Health consumers' ethical concerns towards artificial intelligence in Australian emergency departments.

OBJECTIVES: To investigate health consumers' ethical concerns towards the use of artificial intelligence (AI) in EDs. METHODS: Qualitative semi-structured interviews with health consumers, recruited via health consumer networks and community groups, interviews conducted between January and August 2022. RESULTS: We interviewed 28 health consumers about their perceptions towards the ethical use of AI in EDs. The results discussed in this paper highlight the challenges and barriers for the effective and ethical implementation of AI from the perspective of Australian health consumers. Most health consumers are more likely to support AI health tools in EDs if they continue to be involved in the decision-making process. There is considerably more approval of AI tools that support clinical decision-making, as opposed to replacing it. There is mixed sentiment about the acceptability of AI tools influencing clinical decision-making and judgement. Health consumers are mostly supportive of the use of their data to train and develop AI tools but are concerned with who has access. Addressing bias and discrimination in AI is an important consideration for some health consumers. Robust regulation and governance are critical for health consumers to trust and accept the use of AI. CONCLUSION: Health consumers view AI as an emerging technology that they want to see comprehensively regulated to ensure it functions safely and securely with EDs. Without considerations made for the ethical design, implementation and use of AI technologies, health consumer trust and acceptance in the use of these tools will be limited.

Predicting multifaceted risks using machine learning in atrial fibrillation: insights from GLORIA-AF study.

Aims: Patients with atrial fibrillation (AF) have a higher risk of ischaemic stroke and death. While anticoagulants are effective at reducing these risks, they increase the risk of bleeding. Current clinical risk scores only perform modestly in predicting adverse outcomes, especially for the outcome of death. We aimed to test the multi-label gradient boosting decision tree (ML-GBDT) model in predicting risks for adverse outcomes in a prospective global AF registry. Methods and results: We studied patients from phase II/III of the Global Registry on Long-Term Oral Anti-Thrombotic Treatment in Patients with Atrial Fibrillation registry between 2011 and 2020. The outcomes were all-cause death, ischaemic stroke, and major bleeding within 1 year following the AF. We trained the ML-GBDT model and compared its discrimination with the clinical scores in predicting patient outcomes. A total of 25 656 patients were included [mean age 70.3 years (SD 10.3); 44.8% female]. Within 1 year after AF, ischaemic stroke occurred in 215 (0.8%), major bleeding in 405 (1.6%), and death in 897 (3.5%) patients. Our model achieved an optimized area under the curve in predicting death (0.785, 95% CI: 0.757-0.813) compared with the Charlson Comorbidity Index (0.747, P = 0.007), ischaemic stroke (0.691, 0.626-0.756) compared with CHA2DS2-VASc (0.613, P = 0.028), and major bleeding (0.698, 0.651-0.745) as opposed to HAS-BLED (0.607, P = 0.002), with improvement in net reclassification index (10.0, 12.5, and 23.6%, respectively). Conclusion: The ML-GBDT model outperformed clinical risk scores in predicting the risks in patients with AF. This approach could be used as a single multifaceted holistic tool to optimize patient risk assessment and mitigate adverse outcomes when managing AF.

Attitudes towards artificial intelligence in emergency medicine.

OBJECTIVE: To assess Australian and New Zealand emergency clinicians' attitudes towards the use of artificial intelligence (AI) in emergency medicine. METHODS: We undertook a qualitative interview-based study based on grounded theory. Participants were recruited through ED internal mailing lists, the Australasian College for Emergency Medicine Bulletin, and the research teams' personal networks. Interviews were transcribed, coded and themes presented. RESULTS: Twenty-five interviews were conducted between July 2021 and May 2022. Thematic saturation was achieved after 22 interviews. Most participants were from either Western Australia (52%) or Victoria (16%) and were consultants (96%). More participants reported feeling optimistic (10/25) than neutral (6/25), pessimistic (2/25) or mixed (7/25) towards the use of AI in the ED. A minority expressed scepticism regarding the feasibility or value of implementing AI into the ED. Multiple potential risks and ethical issues were discussed by participants including skill loss from overreliance on AI, algorithmic bias, patient privacy and concerns over liability. Participants also discussed perceived inadequacies in existing information technology systems. Participants felt that AI technologies would be used as decision support tools and not replace the roles of emergency clinicians. Participants were not concerned about the impact of AI on their job security. Most (17/25) participants thought that AI would impact emergency medicine within the next 10 years. CONCLUSIONS: Emergency clinicians interviewed were generally optimistic about the use of AI in emergency medicine, so long as it is used as a decision support tool and they maintain the ability to override its recommendations.

Applications of natural language processing at emergency department triage: A narrative review.

INTRODUCTION: Natural language processing (NLP) uses various computational methods to analyse and understand human language, and has been applied to data acquired at Emergency Department (ED) triage to predict various outcomes. The objective of this scoping review is to evaluate how NLP has been applied to data acquired at ED triage, assess if NLP based models outperform humans or current risk stratification techniques when predicting outcomes, and assess if incorporating free-text improve predictive performance of models when compared to predictive models that use only structured data. METHODS: All English language peer-reviewed research that applied an NLP technique to free-text obtained at ED triage was eligible for inclusion. We excluded studies focusing solely on disease surveillance, and studies that used information obtained after triage. We searched the electronic databases MEDLINE, Embase, Cochrane Database of Systematic Reviews, Web of Science, and Scopus for medical subject headings and text keywords related to NLP and triage. Databases were last searched on 01/01/2022. Risk of bias in studies was assessed using the Prediction model Risk of Bias Assessment Tool (PROBAST). Due to the high level of heterogeneity between studies and high risk of bias, a metanalysis was not conducted. Instead, a narrative synthesis is provided. RESULTS: In total, 3730 studies were screened, and 20 studies were included. The population size varied greatly between studies ranging from 1.8 million patients to 598 triage notes. The most common outcomes assessed were prediction of triage score, prediction of admission, and prediction of critical illness. NLP models achieved high accuracy in predicting need for admission, triage score, critical illness, and mapping free-text chief complaints to structured fields. Incorporating both structured data and free-text data improved results when compared to models that used only structured data. However, the majority of studies (80%) were assessed to have a high risk of bias, and only one study reported the deployment of an NLP model into clinical practice. CONCLUSION: Unstructured free-text triage notes have been used by NLP models to predict clinically relevant outcomes. However, the majority of studies have a high risk of bias, most research is retrospective, and there are few examples of implementation into clinical practice. Future work is needed to prospectively assess if applying NLP to data acquired at ED triage improves ED outcomes when compared to usual clinical practice.

Western Australian medical students' attitudes towards artificial intelligence in healthcare.

INTRODUCTION: Surveys conducted internationally have found widespread interest in artificial intelligence (AI) amongst medical students. No similar surveys have been conducted in Western Australia (WA) and it is not known how medical students in WA feel about the use of AI in healthcare or their understanding of AI. We aim to assess WA medical students' attitudes towards AI in general, AI in healthcare, and the inclusion of AI education in the medical curriculum. METHODS: A digital survey instrument was developed based on a review of available literature and consultation with subject matter experts. The survey was piloted with a group of medical students and refined based on their feedback. We then sent this anonymous digital survey to all medical students in WA (approximately 1539 students). Responses were open from the 7th of September 2021 to the 7th of November 2021. Students' categorical responses were qualitatively analysed, and free text comments from the survey were qualitatively analysed using open coding techniques. RESULTS: Overall, 134 students answered one or more questions (8.9% response rate). The majority of students (82.0%) were 20-29 years old, studying medicine as a postgraduate degree (77.6%), and had started clinical rotations (62.7%). Students were interested in AI (82.6%), self-reported having a basic understanding of AI (84.8%), but few agreed that they had an understanding of the basic computational principles of AI (33.3%) or the limitations of AI (46.2%). Most students (87.5%) had not received teaching in AI. The majority of students (58.6%) agreed that AI should be part of medical training and most (72.7%) wanted more teaching focusing on AI in medicine. Medical students appeared optimistic regarding the role of AI in medicine, with most (74.4%) agreeing with the statement that AI will improve medicine in general. The majority (56.6%) of medical students were not concerned about the impact of AI on their job security as a doctor. Students selected radiology (72.6%), pathology (58.2%), and medical administration (44.8%) as the specialties most likely to be impacted by AI, and psychiatry (61.2%), palliative care (48.5%), and obstetrics and gynaecology (41.0%) as the specialties least likely to be impacted by AI. Qualitative analysis of free text comments identified the use of AI as a tool, and that doctors will not be replaced as common themes. CONCLUSION: Medical students in WA appear to be interested in AI. However, they have not received education about AI and do not feel they understand its basic computational principles or limitations. AI appears to be a current deficit in the medical curriculum in WA, and most students surveyed were supportive of its introduction. These results are consistent with previous surveys conducted internationally.

Deep learning model to predict exercise stress test results: Optimizing the diagnostic test selection strategy and reduce wastage in suspected coronary artery disease patients.

BACKGROUND: Cardiac exercise stress testing (EST) offers a non-invasive way in the management of patients with suspected coronary artery disease (CAD). However, up to 30% EST results are either inconclusive or non-diagnostic, which results in significant resource wastage. Our aim was to build machine learning (ML) based models, using patients demographic (age, sex) and pre-test clinical information (reason for performing test, medications, blood pressure, heart rate, and resting electrocardiogram), capable of predicting EST results beforehand including those with inconclusive or non-diagnostic results. METHODS: A total of 30,710 patients (mean age 54.0 years, 69% male) were included in the study with 25% randomly sampled in the test set, and the remaining samples were split into a train and validation set with a ratio of 9:1. We constructed different ML models from pre-test variables and compared their discriminant power using the area under the receiver operating characteristic curve (AUC). RESULTS: A network of Oblivious Decision Trees provided the best discriminant power (AUC=0.83, sensitivity=69%, specificity=0.78%) for predicting inconclusive EST results. A total of 2010 inconclusive ESTs were correctly identified in the testing set. CONCLUSIONS: Our ML model, developed using demographic and pre-test clinical information, can accurately predict EST results and could be used to identify patients with inconclusive or non-diagnostic results beforehand. Our system could thus be used as a personalised decision support tool by clinicians for optimizing the diagnostic test selection strategy for CAD patients and to reduce healthcare expenditure by reducing nondiagnostic or inconclusive ESTs.

Performance of multilabel machine learning models and risk stratification schemas for predicting stroke and bleeding risk in patients with non-valvular atrial fibrillation.

BACKGROUND: Appropriate anticoagulant therapy for patients with atrial fibrillation (AF) requires assessment of stroke and bleeding risks. However, risk stratification schemas such as CHA2DS2-VASc and HAS-BLED have modest predictive capacity for patients with AF. Multilabel machine learning (ML) techniques may improve predictive performance and support decision-making for anticoagulant therapy. We compared the performance of multilabel ML models with the currently used risk scores for predicting outcomes in AF patients. METHODS: This was a retrospective cohort study of 9670 patients, mean age 76.9 years, 46% women, who were hospitalized with non-valvular AF, and had 1-year follow-up. The outcomes were ischemic stroke (167), major bleeding (430) admissions, all-cause death (1912) and event-free survival (7387). Discrimination and calibration of ML models were compared with clinical risk scores by area under the curve (AUC). Risk stratification was assessed using net reclassification index (NRI). RESULTS: Multilabel gradient boosting classifier chain provided the best AUCs for stroke (0.685 95% CI 0.676, 0.694), major bleeding (0.709 95% CI 0.703, 0.716) and death (0.765 95% CI 0.763, 0.768) compared to multi-layer neural networks and classifier chain using support vector machine. It provided modest performance improvement for stroke compared to AUC of CHA2DS2-VASc (0.652, NRI = 3.2%, p-value = 0.1), but significantly improved major bleeding prediction compared to AUC of HAS-BLED (0.522, NRI = 22.8%, p-value < 0.05). It also achieved greater discriminant power for death compared with AUC of CHA2DS2-VASc (0.606, p-value < 0.05). ML models identified additional risk features such as hemoglobin level, renal function. CONCLUSIONS: Multilabel ML models can outperform clinical risk stratification scores for predicting the risk of major bleeding and death in non-valvular AF patients.

Applications of machine learning to undifferentiated chest pain in the emergency department: A systematic review.

BACKGROUND: Chest pain is amongst the most common reason for presentation to the emergency department (ED). There are many causes of chest pain, and it is important for the emergency physician to quickly and accurately diagnose life threatening causes such as acute myocardial infarction (AMI). Multiple clinical decision tools have been developed to assist clinicians in risk stratifying patients with chest. There is growing recognition that machine learning (ML) will have a significant impact on the practice of medicine in the near future and may assist with diagnosis and risk stratification. This systematic review aims to evaluate how ML has been applied to adults presenting to the ED with undifferentiated chest pain and assess if ML models show improved performance when compared to physicians or current risk stratification techniques. METHODS AND FINDINGS: We conducted a systematic review of journal articles that applied a ML technique to an adult patient presenting to an emergency department with undifferentiated chest pain. Multiple databases were searched from inception through to November 2020. In total, 3361 articles were screened, and 23 articles were included. We did not conduct a metanalysis due to a high level of heterogeneity between studies in both their methods, and reporting. The most common primary outcomes assessed were diagnosis of acute myocardial infarction (AMI) (12 studies), and prognosis of major adverse cardiovascular event (MACE) (6 studies). There were 14 retrospective studies and 5 prospective studies. Four studies reported the development of a machine learning model retrospectively then tested it prospectively. The most common machine learning methods used were artificial neural networks (14 studies), random forest (6 studies), support vector machine (5 studies), and gradient boosting (2 studies). Multiple studies achieved high accuracy in both the diagnosis of AMI in the ED setting, and in predicting mortality and composite outcomes over various timeframes. ML outperformed existing risk stratification scores in all cases, and physicians in three out of four cases. The majority of studies were single centre, retrospective, and without prospective or external validation. There were only 3 studies that were considered low risk of bias and had low applicability concerns. Two studies reported integrating the ML model into clinical practice. CONCLUSIONS: Research on applications of ML for undifferentiated chest pain in the ED has been ongoing for decades. ML has been reported to outperform emergency physicians and current risk stratification tools to diagnose AMI and predict MACE but has rarely been integrated into practice. Many studies assessing the use of ML in undifferentiated chest pain in the ED have a high risk of bias. It is important that future studies make use of recently developed standardised ML reporting guidelines, register their protocols, and share their datasets and code. Future work is required to assess the impact of ML model implementation on clinical decision making, patient orientated outcomes, and patient and physician acceptability. TRIAL REGISTRATION: International Prospective Register of Systematic Reviews registration number: CRD42020184977.

Artificial intelligence-enhanced echocardiography in the emergency department.

A focused cardiac ultrasound performed by an emergency physician is becoming part of the standard assessment of patients in a variety of clinical situations. The development of inexpensive, portable handheld devices promises to make point-of-care ultrasound even more accessible over the coming decades. Many of these handheld devices are beginning to integrate artificial intelligence (AI) for image analysis. The integration of AI into focused cardiac ultrasound will have a number of implications for emergency physicians. This perspective presents an overview of the current state of AI research in echocardiography relevant to the emergency physician, as well as the future possibilities, challenges and risks of this technology.

Utility of Emergency Department Chest Imaging in Patients with Cancer: A Descriptive Study.

BACKGROUND: The use of computed tomography (CT) has been scrutinized in emergency medicine, particularly in patients with cancer. Previous studies have characterized the rate of CT use in this population; however, limited data are available about the yield of this modality compared with radiography and its clinical decision-making effect. OBJECTIVE: To determine whether CT imaging of the chest increases identification of clinically significant results compared with chest radiography (CXR) in patients with cancer. METHODS: This was a retrospective chart review of patients with a history of solid tumors presenting to an emergency department in 2017. Patients who received both CXR and CT (or CT angiography) of the chest during their assessment were identified and the rate of clinically significant findings on imaging was compared. Clinical findings were further categorized as requiring nonurgent, urgent, or emergent attention. Descriptive statistics and chi-squared testing were performed between the 2 imaging modalities. RESULTS: From 839 patients meeting inclusion criteria, 287 were randomly sampled. The predominant malignancies were lung (32.4%), breast (13.9%), and head and neck cancer (13.6%). A greater number of patients had clinically significant findings identified on CT imaging (n = 222) compared with CXR (n = 108). Stratification upon urgency of these findings (nonurgent, urgent, or emergent) reveals a significant difference in all strata (p < 0.05). CONCLUSIONS: Compared with CXR, CT imaging of the chest identified significantly more clinically relevant findings requiring attention and consequently affecting clinical decision making.

The effect of guideline implementation on discharge analgesia prescribing (two years on).

The provision of appropriate discharge analgesia can be challenging and is often prescribed by some of the most junior members of the medical team. Opioid abuse has been considered a growing public health crisis and physician overprescribing is a major contributor. In 2015 an initial audit of discharge analgesia at the Royal Perth Hospital led to the development of discharge analgesia guidelines. Compliance with these guidelines was assessed by a follow-up audit in 2016, which showed improved practice. This audit assesses discharge analgesia prescribing practices two years following guideline implementation. Dispensing data were obtained for analgesic medication over a three-month period from April to July 2017 and 100 unique patients were chosen using computer generated randomisation. Patients' medical records were assessed against the hospital's Postoperative Inpatients Discharge Analgesia Guidelines. The data collected were then compared with equivalent data from the previous 2015 and 2016 audits. Overall 83.4% of the 170 discharge analgesia prescriptions written were compliant with guidelines. The highest overall compliance rates were achieved for paracetamol (100%, up from 95.9% in 2016), celecoxib (96%, down from 100% in 2016), and oxycodone immediate release (IR) (74%, down from 88.9% in 2016). The quantity of oxycodone IR given on discharge complied with quantity guidelines in only 56% of cases. Overall there has been a significant and sustained improvement in appropriateness of discharge analgesia prescribing since 2015, though the results from 2017 show less compliance than 2016 and that achieving compliance with quantity guidelines is an ongoing challenge. This demonstrates the challenge of obtaining high adherence to guidelines over a longer time period.

Artificial intelligence and machine learning in emergency medicine.

Interest in artificial intelligence (AI) research has grown rapidly over the past few years, in part thanks to the numerous successes of modern machine learning techniques such as deep learning, the availability of large datasets and improvements in computing power. AI is proving to be increasingly applicable to healthcare and there is a growing list of tasks where algorithms have matched or surpassed physician performance. Despite the successes there remain significant concerns and challenges surrounding algorithm opacity, trust and patient data security. Notwithstanding these challenges, AI technologies will likely become increasingly integrated into emergency medicine in the coming years. This perspective presents an overview of current AI research relevant to emergency medicine.