous challenges, including lengthy result times and diagnostic inaccuracy, particularly in the context of drug-resistant strains and latent infections. This study proposes a novel approach that integrates multiple diagnostic modalities with machine learning to enhance TB diagnosis accuracy. Methods/Case Report The Machine learning model was tailored for the classification of radiological images, specifically CT scans, using original DICOM images to preserve valuable voxel information. The study involved a curated dataset of CT scans from Pakistan, TB-positive (smear+/Culture+ n=72; smear+/culture- n= 3; smear-/Culture+ n=16; smear-/Culture- n=14; clinical and radiological diagnosis only n= 12) and healthy (n=103) scans from COVID-CTSet were included in the training to assess the specificity of the model. Data preprocessing included noise reduction, histogram matching, and data augmentation techniques to ensure a robust training dataset. The methodology also incorporated using k-fold cross validation and a custom learning rate scheduling strategy to optimize the model’s training process. To further test its robustness, the model was evaluated by swapping random patches and analyzing 60 mm patches from the entire CT scan for local feature categorization. Results (if a Case Study enter NA) The model has demonstrated outstanding performance, with sensitivity and specificity exceeding 95% on our internal dataset (n=117) and aligning well with clinical, smear, culture, and anti-TB antibody findings. It also showed promising results on the external NIH TB Portal’s dataset (n= 1,460). The model is robust to variations in CT scanner image quality and noise profiles, maintaining strong generalization across datasets. Even when analyzing random 60mm patches, effectively classifying all local features Conclusion Our study enhances tuberculosis diagnostic methods in endemic regions by blending machine learning with conventional diagnostic techniques. This approach improves TB detection, particularly the early and accurate identification in areas impacted by drug-resistant strains and asymptomatic latent infections. By integrating radiological and clinico-pathological data, we aim to develop a comprehensive diagnostic score. Moving forward, we plan to further integrate this method into clinical decision support systems and adapt the approach to other infectious diseases.

Ordering and Utilization of Peripheral Blood Flow Cytometry in an Academic Medical Center

Abstract Introduction/Objective The American Society for Clinical Pathology and the American Board of Internal Medicine’s Choosing Wisely campaign have developed best practice utilization guidelines for laboratory testing. Wisconsin Diagnostic Laboratories and Medical College of Wisconsin leadership noticed an increase in inappropriate peripheral blood flow cytometry testing in the last two years. An intervention was developed to help reduce inappropriate testing requests. Methods/Case Report The laboratory implemented a clinical decision support tool in the electronic medical record system that looked at patient diagnosis, history of stem cell transplant, and recommendations from pathology to help improve utilization of peripheral blood flow cytometry testing. Based on patient history, flow cytometry testing would move forward, or an alternative order of a peripheral blood pathology smear review with reflex to flow cytometry was offered. Pathologists would reflex flow cytometry if abnormalities were present by light microscopy or predetermined reflexing criteria were met. Results (if a Case Study enter NA) 314 cases were included for analysis. The intervention reduced volumes of peripheral blood flow cytometry in patients without a history of hematologic malignancy by 49.1% (p=<0.001). Appropriate utilization increased from 47.0% in the pre-intervention period to 80.4% in the post intervention period (p=<0.001). Significant diagnosis rate increased from 42.5% to 50.0% (p=.261). This resulted in a cost savings of $12,232 for the health system during the intervention period. Conclusion Clinical decision support based utilization efforts provide strong and consistent interventions to reduce the unnecessary and inappropriate ordering of peripheral blood flow cytometry.

Artificial Intelligence in Oral and Maxillofacial Surgery: Bridging the Gap between Technology and Clinical Practice a Narrative Review

Objective: To provide a comprehensive overview of current applications and future prospects of artificial intelligence (AI) in oral and maxillofacial surgery (OMFS), while critically analyzing implementation challenges and exploring potential advancements.  Methods A systematic literature review was conducted using PubMed/MEDLINE and Embase databases, encompassing English-language articles up to December 30, 2023. Search terms combined OMFS and AI concepts, with database-specific syntax employed.  Results AI applications in OMFS span multiple domains, including image analysis, surgical planning, intraoperative guidance, and clinical decision support. Deep learning models have demonstrated high accuracy in detecting mandibular fractures, performing cephalometric analyses, and classifying maxillofacial pathologies. AI-enhanced surgical planning and robotic systems show promise in improving precision and outcomes across various OMFS procedures. However, challenges persist in data quality, clinical validation, and seamless workflow integration.  Conclusions AI technologies have the potential to significantly enhance diagnostic accuracy, surgical precision, and treatment outcomes in OMFS. Future research directions include developing multimodal AI systems, advancing AI-powered surgical navigation, and exploring federated learning approaches. Successful implementation of AI in OMFS practice will require collaborative efforts among clinicians, researchers, engineers, and policymakers to address technical, ethical, and regulatory challenges. As these hurdles are overcome, AI is poised to become an integral part of OMFS, augmenting surgical capabilities and elevating patient care standards.

Just another tool in their repertoire: uncovering insights into public and patient perspectives on clinicians' use of machine learning in perioperative care.

Successful implementation of machine learning-augmented clinical decision support systems (ML-CDSS) in perioperative care requires the prioritization of patient-centric approaches to ensure alignment with societal expectations. We assessed general public and surgical patient attitudes and perspectives on ML-CDSS use in perioperative care. A sequential explanatory study was conducted. Stage 1 collected public opinions through a survey. Stage 2 ascertained surgical patients' experiences and attitudes via focus groups and interviews. For Stage 1, a total of 281 respondents' (140 males [49.8%]) data were considered. Among participants without ML awareness, males were almost three times more likely than females to report more acceptance (OR = 2.97; 95% CI, 1.36-6.49) and embrace (OR = 2.74; 95% CI, 1.23-6.09) of ML-CDSS use by perioperative teams. Males were almost twice as likely as females to report more acceptance across all perioperative phases with ORs ranging from 1.71 to 2.07. In Stage 2, insights from 10 surgical patients revealed unanimous agreement that ML-CDSS should primarily serve a supportive function. The pre- and post-operative phases were identified explicitly as forums where ML-CDSS can enhance care delivery. Patients requested for education on ML-CDSS's role in their care to be disseminated by surgeons across multiple platforms. The general public and surgical patients are receptive to ML-CDSS use throughout their perioperative care provided its role is auxiliary to perioperative teams. However, the integration of ML-CDSS into perioperative workflows presents unique challenges for healthcare settings. Insights from this study can inform strategies to support large-scale implementation and adoption of ML-CDSS by patients in all perioperative phases. Key strategies to promote the feasibility and acceptability of ML-CDSS include clinician-led discussions about ML-CDSS's role in perioperative care, established metrics to evaluate the clinical utility of ML-CDSS, and patient education.

Predicting maintenance lithium response for bipolar disorder from electronic health records-a retrospective study.

Optimising maintenance drug treatment selection for people with bipolar disorder is challenging. There is some evidence that clinical and demographic features may predict response to lithium. However, attempts to personalise treatment choice have been limited. We aimed to determine if machine learning methods applied to electronic health records could predict differential response to lithium or olanzapine. From electronic United Kingdom primary care records, we extracted a cohort of individuals prescribed either lithium (19,106 individuals) or olanzapine (12,412) monotherapy. Machine learning models were used to predict successful monotherapy maintenance treatment, using 113 clinical and demographic variables, 8,017 (41.96%) lithium responders and 3,831 (30.87%) olanzapine responders. We found a quantitative structural difference in that lithium maintenance responders were weakly predictable in our holdout sample, consisting of the 5% of patients with the most recent exposure. Age at first diagnosis, age at first treatment and the time between these were the most important variables in all models. Even if we failed to predict successful monotherapy olanzapine treatment, and so to definitively separate lithium vs. olanzapine responders, the characterization of the two groups may be used for classification by proxy. This can, in turn, be useful for establishing maintenance therapy. The further exploration of machine learning methods on EHR data for drug treatment selection could in the future play a role for clinical decision support. Signals in the data encourage further experiments with larger datasets to definitively separate lithium vs. olanzapine responders.

Development and expert inspections of the user interface for a primary care decision support system

BackgroundGeneral practitioners play a unique key role in diagnosing patients with unclear diseases. Decision support systems in primary care can assist with diagnosis provided that they are efficient and user-friendly. ObjectivesThe objective of this study is to develop a high-fidelity prototype of the user interface of a clinical decision support system for primary care, particularly for diagnosis support in unclear diseases, using expert inspections at an early stage of development to ensure a high level of usability. MethodsThe user interface prototype was iteratively developed based on previous research, design principles, and usability guidelines. During the development phase, three usability inspections were carried out by all experts at four-week intervals as heuristic walkthrough. Each inspection consisted of two parts: 1) Task-based inspection 2) Free exploration and evaluation based on usability heuristics. Five domain experts assessed the current status of development.The tasks in the inspections were based on the task model derived in the requirements analysis: perform data entry, review and discuss results, schedule further diagnostics, refer to specialists and close case. ResultsAs a result of this iterative development, a high-fidelity, clickable user interface prototype was created that is able to fulfil all six tasks of our task model. The usability inspections identified a total of 196 usability issues (for all 3 inspections; Part 1: 90 issues, Part 2: 106 issues), ranging in severity from minor to severe. These served the continuous adjustment and improvement of the prototype. All main tasks were completed successfully despite these problems. ConclusionUsability inspections through heuristic walkthroughs can support and optimise the development of a user-centred decision support system in order to ensure its suitability for performing relevant tasks.

Patients' attitudes toward artificial intelligence in dentistry and their trust in dentists.

This study intended to evaluate patients' attitudes toward the use of AI in dental radiographic detection of occlusal caries and the impact of AI-based diagnosis on their trust in dentists. A total of 272 completed questionnaires were included in this study. In the first part of the study, approval was obtained from the patients, and data were collected about their socio-demographic characteristics. In the second part the 11-item Dentist Trust Scale was applied. In the third and fourth parts, there were questions about two clinical scenarios, the patients' knowledge of attitudes toward AI, and how the AI-based diagnosis had affected their trust. Evaluation was performed using a Likert-type scale. Data were analyzed with the Chi-square, one-way ANOVA, and ordinal logistic regression tests (p < 0.05). The patients believed that "AI is useful" (3.86 ± 1.03) and were not afraid of the use of AI in dentistry (2.40 ± 1.05). Educational level was considerably related to the patients' attitudes to the use of AI for dental diagnostics (p < 0.05). The patients stated that "dentists are extremely thorough and careful" (4.39 ± 0.77). The patients displayed a positive attitude to AI-based diagnosis in the dental field and appear to exhibit trust in dentists. The use of Al in routine clinical practice canprovide important benefit to physicians as a clinical decision support system in dentistry and understanding patients' attitudes may allow dentists to shape AI-supported dentistry in the future.

Risk-Specific Training Cohorts to Address Class Imbalance in Surgical Risk Prediction.

Machine learning tools are increasingly deployed for risk prediction and clinical decision support in surgery. Class imbalance adversely impacts predictive performance, especially for low-incidence complications. To evaluate risk-prediction model performance when trained on risk-specific cohorts. This cross-sectional study performed from February 2024 to July 2024 deployed a deep learning model, which generated risk scores for common postoperative complications. A total of 109 445 inpatient operations performed at 2 University of Florida Health hospitals from June 1, 2014, to May 5, 2021 were examined. The model was trained de novo on separate cohorts for high-risk, medium-risk, and low-risk Common Procedure Terminology codes defined empirically by incidence of 5 postoperative complications: (1) in-hospital mortality; (2) prolonged intensive care unit (ICU) stay (≥48 hours); (3) prolonged mechanical ventilation (≥48 hours); (4) sepsis; and (5) acute kidney injury (AKI). Low-risk and high-risk cutoffs for complications were defined by the lower-third and upper-third prevalence in the dataset, except for mortality, cutoffs for which were set at 1% or less and greater than 3%, respectively. Model performance metrics were assessed for each risk-specific cohort alongside the baseline model. Metrics included area under the receiver operating characteristic curve (AUROC), area under the precision-recall curve (AUPRC), F1 scores, and accuracy for each model. A total of 109 445 inpatient operations were examined among patients treated at 2 University of Florida Health hospitals in Gainesville (77 921 procedures [71.2%]) and Jacksonville (31 524 procedures [28.8%]). Median (IQR) patient age was 58 (43-68) years, and median (IQR) Charlson Comorbidity Index score was 2 (0-4). Among 109 445 operations, 55 646 patients were male (50.8%), and 66 495 patients (60.8%) underwent a nonemergent, inpatient operation. Training on the high-risk cohort had variable impact on AUROC, but significantly improved AUPRC (as assessed by nonoverlapping 95% confidence intervals) for predicting mortality (0.53; 95% CI, 0.43-0.64), AKI (0.61; 95% CI, 0.58-0.65), and prolonged ICU stay (0.91; 95% CI, 0.89-0.92). It also significantly improved F1 score for mortality (0.42; 95% CI, 0.36-0.49), prolonged mechanical ventilation (0.55; 95% CI, 0.52-0.58), sepsis (0.46; 95% CI, 0.43-0.49), and AKI (0.57; 95% CI, 0.54-0.59). After controlling for baseline model performance on high-risk cohorts, AUPRC increased significantly for in-hospital mortality only (0.53; 95% CI, 0.42-0.65 vs 0.29; 95% CI, 0.21-0.40). In this cross-sectional study, by training separate models using a priori knowledge for procedure-specific risk classes, improved performance in standard evaluation metrics was observed, especially for low-prevalence complications like in-hospital mortality. Used cautiously, this approach may represent an optimal training strategy for surgical risk-prediction models.

Predicting clinical trial success for Clostridium difficile infections based on preclinical data.

Preclinical models are ubiquitous and essential for drug discovery, yet our understanding of how well they translate to clinical outcomes is limited. In this study, we investigate the translational success of treatments for Clostridium difficile infection from animal models to human patients. Our analysis shows that only 36% of the preclinical and clinical experiment pairs result in translation success. Univariate analysis shows that the sustained response endpoint is correlated with translation failure (SRC = -0.20, p-value = 1.53 × 10-54), and explainability analysis of multi-variate random forest models shows that both sustained response endpoint and subject age are negative predictors of translation success. We have developed a recommendation system to help plan the right preclinical study given factors such as drug dosage, bacterial dosage, and preclinical/clinical endpoint. With an accuracy of 0.76 (F1 score of 0.71) and by using only 7 features (out of 68 total), the proposed system boosts translational efficiency by 25%. The method presented can extend to any disease and can serve as a preclinical to clinical translation decision support system to accelerate drug discovery and de-risk clinical outcomes.