Analysis of 'One in a Million' primary care consultation conversations using natural language processing

BACKGROUND

Modern patient electronic health records form a core part of primary care; they contain both clinical codes and free text entered by the clinician. Natural language processing (NLP) could be employed to generate these records through 'listening' to a consultation conversation.

OBJECTIVES

This study develops and assesses several text classifiers for identifying clinical codes for primary care consultations based on the doctor-patient conversation. We evaluate the possibility of training classifiers using medical code descriptions, and the benefits of processing transcribed speech from patients as well as doctors. The study also highlights steps for improving future classifiers.

METHODS

Using verbatim transcripts of 239 primary care consultation conversations (the 'One in a Million' dataset) and novel additional datasets for distant supervision, we trained NLP classifiers (naïve Bayes, support vector machine, nearest centroid, a conventional BERT classifier and few-shot BERT approaches) to identify the International Classification of Primary Care-2 clinical codes associated with each consultation.

RESULTS

Of all models tested, a fine-tuned BERT classifier was the best performer. Distant supervision improved the model's performance (F1 score over 16 classes) from 0.45 with conventional supervision with 191 labelled transcripts to 0.51. Incorporating patients' speech in addition to clinician's speech increased the BERT classifier's performance from 0.45 to 0.55 F1 (p=0.01, paired bootstrap test).

CONCLUSIONS

Our findings demonstrate that NLP classifiers can be trained to identify clinical area(s) being discussed in a primary care consultation from audio transcriptions; this could represent an important step towards a smart digital assistant in the consultation room.

Estimands for atopic dermatitis clinical trials: Expert opinion on the importance of intercurrent events

Despite the emergence of novel targeted treatments for atopic dermatitis (AD), there is a lack of guidelines on standardizing analysis of clinical trial data. To define and estimate meaningful treatment comparisons, several factors, including intercurrent events, must be taken into account. Intercurrent events are defined as events occurring after treatment initiation that affect either the interpretation or existence of the measurements associated with clinical questions of interest. Due to the relapsing, unpredictable nature of AD, intercurrent events frequently occur in AD trials, such as use of rescue therapy for intense itch and sleep deprivation. Despite the impact of intercurrent events in AD, they are often handled in an inconsistent manner across trials, which limits results interpretation. The estimand framework is increasingly used to estimate treatment effects while accounting for intercurrent events. This review explores how guidance from the International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH) on the use of estimands can be applied to support AD clinical trial design and analysis. We propose that estimands are used in AD trials and defined early during trial design. The use of estimands can provide clinicians with interventional trial results that are more reflective of clinical practice, help facilitate comparisons across clinical trials, and are more informative to enable improved treatment selection for patients.

Sickle cell bone disease and response to intravenous bisphosphonates in children

Children with sickle cell disease (SCD) have the potential for extensive and early-onset bone morbidity. This study reports on the diversity of bone morbidity seen in children with SCD followed at three tertiary centers. IV bisphosphonates were effective for bone pain analgesia and did not trigger sickle cell complications.

INTRODUCTION

To evaluate bone morbidity and the response to intravenous (IV) bisphosphonate therapy in children with SCD.

METHODS

We conducted a retrospective review of patient records from 2003 to 2019 at three Canadian pediatric tertiary care centers. Radiographs, magnetic resonance images, and computed tomography scans were reviewed for the presence of avascular necrosis (AVN), bone infarcts, and myositis. IV bisphosphonates were offered for bone pain management. Bone mineral density was assessed by dual-energy X-ray absorptiometry (DXA).

RESULTS

Forty-six children (20 girls, 43%) had bone morbidity at a mean age of 11.8 years (SD 3.9) including AVN of the femoral (17/46, 37%) and humeral (8/46, 17%) heads, H-shaped vertebral body deformities due to endplate infarcts (35/46, 76%), and non-vertebral body skeletal infarcts (15/46, 32%). Five children (5/26, 19%) had myositis overlying areas of AVN or bone infarcts visualized on magnetic resonance imaging. Twenty-three children (8/23 girls) received IV bisphosphonate therapy. They all reported significant or complete resolution of bone pain. There were no reports of sickle cell hemolytic crises, pain crises, or stroke attributed to IV bisphosphonate therapy.

CONCLUSION

Children with SCD have the potential for extensive and early-onset bone morbidity. In this series, IV bisphosphonates were effective for bone pain analgesia and did not trigger sickle cell complications.

Online KidClot education for patients and families initiating warfarin therapy: The eKITE study

Anticoagulation with Vitamin K antagonists (VKA) has always posed challenges in terms of monitoring requirements. These challenges were further exacerbated in the setting of the COVID-19 pandemic, with limited access to and/or avoidance of laboratory testing. The importance of utilizing point of care (POC) health technology for individualized patient management is salient. The foundation of effective home INR monitoring is establishing patient knowledge about their therapy and INR testing proficiency. The eKITE series was developed to support patients in establishing foundational knowledge required for VKA (warfarin) management and INR monitoring. The primary objectives were to evaluate eKITE, a patient-oriented innovative online education program for warfarin therapy, participant learning stress, and patient preference for online learning. This multi-center prospective study provided patients access to warfarin online education. Participants were required to complete written quizzes assessing warfarin knowledge of key concepts proficiency and identifying knowledge deficits. Patient preference, evaluating calm (lack of anxiety) while learning, and an INR on a home meter was completed. Participants performed INR tests at home and reported INRs by telephone. The analysis included 144 children and caregivers enrolled at five US and CDN sites. Most indications for anticoagulation were cardiac (congenital or acquired heart disease) with varied INR target ranges. Mean knowledge scores for warfarin and INR self-testing modules were 97%, with low anxiety with TTR of 84%. Patient preferred online learning. eKITE is an effective teaching modality for warfarin/home INR monitoring with safe INR testing and warfarin management that is appropriate for pediatrics and adults alike. PROLOGUE: The whir in the hallways is deafening. Lights bright, alarms are ringing in a chorus of unsynchronized beeps and screeches. It has been more than a week since I have slept. Snuggled beside me is my precious child, whining and equally irritated with the asynchronous symphony, further compounded by anxiety, procedures, and pain. The sun has broken. The staff smiles are welcoming and incessant, as one after one, they approach hurried, urgent, assiduous, their need to coach me for our upcoming departure to the warmth of home. Each provides essential information that I will require to keep my child, my treasure, safe and healthy. Yet, my eyes are heavy, blurred, and my brain foggy, trapped in a dark heavy cloud. How am I to follow? Comprehend? and retain anything? As they instruct, my precious child yearns for loving arms, compassion and love, whining, crying in disquiet. Overwhelmed does not adequately describe my ineffable exhaustion. Amidst this, how am I to learn about warfarin? Such a challenging medication, with so much to know. Concentrate, I tell myself, focus; now is my only opportunity to learn. I must be alert. It seems to be nonsensical.

User attitudes towards virtual home assessment technologies

Telehealth has long been highlighted as a way to solve issues of efficiency and effectiveness in healthcare and to improve patients' care and has become fundamental to address patients' needs during the COVID-19 pandemic; however previous studies have shown mixed results in the user acceptance of such technologies. Whilst many previous studies have focussed on clinical application of telehealth, we focus on the adoption of telehealth for virtual assessments visits aimed to evaluate the suitability of a property where a patient is discharged, and eventual adaptations needed. We present a study of stakeholders' attitudes towards such virtual assessment visits. The study has been carried out with healthcare professionals and patients and allowed us to identify user attitudes, barriers and facilitators for the success of virtual assessment visits from the point of view of healthcare professionals and patients. Finally, we discuss implications for designers of telehealth services and guidelines that can be derived from our study.

Pioneering Global Best Practices in Atopic Dermatitis: Results from the Atopic Dermatitis Quality of Care Initiative

BACKGROUND

Atopic dermatitis (AD) is a chronic inflammatory skin disease characterized by unrelenting pruritus and recurrent eczematous lesions. It affects up to 15% of children and adolescents and up to 5% of adults and confers a high and multifactorial burden to patients, families, and society. With increasing awareness of this substantial burden, AD has become a priority for healthcare systems.

OBJECTIVE

The Atopic Dermatitis Quality of Care (ADQoC) Initiative set out to describe good practices for addressing the challenges that impede the management of AD.

METHODS

The Initiative carried out a literature review and surveyed 32 expert care centres, catalogued findings, and analysed and elucidated global challenges to AD care along with good practice implementations that can address them.

RESULTS

The four challenges to quality care for AD are: misconceptions about AD; delayed referral and access to AD specialists; poor patient access to AD treatments and poor adherence to medications; and managing the complexity of atopic dermatitis and its comorbidities. The Initiative highlighted five of ten good practice implementations as high priority for any AD care centre to focus: clinical assessment and diagnosis; a structured multidisciplinary care team; monitoring and evaluating care quality; patient education and communication; and collaboration and exchange with patient groups.

CONCLUSIONS

These implementations can provide benefits for patients, healthcare providers, and the healthcare system. They directly contribute to the efficacy of treatment; improved healthcare provider efficiency; improved education; and improved costs to healthcare systems. The Initiative was launched on https://atopicdermatitiscare.kpmg.co.uk/ to provide an easy-to-use educational platform.

Recommended core outcome instruments for health-related quality of life, long-term control and itch intensity in atopic eczema trials: results of the HOME VII consensus meeting

Background

The Harmonising Outcome Measures for Eczema (HOME) initiative has established a core outcome set of domains for atopic eczema (AE) clinical trials. Previous consensus meetings have agreed on preferred instruments for clinician‐reported signs (Eczema Area and Severity Index, EASI) and patient‐reported symptoms (Patient‐Oriented Eczema Measure, POEM). This paper reports consensus decisions from the HOME VII meeting.

Objectives

To complete the core outcome set for AE by agreeing on core outcome instruments for the domains of quality of life (QoL), long‐term control and itch intensity.

Methods

A face‐to‐face consensus meeting was held in Tokyo, Japan (8–10 April 2019) including 75 participants (49 healthcare professionals/methodologists, 14 patients, 12 industry representatives) from 16 countries. Consensus decisions were made by presentations of evidence, followed by whole and small group discussions and anonymous voting using predefined consensus rules.

Results

It was agreed by consensus that QoL should be measured using the Dermatology Life Quality Index (DLQI) for adults, the Children’s Dermatology Life Quality Index (CDLQI) for children and the Infant’s Dermatology Quality of Life Index (IDQoL) for infants. For long‐term control, the Recap of Atopic Eczema (RECAP) instrument or the Atopic Dermatitis Control Test (ADCT) should be used. Consensus was not reached over the frequency of data collection for long‐term control. The peak itch numerical rating scale (NRS)‐11 past 24 h was recommended as an additional instrument for the symptom domain in trials of older children and adults. Agreement was reached that all core outcome instruments should be captured at baseline and at the time of primary outcome assessment as a minimum.

Conclusions

For now, the core outcome set for clinical trials in AE is complete. The specified domains and instruments should be used in all new clinical trials and systematic reviews of eczema treatments.

What is already known about this topic?

Core outcomes sets improve the design and reporting of clinical trials, reduce selective outcome reporting bias and facilitate meta‐analysis of results in systematic reviews.

The HOME core outcome set for eczema recommends the inclusion of four core domains in all atopic eczema trials: clinician‐reported signs, patient‐reported symptoms, health‐related quality of life (HrQoL) and long‐term control.

Clinician‐reported signs should be captured using the Eczema Area and Severity Index (EASI) and patient‐reported symptoms using the Patient‐Oriented Eczema Measure (POEM).

What does this study add?

The HOME core outcome set is now complete and recommended core outcome instruments have been agreed on for all four domains.

Core outcome instruments for HrQoL: Dermatology Life Quality Index (DLQI) for adults, Children’s Dermatology Life Quality Index (CDLQI) for children and Infant’s Dermatology Quality of Life Index (IDQoL) for infants.

Core outcome instruments for long‐term control: either the Recap of Atopic Eczema (RECAP) or the Atopic Dermatitis Control Test (ADCT).

In addition, itch intensity should be measured using the peak NRS‐11 past 24 h for trials including older children and adults.

What are the clinical implications of this work?

If all future trials of eczema treatments include the HOME core outcome instruments, then trial results will be more readily incorporated into meta‐analyses in systematic reviews and clinical care will be informed by the best available evidence.

Linked Comment: D.F. Murrell and C.F. Paul. Br J Dermatol 2021; 185:13–14.

Recommended core outcome instruments for health-related quality of life, long-term control and itch intensity in atopic eczema trials: results of the HOME VII consensus meeting

BACKGROUND

The Harmonising Outcome Measures for Eczema (HOME) initiative has established a core outcome set of domains for atopic eczema clinical trials. Previous consensus meetings have agreed upon preferred instruments for clinician-reported signs (Eczema Area and Severity Index - EASI) and patient-reported symptoms (Patient-Oriented Eczema Measure - POEM). This paper reports consensus decisions from the HOME VII meeting.

OBJECTIVE

To complete the core outcome set for atopic eczema by agreeing upon core outcome instruments for the domains of quality of life, long-term control and itch intensity.

METHODS

Face-to-face consensus meeting held in Tokyo, Japan (8th to 10th April, 2019) including 74 participants (47 healthcare professionals/methodologists, 14 patients, 13 industry representatives), from 16 countries. Consensus decisions were made by presentations of evidence, followed by whole and small group discussions and anonymous voting using pre-defined consensus rules.

RESULTS

It was agreed by consensus that quality of life should be measured using the Dermatology Life Quality Index (DLQI) for adults, the Children's Dermatology Life Quality Index (CDLQI) for children, and the Infant's Dermatology Quality of Life Index (IDQoL) for infants. For long-term control, the Recap of Atopic Eczema (RECAP) instrument or the Atopic Dermatitis Control Test (ADCT) should be used. Consensus was not reached over the frequency of data collection for long-term control. The peak itch numerical rating scale(NRS)-11 past 24 hours was recommended as an additional instrument for the symptom domain in trials of older children and adults. Agreement was reached that all core outcome instruments should be captured at baseline and at the time of primary outcome assessment as a minimum.

CONCLUSIONS

For now, the core outcome set for clinical trials in atopic eczema is complete. The specified domains and instruments should be used in all new clinical trials and systematic reviews of eczema treatments.

Scalable Bayesian preference learning for crowds

We propose a scalable Bayesian preference learning method for jointly predicting the preferences of individuals as well as the consensus of a crowd from pairwise labels. Peoples’ opinions often differ greatly, making it difficult to predict their preferences from small amounts of personal data. Individual biases also make it harder to infer the consensus of a crowd when there are few labels per item. We address these challenges by combining matrix factorisation with Gaussian processes, using a Bayesian approach to account for uncertainty arising from noisy and sparse data. Our method exploits input features, such as text embeddings and user metadata, to predict preferences for new items and users that are not in the training set. As previous solutions based on Gaussian processes do not scale to large numbers of users, items or pairwise labels, we propose a stochastic variational inference approach that limits computational and memory costs. Our experiments on a recommendation task show that our method is competitive with previous approaches despite our scalable inference approximation. We demonstrate the method’s scalability on a natural language processing task with thousands of users and items, and show improvements over the state of the art on this task. We make our software publicly available for future work (https://github.com/UKPLab/tacl2018-preference-convincing/tree/crowdGPPL).

Lebrikizumab, a High-Affinity IL-13 Inhibitor, Improves Clinical Manifestations in Moderate-to-Severe Atopic Dermatitis: Primary Results From a Randomized, Double-Blinded, Placebo-Controlled, Dose-Ranging, Phase 2b Study

Abstract not available.

P3440Can pre-operative echocardiography predict RV failure post-LVAD implantation in children?

Background

One of the most frequent complications of left ventricular assist device (LVAD) implantation is the development of right ventricular (RV) failure, which occurs in 10%-40% of patients. It is, therefore, essential to identify which LVAD candidates are at risk for RV failure and will require additional RV mechanical support with a biventricular assist device (BiVAD). However, the echocardiographic assessment of the RV is challenging due to its complex geometry and marked load dependence of its function indices. To our knowledge, there is no published evidence regarding which specific RV echocardiographic parameters should inform this decision in the paediatric population.

Aims

We sought to determine which pre-operative RV echocardiographic parameters best correlate with the need for a BiVAD, as opposed to LVAD alone, in a paediatric population undergoing VAD insertion at a tertiary care institution.

Methods

Retrospective review of the pre-operative echocardiograms of children (<18 years) with dilated cardiomyopathy undergoing elective VAD implantation at our institution, from November 2007 to December 2018. Preoperative quantitative and qualitative RV echocardiographic parameters described in the literature to be associated with RV failure after LVAD implantation in adults were collected. Moreover, qualitative RV function was independently assessed by three echocardiographers, blinded to the outcome of the patients.

Results

89 patients were included, 39 (43.8%) males, median age = 1.7 years (IQR = 6.9), median weight = 11.6 kg (IQR = 13.8). 49 (55.1%) patients had an LVAD implanted, whereas 40 (44.9%) were deemed to need biventricular support. 45 (50.6%) patients received an EXCOR Berlin Heart, 16 (18%) a HeartWare HVAD and 28 (31.4%) a Levitronix Centrimag device. Requirement of BiVAD support was significantly more common in patients with moderate to severe RV impairment, as per expert assessment (OR = 2.864; 95% CI: 1.188–6.903, p=0.018), and tricuspid regurgitation > grade III (OR = 3.154; 95% CI: 1.124–8.850, p=0.025). All the other parameters collected – tricuspid annular plane systolic excursion (TAPSE), tricuspid regurgitant jet velocity, tricuspid regurgitation duration corrected for heart rate, RV tissue Doppler indices, RV fractional area change, and RV/LV diameter ratio– were not significantly different among groups.

Conclusions

In our paediatric population, expert assessment of RV function and degree of tricuspid regurgitation were strong predictors of RV failure among patients undergoing LVAD implantation, allowing for pre-emptive RVAD implantation and improving patient outcomes.

HLA Genotype of Patients with Severe Haemophilia A due to Intron 22 Inversion with and without Inhibitors of Factor VIII

Molecular genetic studies have shown that development of antibodies to factor VIII (inhibitors) occurs most frequently in patients with severe haemophilia due to major gene lesions including inversions, stop codons and large deletions. Previous studies of HLA type were performed on inhibitor and non-inhibitor patients with diverse uncharacterised mutations which may have confounded detection of significant associations. We therefore selected a group of patients with a single mutation type, the prevalent intron 22 inversion, with or without inhibitors, to determine HLA genotype. Seventy-one such patients, 42 without and 29 with inhibitors (13 high, 9 low and 7 transient responders) were genotyped for MHC Class I HLA-A, -B, -C and Class II HLA-DQA, -DQB and -DRB loci. No strong correlation of any HLA-allele to inhibitor or non-inhibitor status was found. However, alleles of the haplotype HLA-A3, HLA-B7, HLA-C7, HLA-DQA0102, HLA-DQB0602, HLA-DR15 occurred more often in inhibitor patients. Since the alleles of this extended haplotype are common in the North European population only a very strong association would achieve statistical significance. Further studies of groups of patients similar to those studied here will be needed to confirm or exclude this association.

Report from the fifth international consensus meeting to harmonize core outcome measures for atopic eczema/dermatitis clinical trials (HOME initiative)

This is the report from the fifth meeting of the Harmonising Outcome Measures for Eczema initiative (HOME V). The meeting was held on 12-14 June 2017 in Nantes, France, with 81 participants. The main aims of the meeting were (i) to achieve consensus over the definition of the core domain of long-term control and how to measure it and (ii) to prioritize future areas of research for the measurement of the core domain of quality of life (QoL) in children. Moderated whole-group and small-group consensus discussions were informed by presentations of qualitative studies, systematic reviews and validation studies. Small-group allocations were performed a priori to ensure that each group included different stakeholders from a variety of geographical regions. Anonymous whole-group voting was carried out using handheld electronic voting pads according to predefined consensus rules. It was agreed by consensus that the long-term control domain should include signs, symptoms, quality of life and a patient global instrument. The group agreed that itch intensity should be measured when assessing long-term control of eczema in addition to the frequency of itch captured by the symptoms domain. There was no recommendation of an instrument for the core outcome domain of quality of life in children, but existing instruments were assessed for face validity and feasibility, and future work that will facilitate the recommendation of an instrument was agreed upon.

Selective duodenography for controlled first-pass bolus distention of the duodenum in neonates and young children with bile-stained vomiting

AIM

To compare "selective duodenograms" performed through nasogastric tubes advanced into the proximal duodenum against traditional contrast studies regarding visualisation of a first-pass bolus distention of the duodenum and frequency of inconclusive results.

MATERIALS AND METHODS

Retrospective comparison of "selective duodenograms" and traditional upper gastrointestinal contrast studies in neonates with bile-stained vomiting, regarding demonstration of the duodenal C-loop, first-pass bolus capture, degree of distention of the duodenum, and number of inconclusive studies. Statistical comparison was performed using the two-tailed Fishers exact and chi-squared tests.

RESULTS

There were 31 "selective duodenograms" and 70 traditional studies. The C-loop was demonstrated in 93% of traditional studies versus 100% in "selective duodenograms" and was demonstrated significantly more often during the first-pass (94% versus 73%; p=0.018) and more often with distention of the duodenum for "selective duodenography" (94% versus 56%, p<0.001). There were more inconclusive traditional studies (7% versus 3%; non-significant).

CONCLUSION

Emergency upper gastrointestinal tract studies performed in neonates using the "selective duodenogram" technique demonstrated the duodenum with 100% success, with significantly more frequent first-pass bolus visualisation and duodenal distention than traditional studies. The five (7%) inconclusive traditional studies, present a significant diagnostic conundrum in the emergency setting because emergency surgery must be contemplated without proof.

A Disaster Response System based on Human-Agent Collectives

Major natural or man-made disasters such as Hurricane Katrina or the 9/11 terror attacks pose significant challenges for emergency responders. First, they have to develop an understanding of the unfolding event either using their own resources or through third-parties such as the local population and agencies. Second, based on the information gathered, they need to deploy their teams in a flexible manner, ensuring that each team performs tasks in The most effective way. Third, given the dynamic nature of a disaster space, and the uncertainties involved in performing rescue missions, information about the disaster space and the actors within it needs to be managed to ensure that responders are always acting on up-to-date and trusted information. Against this background, this paper proposes a novel disaster response system called HAC-ER. Thus HAC-ER interweaves humans and agents, both robotic and software, in social relationships that augment their individual and collective capabilities. To design HAC-ER, we involved end-users including both experts and volunteers in a several participatory design workshops, lab studies, and field trials of increasingly advanced prototypes of individual components of HAC-ER as well as the overall system. This process generated a number of new quantitative and qualitative results but also raised a number of new research questions. HAC-ER thus demonstrates how such Human-Agent Collectives (HACs) can address key challenges in disaster response. Specifically, we show how HAC-ER utilises crowdsourcing combined with machine learning to obtain most important situational awareness from large streams of reports posted by members of the public and trusted organisations. We then show how this information can inform human-agent teams in coordinating multi-UAV deployments, as well as task planning for responders on the ground. Finally, HAC-ER incorporates an infrastructure and the associated intelligence for tracking and utilising the provenance of information shared across the entire system to ensure its accountability. We individually validate each of these elements of HAC-ER and show how they perform against standard (non-HAC) baselines and also elaborate on the evaluation of the overall system.