A Comparison of Artificial Intelligence and Human Doctors for the Purpose of Triage and Diagnosis.

AI virtual assistants have significant potential to alleviate the pressure on overly burdened healthcare systems by enabling patients to self-assess their symptoms and to seek further care when appropriate. For these systems to make a meaningful contribution to healthcare globally, they must be trusted by patients and healthcare professionals alike, and service the needs of patients in diverse regions and segments of the population. We developed an AI virtual assistant which provides patients with triage and diagnostic information. Crucially, the system is based on a generative model, which allows for relatively straightforward re-parameterization to reflect local disease and risk factor burden in diverse regions and population segments. This is an appealing property, particularly when considering the potential of AI systems to improve the provision of healthcare on a global scale in many regions and for both developing and developed countries. We performed a prospective validation study of the accuracy and safety of the AI system and human doctors. Importantly, we assessed the accuracy and safety of both the AI and human doctors independently against identical clinical cases and, unlike previous studies, also accounted for the information gathering process of both agents. Overall, we found that the AI system is able to provide patients with triage and diagnostic information with a level of clinical accuracy and safety comparable to that of human doctors. Through this approach and study, we hope to start building trust in AI-powered systems by directly comparing their performance to human doctors, who do not always agree with each other on the cause of patients' symptoms or the most appropriate triage recommendation.

Bronchiolitis at a specialist paediatric centre: The electronic medical record helps to evaluate low-value care.

AIM: Low-value care (LVC) is common. We aimed, using infants presenting to a major tertiary paediatric hospital with bronchiolitis between April 2016 and July 2018, to: (i) assess rates of chest X-ray (CXR) and medication use; (ii) identify associated factors; and (iii) measure the harm of not performing these practices. METHODS: We extracted data from the electronic medical record for all children aged 1-12 months given a diagnosis of bronchiolitis in the emergency department. Factors potentially associated with LVC practices were extracted, including patient demographics, ordering physician characteristics, order indication, medications prescribed and admission ward. To assess for harm, a radiologist, blinded to CXR indication, reviewed all CXRs ordered over the winter of 2017 for infants with bronchiolitis. RESULTS: A CXR was ordered for 439 (11.2%) infants, most commonly to rule out consolidation and collapse (65%). CXRs were more likely to be ordered for admitted infants (40.9% admitted to the general medical ward), and 62% were ordered by emergency department staff. Salbutamol was prescribed for 9.3% (n = 199). Amongst those who had a CXR, 28% were prescribed an antibiotic compared to 2.1% for those who did not. In an audit of 98 CXRs ordered over the winter of 2017, there were no CXR findings that meaningfully affected patient outcomes. CONCLUSION: Using electronic medical record data, we found that CXR and medication use in bronchiolitis were higher than expected given our hospital guideline advice. Future research needs to understand why and develop interventions to reduce LVC.

An immersive "simulation week" enhances clinical performance of incoming surgical interns improved performance persists at 6 months follow-up.

BACKGROUND: The transition from student to intern can be challenging. The "August" or "July effect" describes increased errors and reduced patient safety during this transition. The study objectives were to develop, pilot, and evaluate clinical performance after an immersive simulation course for incoming interns. METHODS: Graduating students were recruited for a 1-week immersive simulation course. Controls received no simulation training. Primary outcome (at baseline, and 1 and 6 months) was clinical performance on Objective Structured Clinical Examinations (OSCE) of clinical procedures and surgical technical skills. Secondary outcomes were self-reported confidence and clinical procedure logbook data. RESULTS: Nineteen students were recruited. Sixteen completed the 6-month follow-up, 10 in the intervention group and 6 in the control group. No differences were demonstrated between interventions and controls at baseline (OSCE [median, 66 vs. 78; P = .181], technical skills [48 vs. 52.5; P = .381], and confidence [101 vs 96; P = .368]). Interventions outperformed controls at 1 month (OSCE [111 vs 82; P = .001], technical skills [78.5 vs 63; P = .030], and confidence [142 vs. 119; P < .001]), and 6 months (OSCE [107 vs. 93; P = .007], technical skills [92.5 vs. 69; P = .044], and confidence [148 vs. 129; P = .022]). No differences were observed in numbers of clinical procedures performed at 1 (P = .958), 4 (P = .093), or 6 months (P = .713). CONCLUSION: The immersive simulation course objectively improved subjects' clinical skills, technical skills, and confidence. Despite similar clinical experience as controls, the intervention group's improved performance persisted at 6 months follow-up. This feasible and effective intervention to ease transition from student to intern could reduce errors and enhance patient safety.

Resumption of work after acute coronary syndrome or coronary artery bypass graft surgery.

BACKGROUND: Return to work is an important indicator of recovery after acute cardiac events. This study aimed to determine rates of work resumption and identify predictors of non-return to work and delayed resumption of work. METHODS: 401 currently employed patients consecutively admitted after acute coronary syndrome or to undergo coronary artery bypass graft surgery were recruited. Patient characteristics, perceptions and occupational outcomes were investigated via interviews and self-report questionnaires. RESULTS: Twenty-three patients were lost to follow-up. Of the 378 completers, 343 (90.7%) patients resumed work, while 35 (9.3%) did not. By four months, 309 (91.1%) patients had returned to work. At 12 months, 302 (79.9%) of the 378 patients were employed, 32 (8.5%) unemployed and 20 (5.3%) retired. The employment status of 24 (6.3%) patients was unknown. Non-return to work was significantly more likely if patients were not intending to return to work or were uncertain, had a negative perception of health, had a comorbidity other than diabetes and reported financial stress. Significant predictors of delayed return to work were cardiac rehabilitation attendance, longer hospital stay, past angina, having a manual job, physically active work, job dissatisfaction, no confidante and depression. CONCLUSIONS: Patients at risk of poor occupational outcomes can be identified early. Strategies to improve vocational rehabilitation require further investigation.

Assessment of left ventricular diastolic function by early diastolic mitral annulus peak acceleration rate: experimental studies and clinical application.

We sought to examine the hemodynamic determinants and clinical application of the peak acceleration rate of early (Ea) diastolic velocity of the mitral annulus by tissue Doppler. Simultaneous left atrial and left ventricular (LV) catheterization and Doppler echocardiography were performed in 10 dogs. Preload was altered using volume infusion and caval occlusion, whereas myocardial lusitropic state was altered with dobutamine and esmolol. The clinical application was examined in 190 consecutive patients (55 control, 41 impaired relaxation, 46 pseudonormal, and 48 restrictive LV filling). In addition, in 60 consecutive patients, we examined the relation between it and mean wedge pressure with simultaneous Doppler echocardiography and right heart catheterization. In canine studies, a significant positive relation was present between peak acceleration rate of Ea and transmitral pressure gradient only in the stages with normal or enhanced LV relaxation, but with no relation in the stages where the time constant of LV relaxation (tau) was > or =50 ms. Its hemodynamic determinants were tau, LV minimal pressure, and transmitral pressure gradient. In clinical studies, peak acceleration rate of Ea was significantly lower in patients with impaired LV relaxation irrespective of filling pressures (P < 0.001) and with similar accuracy to peak Ea velocity (area under the curve for septal and lateral peak acceleration rates: both 0.78) in identifying these patients. No significant relation was observed between peak acceleration rate and mean wedge pressure. Peak acceleration rate of Ea appears to be a useful index of LV relaxation but not of filling pressures and can be applied to identify patients with impaired LV relaxation irrespective of their filling pressures.

Haemodynamic insights into the effects of ischaemia and cycle length on tissue Doppler-derived mitral annulus diastolic velocities.

In the present study, we performed simultaneous epicardial echocardiography and left heart catheterization on ten adult dogs to investigate the effects of ischaemia and tachycardia on the mitral annulus early (E(a)) and late (A(a)) diastolic velocities and the haemodynamic mechanisms involved. Left atrial pressure and left ventricular (LV) volumes and pressures were measured with 5 French Millar catheters. In each dog, inferior vena cava occlusion was used to alter preload and circumflex coronary artery occlusion was applied to induce ischaemia at two different cycle lengths: 450 and 550 ms. At both cycle lengths, ischaemia resulted in a reduction in LV relaxation, LV global and ipsilateral systolic function, transmitral pressure gradient (TMG), E(a) and A(a) ( P <0.05). The shorter cycle length was associated with a shorter tau (time constant of LV relaxation), reduced TMG and reduced septal and lateral E(a) ( P <0.05 for all variables). Both septal and lateral A(a) were significantly increased ( P <0.05). Ischaemia influences E(a) through changes in LV relaxation, global and regional systolic function and TMG. An increase in heart rate reduces E(a), but increases A(a).

Acute effect of nonsurgical septal reduction therapy on regional left ventricular asynchrony in patients with hypertrophic obstructive cardiomyopathy.

BACKGROUND: Patients with hypertrophic obstructive cardiomyopathy have left ventricular (LV) diastolic dysfunction due, in part, to temporal heterogeneity in regional function. The acute effect of the relief of LV outflow tract obstruction is unknown. Therefore, we investigated the effects of nonsurgical septal reduction therapy (NSRT) on regional function. METHODS AND RESULTS: Twenty-two patients (aged 56+/-17 years) underwent echocardiographic examination, including tissue Doppler imaging, simultaneously with left heart catheterization before and after NSRT. LV regional function was assessed in 12 segments from which myocardial strain was obtained. Asynchrony was calculated as the coefficient of variation of the time interval from the QRS complex to the onset of expansion and to early diastolic strain. After NSRT, a significant reduction in LV outflow tract obstruction (from 57+/-5 to 12+/-3 mm Hg) occurred with shortening of the time constant of LV relaxation (71+/-4 to 61+/-3 ms; both P<0.05). The coefficient of variation of the time interval to onset of regional expansion decreased significantly and related well to the changes in the time constant of LV relaxation (r=0.81, P<0.01). CONCLUSIONS: NSRT has a favorable effect on LV regional asynchrony, which accounts for the acute changes in LV relaxation.