Showing papers by "Derek Bell published in 2019"

An Observational Study of Deep Learning and Automated Evaluation of Cervical Images for Cancer Screening.

Abstract: Background Human papillomavirus vaccination and cervical screening are lacking in most lower resource settings, where approximately 80% of more than 500 000 cancer cases occur annually. Visual inspection of the cervix following acetic acid application is practical but not reproducible or accurate. The objective of this study was to develop a “deep learning”-based visual evaluation algorithm that automatically recognizes cervical precancer/cancer.

An Observational Study of Deep Learning and Automated Evaluation of Cervical Images for Cancer Screening

Abstract: Background Human papillomavirus vaccination and cervical screening are lacking in most lower resource settings, where approximately 80% of more than 500 000 cancer cases occur annually. Visual inspection of the cervix following acetic acid application is practical but not reproducible or accurate. The objective of this study was to develop a “deep learning”-based visual evaluation algorithm that automatically recognizes cervical precancer/cancer.

Priority use cases for antibody-detecting assays of recent malaria exposure as tools to achieve and sustain malaria elimination.

Abstract: Measurement of malaria specific antibody responses represents a practical and informative method for malaria control programs to assess recent exposure to infection. Technical advances in recombinant antigen production, serological screening platforms, and analytical methods have enabled the identification of several target antigens for laboratory based and point-of-contact tests. Questions remain as to how these serological assays can best be integrated into malaria surveillance activities to inform programmatic decision-making. This report synthesizes discussions from a convening at Institut Pasteur in Paris in June 2017 aimed at defining practical and informative use cases for serology applications and highlights five programmatic uses for serological assays including: documenting the absence of transmission; stratification of transmission; measuring the effect of interventions; informing a decentralized immediate response; and testing and treating P. vivax hypnozoite carriers.

Evolving quality improvement support strategies to improve Plan-Do-Study-Act cycle fidelity: a retrospective mixed-methods study.

Abstract: Background Although widely recommended as an effective approach to quality improvement (QI), the Plan–Do–Study–Act (PDSA) cycle method can be challenging to use, and low fidelity of published accounts of the method has been reported. There is little evidence on the fidelity of PDSA cycles used by front-line teams, nor how to support and improve the method’s use. Data collected from 39 front-line improvement teams provided an opportunity to retrospectively investigate PDSA cycle use and how strategies were modified to help improve this over time. Methods The fidelity of 421 PDSA cycles was reviewed using a predefined framework and statistical analysis examined whether fidelity changed over three annual rounds of projects. The experiences of project teams and QI support staff were investigated through document analysis and interviews. Results Although modest, statistically significant improvements in PDSA fidelity occurred; however, overall fidelity remained low. Challenges to achieving greater fidelity reflected problems with understanding the PDSA methodology, intention to use and application in practice. These problems were exacerbated by assumptions made in the original QI training and support strategies: that PDSA was easy to understand; that teams would be motivated and willing to use PDSA; and that PDSA is easy to apply. QI strategies that evolved to overcome these challenges included project selection process, redesign of training, increased hands-on support and investment in training QI support staff. Conclusion This study identifies support strategies that may help improve PDSA cycle fidelity. It provides an approach to assess minimum standards of fidelity which can be replicated elsewhere. The findings suggest achieving high PDSA fidelity requires a gradual and negotiated process to explore different perspectives and encourage new ways of working.

Successful Healthcare Improvements From Translating Evidence in complex systems (SHIFT-Evidence): simple rules to guide practice and research.

Abstract: Background Evidence translation and improvement research indicate that healthcare contexts are complex systems, characterized by uncertainty and surprise, which often defy orchestrated intervention attempts. This article reflects on the implications of complexity on attempts to translate evidence, and on a newly published framework for Successful Healthcare Improvements From Translating Evidence in complex systems (SHIFT-Evidence). Discussion SHIFT-Evidence positions the challenge of evidence translation within the complex and evolving context of healthcare, and recognizes the wider issues practitioners routinely face. It is empirically grounded, and designed to be comprehensive, practically relevant and actionable. SHIFT-evidence is summarized by three principles designed to be intuitive and memorable: 'act scientifically and pragmatically'; 'embrace complexity'; and 'engage and empower'. Common challenges and strategies to overcome them are summarized in 12 'simple rules' that provide actionable guidance. Conclusion SHIFT-Evidence provides a practical tool to guide practice and research of evidence translation and improvement within complex dynamic healthcare settings. Implications are that improvement initiatives and research study designs need to take into account the unique initial conditions in each local setting; conduct needs to respond to unpredictable effects and address dependent problems; and evaluation needs to be sensitive to evolving priorities and the emergent range of activities required to achieve improvement.

Dr Foster global frailty score: an international retrospective observational study developing and validating a risk prediction model for hospitalised older persons from administrative data sets

Abstract: Objectives This study aimed to examine the prevalence of frailty coding within the Dr Foster Global Comparators (GC) international database. We then aimed to develop and validate a risk prediction model, based on frailty syndromes, for key outcomes using the GC data set. Design A retrospective cohort analysis of data from patients over 75 years of age from the GC international administrative data. A risk prediction model was developed from the initial analysis based on seven frailty syndrome groups and their relationship to outcome metrics. A weighting was then created for each syndrome group and summated to create the Dr Foster Global Frailty Score. Performance of the score for predictive capacity was compared with an established prognostic comorbidity model (Elixhauser) and tested on another administrative database Hospital Episode Statistics (2011-2015), for external validation. Setting 34 hospitals from nine countries across Europe, Australia, the UK and USA. Results Of 6.7 million patient records in the GC database, 1.4 million (20%) were from patients aged 75 years or more. There was marked variation in coding of frailty syndromes between countries and hospitals. Frailty syndromes were coded in 2% to 24% of patient spells. Falls and fractures was the most common syndrome coded (24%). The Dr Foster Global Frailty Score was significantly associated with in-hospital mortality, 30-day non-elective readmission and long length of hospital stay. The score had significant predictive capacity beyond that of other known predictors of poor outcome in older persons, such as comorbidity and chronological age. The score’s predictive capacity was higher in the elective group compared with non-elective, and may reflect improved performance in lower acuity states. Conclusions Frailty syndromes can be coded in international secondary care administrative data sets. The Dr Foster Global Frailty Score significantly predicts key outcomes. This methodology may be feasibly utilised for case-mix adjustment for older persons internationally.

Impact of protein fouling on nanoparticle capture within the Viresolve® Pro and Viresolve® NFP virus removal membranes

Abstract: Virus filtration is a robust size-based technique that can provide the high level of viral clearance required for the production of mammalian-derived biotherapeutics such as monoclonal antibodies. Several studies have shown that the retention characteristics of some, but not all, virus filters can be significantly affected by membrane fouling, but there have been no direct measurements of how protein fouling might alter the location of virus capture within these membranes. The objective of this study was to directly examine the effect of protein fouling by human immunoglobulin G (IgG) on virus capture within the Viresolve® Pro and Viresolve® NFP membranes by scanning electron microscopy using different size gold nanoparticles. IgG fouling shifted the capture location of 20 nm gold nanoparticles further upstream within the Viresolve® Pro filter due to the constriction and/or blockage of the pores in the virus retentive region of the filter. In contrast, IgG fouling had no measurable effect on the capture of 20 nm nanoparticles in the Viresolve® NFP membrane, and IgG fouling had no effect on the capture of larger 40 and 100 nm nanoparticles in either membrane. These results provide important insights into how protein fouling alters the virus retention characteristics of different virus filters.

Knowledge gaps in the construction of rural healthy homes: A research agenda for improved low-cost housing in hot-humid Africa.

Abstract: Lorenz von Seidlein and colleagues discuss improving house designs in rural Africa to benefit health.

Off-Colony Screening of Biosynthetic Libraries by Rapid Laser-Enabled Mass Spectrometry.

Abstract: By leveraging advances in DNA synthesis and molecular cloning techniques, synthetic biology increasingly makes use of large construct libraries to explore large design spaces. For biosynthetic pathway engineering, the ability to screen these libraries for a variety of metabolites of interest is essential. If the metabolite of interest or the metabolic phenotype is not easily measurable, screening soon becomes a major bottleneck involving time-consuming culturing, sample preparation, and extraction. To address this, we demonstrate the use of automated laser-assisted rapid evaporative ionization mass spectrometry (LA-REIMS)-a form of ambient laser desorption ionization mass spectrometry-to perform rapid mass spectrometry analysis direct from agar plate yeast colonies without sample preparation or extraction. We use LA-REIMS to assess production levels of violacein and betulinic acid directly from yeast colonies at a rate of 6 colonies per minute. We then demonstrate the throughput enabled by LA-REIMS by screening over 450 yeast colonies within <4 h, while simultaneously generating recoverable glycerol stocks of each colony in real time. This showcases LA-REIMS as a prescreening tool to complement downstream quantification methods such as liquid chromatography-mass spectroscopy (LCMS). By prescreening several hundred colonies with LA-REIMS, we successfully isolate and verify a strain with a 2.5-fold improvement in betulinic acid production. Finally, we show that LA-REIMS can detect 20 out of a panel of 27 diverse biological molecules, demonstrating the broad applicability of LA-REIMS to metabolite detection. The rapid and automated nature of LA-REIMS makes this a valuable new technology to complement existing screening technologies currently employed in academic and industrial workflows.

Fully-automated patient-level malaria assessment on field-prepared thin blood film microscopy images, including Supplementary Information.

Abstract: Malaria is a life-threatening disease affecting millions. Microscopy-based assessment of thin blood films is a standard method to (i) determine malaria species and (ii) quantitate high-parasitemia infections. Full automation of malaria microscopy by machine learning (ML) is a challenging task because field-prepared slides vary widely in quality and presentation, and artifacts often heavily outnumber relatively rare parasites. In this work, we describe a complete, fully-automated framework for thin film malaria analysis that applies ML methods, including convolutional neural nets (CNNs), trained on a large and diverse dataset of field-prepared thin blood films. Quantitation and species identification results are close to sufficiently accurate for the concrete needs of drug resistance monitoring and clinical use-cases on field-prepared samples. We focus our methods and our performance metrics on the field use-case requirements. We discuss key issues and important metrics for the application of ML methods to malaria microscopy.

Strengths and weaknesses of the acute care systems in the United Kingdom and the Netherlands: what can we learn from each other?

Abstract: The demand on Emergency Departments and acute medical services is increasing internationally, creating pressure on health systems and negatively influencing the quality of delivered care. Visible consequences of the increased demand on acute services is crowding and queuing. This manifests as delays in the Emergency Departments, adverse clinical outcomes and poor patient experience. Despite the similarities in the UK’s and Dutch health care systems, such as universal health coverage, there are differences in the number of patients presenting at the Emergency Departments and the burden of crowding between these countries. Given the similarities in funding, this paper explores the similarities and differences in the organisational structure of acute care in the UK and the Netherlands. In the Netherlands, less patients are seen at the ED than in England and the admission rate is higher. GPs and so-called GP-posts serve 24/7 as gatekeepers in acute care, but EDs are heterogeneously organised. In the UK, the acute care system has a number of different access points and the accessibility of GPs seems to be suboptimal. Acute ambulatory care may relieve the pressure from EDs and Acute Medical Units. In both countries the ageing population leads to a changing case mix at the ED with an increased amount of multimorbid patients with polypharmacy, requiring generalistic and multidisciplinary care. The acute and emergency care in the Netherlands and the UK face similar challenges. We believe that each system has strengths that the other can learn from. The Netherlands may benefit from an acute ambulatory care system and the UK by optimizing the accessibility of GPs 24/7 and improving signposting for urgent care services. In both countries the changing case mix at the ED needs doctors who are superspecialists instead of subspecialists. Finally, to improve the organisation of health care, doctors need to be visible medical leaders and participate in the organisation of care.

Fully-automated patient-level malaria assessment on field-prepared thin blood film microscopy images

Abstract: Malaria is a life-threatening disease affecting millions. Microscopy-based assessment of thin blood films is a standard method to (i) determine malaria species and (ii) quantitate high-parasitemia infections. Full automation of malaria microscopy by machine learning (ML) is a challenging task because field-prepared slides vary widely in quality and presentation, and artifacts often heavily outnumber relatively rare parasites. In this work, we describe a complete, fully-automated framework for thin film malaria analysis that applies ML methods, including convolutional neural nets (CNNs), trained on a large and diverse dataset of field-prepared thin blood films. Quantitation and species identification results are close to sufficiently accurate for the concrete needs of drug resistance monitoring and clinical use-cases on field-prepared samples. We focus our methods and our performance metrics on the field use-case requirements. We discuss key issues and important metrics for the application of ML methods to malaria microscopy.

How New are New Harms Really? Climate Change, Historical Reasoning and Social Change

Abstract: Climate change and other contemporary harms are often depicted as New Harms because they seem to constitute unprecedented challenges. This New Harms Discourse rests on two important premises, both of which we criticise on empirical grounds. First, we argue that the Premise of changed conditions of human interaction—according to which the conditions regarding whom people affect (and how) have changed recently and which emphasises the difference with past conditions of human interaction—risks obfuscating how humanity’s current predicament is merely the transient result of long-term, gradual processes and developments. Second, we dispute the Premise that New Harms have certain features that render them new and argue that New Harms share characteristics with other (past) harms. On the basis of these premises, the New Harms Discourse concludes that climate change is a unique social challenge that requires radically new moral thinking, but we argue that this Uniqueness Myth distracts attention from the valuable lessons we can draw from humanity’s successes and failures in dealing with past harms. We will illustrate how action to tackle climate change and other complex, systemic harms can be informed by the interdisciplinary study of historic harms. We will argue that rejecting the New Harms Discourse is not only empirically justified, it also gives cause for optimism, because it opens up the possibility to draw upon the past to face problems in the present and future.

Climate ethics with an ethnographic sensibility

Abstract: What responsibilities does each of us have to reduce or limit our greenhouse gas emissions? Advocates of individual emissions reductions acknowledge that there are limits to what we can reasonably demand from individuals. Climate ethics has not yet systematically explored those limits. Instead, it has become popular to suggest that such judgements should be ‘context-sensitive’ but this does not tell us what role different contextual factors should play in our moral thinking. The current approach to theory development in climate ethics is not likely to be the most effective way to fill this gap. In existing work, climate ethicists use hypothetical cases to consider what can be reasonably demanded of individuals in particular situations. In contrast, ‘climate ethics with an ethnographic sensibility’ uses qualitative social science methods to collect original data in which real individuals describe their own situations. These real-life cases are more realistic, more detailed and cover a broader range of circumstances than hypothetical cases. Normative analysis of real-life cases can help us to develop a more systematic understanding of the role that different contextual factors should play in determining individual climate responsibilities. It can also help us to avoid the twin dangers of ‘idealization’ and ‘special pleading’.

A Robust Incubator to Improve Access to Microbiological Culture in Low Resource Environments.

Abstract: To help address the limitations of operating conventional microbiological culture incubators in low resource environments, a new incubator design was developed and tested to meet the requirements of operation in laboratories without reliable power (power outages up to 12 contiguous hours) or climate control (ambient indoor temperatures from 5 °C to 45 °C). The device is designed to enable adherence to incubation temperatures recommended for growth detection, identification, and drug susceptibility testing (DST) of human pathogenic bacteria. During power outages, stable temperatures are maintained in the device's internal sample compartment by employing phase change material (PCM) as a bi-directional thermal battery to maintain incubation temperature. Five prototypes were tested in a laboratory setting using environmental test chambers and programmable power supplies, and three were field tested in the Lao PDR in situations of intended use. The prototypes successfully held their temperature to within ±1 °C in both laboratory environmental chamber testing as well as during the field test. The results indicate that the device will maintain stable culture temperatures across periods of intermittent power supply, while enabling normal workflow of this could greatly increase the availability of microbiological culture for diagnosis and antimicrobial resistance (AMR) monitoring.

Improving the quality of care for patients with or at risk of atrial fibrillation: an improvement initiative in UK general practices.

Abstract: Objective Atrial fibrillation (AF) is a growing problem internationally and a recognised cause of cardiovascular morbidity and mortality. The London borough of Hounslow has a lower than expected prevalence of AF, suggesting poor detection and associated undertreatment. To improve AF diagnosis and management, a quality improvement (QI) initiative was set up in 48 general practices in Hounslow. We aimed to study whether there was evidence of a change in AF diagnosis and management in Hounslow following implementation of interventions in this QI initiative. Methods Using the general practice information system (SystmOne), data were retrospectively collected for 415 626 patients, who were actively registered at a Hounslow practice between 1 January 2011 and 31 August 2018. Process, outcome and balancing measures were analysed using statistical process control and interrupted time series regression methods. The baseline period was from 1 January 2011 to 30 September 2014 and the intervention period was from 1 October 2014 to 31 August 2018. Results When comparing the baseline to the intervention period, (1) the rate of new AF diagnoses increased by 27% (relative risk 1.27; 95% CI 1.05 to 1.52; p Conclusion Implementation of interventions in the Hounslow QI initiative coincided with improved AF diagnosis and management. Areas with perceived underdetection of AF should consider similar interventions and methodology.

Better Use of Data to improve parent Satisfaction (BUDS): protocol for a prospective before-and-after pilot study employing mixed methods to improve parent experience of neonatal care

Abstract: Introduction Having a baby that requires neonatal care is stressful and traumatic. Parents often report dissatisfaction with communication of clinical information. In the UK neonatal care data are recorded daily using electronic patient record systems (EPR), from which deidentified data form the National Neonatal Research Database (NNRD). We aim to evaluate the impact of sharing neonatal EPR data with parents, on parent-reported satisfaction, parent–staff interactions, staff workload and data completeness. Methods A prospective, before-and-after, mixed-method study. Participants are parents of inpatient babies (maximum 90) and staff in a tertiary neonatal intensive care unit, London, UK. The intervention was developed by former neonatal parents, neonatologists and neonatal nurses: a communication tool for parents comprising individualised, written, daily infant updates for parents, derived from EPR data. The intervention will be provided to parents over 6 weeks. Plan-Do-Study-Act cycles will inform the tool’s iterative development and improvement. The tool’s impact will be measured using a validated parent survey, staff survey, data completeness measures and interviews. Analysis Primary outcome: parent satisfaction ‘with communication of clinical information and involvement in care’. Secondary outcomes: parent–staff interactions, staff workload, data completeness. Baseline survey data will be obtained from clinical service evaluation preceding the intervention. Baseline data completeness will be derived from the NNRD. During the intervention, surveys will be administered biweekly and data completeness assessed daily. We will analyse outcomes using run charts and partially paired statistical tests. Parent and staff interviews will explore information exchange and the communication tool’s impact. Discussion This study will evaluate the impact of a parent co-designed intervention on communication with parents in neonatal care and the completeness of routinely recorded electronic clinical data. Better use of routinely recorded clinical data provides the opportunity to improve parent satisfaction and increase the research utility of such data, benefiting clinical care. Ethics and dissemination Reviewed and approved by the West Midlands—South Birmingham REC (18/WM/0175). Registration number ISRCTN62718241.

Unscheduled hospital contacts after inpatient discharge: A national observational study of COPD and heart failure patients in England.

Abstract: Introduction Readmissions are a recognised challenge for providers of healthcare and incur financial penalties in a growing number of countries. However, the scale of unscheduled hospital contacts including attendances at emergency departments that do not result in admission is not well known. In addition, little is known about the route to readmission for patients recently discharged from an emergency hospital stay. Methods This is an observational study of national hospital administration data for England. In this retrospective cohort study, we tracked patients for 30 days after discharge from an emergency admission for heart failure (HF) or chronic obstructive pulmonary disorder (COPD). Results The majority of patients (COPD:79%; HF:75%) had no unscheduled contact with secondary health care within 30 days of discharge. Of those who did have unscheduled contact, the most common first unscheduled contact was emergency department (ED) attendance (COPD:16%; HF:18%). A further 5% of COPD patients and 4% of HF patients were admitted for an emergency inpatient stay, but not through the ED. A small percentage of patients (COPD:<1%, HF:2%) died without any known contact with secondary care. ED conversion rates at first attendance for both COPD and HF were high: 75% and 79% respectively. A quarter of patients who were not admitted during this first ED attendance attended the ED again within the 30-day follow-up period, and around half (COPD:56%; HF:63%) of these were admitted at this point. Patients who live alone, had an index admission which included an overnight stay and were comorbid had higher odds of being admitted through the ED than via other routes. Conclusion While the majority of patients did not have unscheduled contact with secondary care in the 30 days after index discharge, many patients attended the ED, often multiple times, and many were admitted to hospital, not always via the ED. More frail patients were more likely to be admitted through the ED, suggesting a possible area of focus as discharge bundles are developed.

What do care home managers believe constitutes an ‘assessment for frailty’ of care home residents in North-West London? A survey

Abstract: Frail individuals are at risk of significant clinical deterioration if their frailty is not identified and managed appropriately. Research suggests that any interaction between an older person and a health or social care professional should include an assessment for frailty. Many older care home residents are frail when admitted, but we have little knowledge of whether or how this is assessed. The aim of this paper is to understand and establish the characteristics of the reported ‘assessments for frailty’ used in care homes with nursing (nursing homes) across North-West London. This will help understand what an ‘assessment for frailty’ of care home residents mean in practice in North-West London. Telephone contact was made with every Care Quality Commission (CQC) (independent regulator of health and adult social care in England) regulated nursing home across North-West London [n = 87]. An online survey was sent to all that expressed interest [n = 73]. The survey was developed through conversations with healthcare professionals, based on literature and tested with academics and clinicians. Survey responses were analysed using descriptive statistics. The Mann-Whitney U test was used for statistical analyses. 24/73 nursing homes completed the survey (33%). Differences in the characteristics of reported ‘assessments for frailty’ across nursing homes were evident. Variation in high level domains assessed (physical, social, mental and environmental) was observed. Nurses were the most common professional group completing assessments for frailty, with documentation and storage being predominantly paper based. A statistically significant difference between the number of assessments used in corporate chain owned nursing homes (3.9) versus independently owned nursing homes (2.1) was observed (U = 21, p = .005). Great variation existed in the characteristics of reported ‘assessments for frailty’ in nursing homes. Our study suggests that not all physical, social, mental and environmental domains of frailty are routinely assessed: it appears that frailty is still primarily viewed only in terms of physical health. The consequences of this could be severe for patients, staff and healthcare settings. Research illustrates that frailty is a broad, multifactorial health state and, as such, an overall ‘assessment for frailty’ should reflect this.

Bombs and cocaine: detecting nefarious nitrogen sources using remote sensing and machine learning

Abstract: Plants are treated with synthetic or organic nitrogen sources to increase growth and yield, the most common being calcium ammonium nitrate. However, some nitrogen sources are used in illicit activities. Ammonium nitrate is used in explosive manufacture and ammonium sulphate in the cultivation and extraction of the narcotic cocaine from Erythroxylum spp. Here we show that hyperspectral sensing, multispectral imaging and machine learning image analysis can be used to visualise and differentiate plants exposed to different nefarious nitrogen sources. Metabolomic analysis of leaves from plants exposed to different nitrogen sources reveals shifts in colourful metabolites that may contribute to altered reflectance signatures. Overall this suggests that different nitrogen feeding regimes alter plant secondary metabolism leading to changes in the reflectance spectrum detectable via machine learning of multispectral data but not the naked eye. Our results could facilitate the development of technologies to monitor illegal activities involving various nitrogen sources and further inform nitrogen application requirements in agriculture.