We previously reported a study of features of emergency healthcare response to COVID-19 that could be modified to mitigate against future excess deaths. Here we determined what themes persisted in later waves. This was an expert panel review of all components of care delivered to COVID-19 patients who died (primary and secondary care, community services, NHS 111 and 999, COVID oximetry at home, virtual wards). 174 deaths were included. 5% were deemed >50% avoidable, 75% included avoidability themes. Contact with primary care remains mostly via telephone, creating diagnostic risk. Patient decision to avoid healthcare contact was common. Recommendations include: better utilisation of home monitoring in future pandemics; improved avoidance of nosocomial spread; patients be encouraged to seek medical advice earlier.

Background The response to the COVID-19 pandemic in the United Kingdom included large scale changes to healthcare delivery, without fully understanding the potential for unexpected effects caused by these changes. The aim was “ to ascertain the characteristics of patients, uncertainty over diagnosis, or features of the emergency response to the pandemic that could be modified to mitigate against future excess deaths”. Methods Review of the entire pathway of care of patients whose death was registered in Salford during the 8 week period of the first wave (primary care, secondary care, 111 and 999 calls) in order to create a single record of healthcare prior to death. An expert panel judged avoidability of death against the National Mortality Case Record Review Programme scale. The panel identified themes using a structured judgement review format. Results There were 522 deaths including 197 in hospital, and 190 in care homes. 51% of patients were female, 81% Caucasian, age 79 ± 9 years. Dementia was present in 35%, COVID-19 was cause of death in 44%. Healthcare contact prior to death was most frequently with primary care (81% of patients). Forty-six patients (9%) had healthcare appointments cancelled (median 1 cancellation, range 1–9). Fewer than half of NHS 111 calls were answered during this period. 18% of deaths contained themes consistent with some degree of avoidability. In people aged ≥75 years who lived at home this was 53%, in care home residents 29% and in patients with learning disability 44% ( n = 9). Common themes were; delays in patients presenting to care providers (10%), delays in testing (17%), avoidable exposure to COVID-19 (26%), delays in provider response (5%), and sub-optimal care (11%). For avoidability scores of 2 or 3 (indicating more than 50% chance of avoidability), 44% of cases had > 2 themes. Conclusions The initial emergency response had unforeseen consequences resulting in late presentation, sub-optimal assessments, and delays in receiving care. Death in more vulnerable groups was more likely to display avoidability themes.

Background The United Kingdom COVID-19 pandemic response included large scale changes to emergency healthcare delivery across primary care, secondary care and ambulance services without fully understanding potential unintended effects. Our aim was to ascertain themes within the emergency response that could be modified to mitigate against future excess deaths. Methods A structured judgement review of the entire care pathway (primary care, secondary care, 111 and 999 calls) for deaths registered in Salford during weeks 12-19 of 2020, creating a single integrated record of all Healthcare interactions and outcomes between 1 st March 2020 and death for all patients. An expert panel judged avoidability of death against the National Mortality Case Record Review Programme scale, and themes associated with potential harm. Results 522 deaths (197 hospital, 190 care homes): 51% female, 81% Caucasian, 35% dementia, age 79±9 years, 44% COVID-19 cause of death. 18% of deaths contained avoidability themes. In people aged ≥75 years who lived at home this was 53%, in care home residents 29%, in learning disability patients 44%. For COVID-19 and non-COVID-19 deaths the figures were 49% and 23%. 15 deaths were judged more than 50:50 avoidable. For higher avoidability (score 2 or 3), 44% had >2 themes. Common themes were delays in patients presenting to care providers (10%, n=53), delays in testing (27%, 139), avoidable exposure to COVID-19 (29%, 149), delays in provider response (6%, 32), and sub-optimal care (34%, 177). Pre-hospital healthcare contact frequently was primary care 81%, 999 56%, NHS 111 11%. The most common outcome of 111 calls was advice to contact GP (34%). 46 patients (9%) had healthcare appointments cancelled (median 1 cancellation, range 1-9). Conclusions The initial emergency response to COVID-19 was associated with late presentation, sub-optimal assessments, and delays in receiving care. The effects had multifaceted impacts and will require complex remedy.

Background The response to the COVID-19 pandemic in the United Kingdom included large scale changes to healthcare delivery, without fully understanding the potential for unexpected effects caused by these changes. The aim was “ to ascertain the characteristics of patients, uncertainty over diagnosis, or features of the emergency response to the pandemic that could be modified to mitigate against future excess deaths”. Methods Review of the entire pathway of care of patients whose death was registered in Salford during the 8 week period of the first wave (primary care, secondary care, 111 and 999 calls) in order to create a single record of healthcare prior to death. An expert panel judged avoidability of death against the National Mortality Case Record Review Programme scale. The panel identified themes using a structured judgement review format. Results There were 522 deaths including 197 in hospital, and 190 in care homes. 51% of patients were female, 81% Caucasian, age 79 ± 9 years. Dementia was present in 35%, COVID-19 was cause of death in 44%. Healthcare contact prior to death was most frequently with primary care (81% of patients). Forty-six patients (9%) had healthcare appointments cancelled (median 1 cancellation, range 1–9). Fewer than half of NHS 111 calls were answered during this period. 18% of deaths contained themes consistent with some degree of avoidability. In people aged ≥75 years who lived at home this was 53%, in care home residents 29% and in patients with learning disability 44% ( n = 9). Common themes were; delays in patients presenting to care providers (10%), delays in testing (17%), avoidable exposure to COVID-19 (26%), delays in provider response (5%), and sub-optimal care (11%). For avoidability scores of 2 or 3 (indicating more than 50% chance of avoidability), 44% of cases had > 2 themes. Conclusions The initial emergency response had unforeseen consequences resulting in late presentation, sub-optimal assessments, and delays in receiving care. Death in more vulnerable groups was more likely to display avoidability themes.

Background UK Ambulance services are under pressure to safely stream appropriate patients away from the Emergency Department (ED). Even so, there has been little evaluation of patient outcomes. We investigated differences between patients who are conveyed directly to ED after calling 999 and those referred by an ambulance crew to a novel GP referral scheme. Methods This was a prospective study comparing patients from two cohorts, one conveyed directly to the ED ( n = 4219) and the other referred to a GP by the on-scene paramedic ( n = 321). To compare differences in patient outcomes, we include follow-up data of a smaller subset of each cohort (up to n = 150 in each) including hospital admission, history of long-term illness, previous ED attendance, length of stay, hospital investigations, internal transfers, 30-day re-admission and 10-month mortality. Results Older individuals, females, and those with minor incidents were more likely to be referred to a GP than conveyed directly to ED. Of those patients referred to the GP, only 22.4% presented at ED within 30 days. These patients were more likely to be admitted then than were those initially conveyed directly to ED (59% vs 31%). Those conveyed to ED had a higher risk of death compared to those who were referred to the GP (HR: 2.59; 95% CI 1.14–5.89), however when analyses were restricted to those who presented at ED within 30 days, there was no difference in mortality risk (HR: 1.45; 95% CI 0.58–3.65). Conclusions Despite limited data and a small sample size, there were differences between patients conveyed directly to ED and those who were referred into GP care. Initial evidence suggests that referring individuals to a GP may provide an appropriate and safe alternative path of care. This pilot study demonstrated a need for larger scale, methodologically rigorous study to demonstrate the benefits of alternative conveyance schemes and recommend changes to the current system of urgent and emergency care.

Many long-standing image processing problems in applied science domains are finding solutions through the application of deep learning approaches to image processing. Here we present one such application; the case of classifying images of protein crystallisation droplets. The Collaborative Crystallisation Centre in Melbourne, Australia is a medium throughput service facility that produces between five and twenty thousand images per day. This submission outlines a reliable and robust machine learning pipeline that autonomously classifies these images using CSIRO’s high-performance computing facilities. Our pipeline achieves improved accuracies over existing implementations and delivers these results in real time. We discuss the specific tools and techniques used to construct the pipeline, as well as the methodologies for testing and validating externally developed classification models.

Scientific workflows describe steps for orchestrating the execution of a network of computational operators toward some goal, such as data transformation for analysis or visualization. Typically, these operators consume and emit transformed data, or cause some effect. In most scientific workflow systems, the operators are typed to enable compatibility checks for their composition that make up a workflow. However, type checking performed by most such systems today is still largely confined to syntactic checking with limited, if any, semantic type checking support. In this paper, we present a type system incorporating the W3C OWL ontology language to aid in representing the semantics of data and workflow operators. We show how this type system supports the detection of type incompatibility errors in workflow compositions, and how it facilitates a (semi-)automatic type correction procedure using type transformations. We have incorporated our solution into Kepler, enabling users to statically test the type-consistency of workflows typed using our type language, and demonstrate that inconsistent bindings between expressively typed operators can be automatically corrected via a procedure that seeks to compose adapter/shim functions, such as unit transformations, time series interpolations, or some other arbitrarily complex data transformations.

The Resource Description Framework (RDF) and Web Ontology Language (OWL) have been widely used in recent years, and automated methods for the analysis of data and knowledge directly within these formalisms are of current interest. Concept induction is a technique for discovering descriptions of data, such as inducing OWL class expressions to describe RDF data. These class expressions capture patterns in the data which can be used to characterise interesting clusters or to act as classification rules over unseen data. The semantics of OWL is underpinned by Description Logics (DLs), a family of expressive and decidable fragments of first-order logic. Recently, methods of concept induction which are well studied in the field of Inductive Logic Programming have been applied to the related formalism of DLs. These methods have been developed for a number of purposes including unsupervised clustering and supervised classification. Refinement-based search is a concept induction technique which structures the search space of DL concept/OWL class expressions and progressively generalises or specialises candidate concepts to cover example data as guided by quality criteria such as accuracy. However, the current state-of-the-art in this area is limited in that such methods: were not primarily designed to scale over large RDF/OWL knowledge bases; do not support class languages as expressive as OWL2-DL; or, are limited to one purpose, such as learning OWL classes for integration into ontologies. Our work addresses these limitations by increasing the efficiency of these learning methods whilst permitting a concept language up to the expressivity of OWL2-DL classes. We describe methods which support both classification (predictive induction) and subgroup discovery (descriptive induction), which, in this context, are fundamentally related. We have implemented our methods as the system called OWL-Miner and show by evaluation that our methods outperform state-of-the-art systems for DL learning in both the quality of solutions found and the speed in which they are computed. Furthermore, we achieve the best ever ten-fold cross validation accuracy results on the long-standing benchmark problem of carcinogenesis. Finally, we present a case study on ongoing work in the application of OWL-Miner to a real-world problem directed at improving the efficiency of biological macromolecular crystallisation.

An innovative policy implemented by a UK Ambulance Service allows paramedics to refer patients to a GP Acute Visiting Service scheme. Initial evidence suggests that this alternate route of care can decrease hospital admission rates, decrease A&E waiting time and provide substantial savings for the NHS. However, there are many unrecognised barriers to referral that are not captured by the quantitative analysis. The goal of this qualitative-observational study was to gain insight into the GP referral scheme from a paramedic's perspective. All notes were transcribed, coded and analysed using a Grounded Theory approach. Four main themes emerged: 1) barriers to referral including wait time, process, and lack of confidence, experience and training 2) approaching the patient with the GP referral scheme in mind 3) frustrations with GP decision making and 4) awareness/understanding of the scheme's impacts. This study provided valuable insight into the paramedic's perspective of the GP referral scheme. Maximising understanding of the scheme, investigating the GP's perspective in decision making and ensuring knowledge and accountability of paramedics, GPs and the public were identified as solutions to strengthen and increase referral rates and scheme success.

Beam-search may be used to iteratively explore and evaluate refinements of candidate hypotheses expressed in logical formalisms such as description logic. In this paper, we analyse heuristics for beam search methods over OWL classes and present a novel search algorithm, OWL-Miner which leverages the properties of convex measure functions to deliver an improved memory-bounded beam search. We present performance results on the mutagenesis benchmark problem and demonstrate superior performance relative to another state-of-the-art implementation, and present 10-fold cross-validated accuracy results which are comparable with those from a variety of other methods. Our improvements to the space and time-based efficiency of refinement-based learning algorithms are significant for expanding the size of learning problems that can be feasibly addressed by refinement learning and the quality of solutions that can be found with limited resources.

Background An innovative policy developed and implemented by a UK Ambulance Service allows paramedics to refer patients to the GP Acute Visiting Service scheme. Initial evidence suggests that using this alternate route of care can decrease hospital admission rates, increase bed availability, decrease wait time in A and Es and provide substantial savings for the NHS. However, there are many unrecognised barriers to referral that have not been captured by the quantitative analysis. The goal of this qualitative-observational study was to gain insight into the GP referral scheme from a paramedic’s perspective. Methods We observed eight paramedics throughout full shifts of 8–12 hours. Data was collected using participant demographics, researcher observations and informal semi-structured interviews. All notes were transcribed, coded and analysed using a Grounded Theory approach to identify emerging themes. Results Paramedics expressed a wide range of frustrations with the scheme, identifying the waiting time, the process and a lack of confidence, experience and training as the three major barriers to referrals. They described how they approached patients with the GP referral scheme in mind, identified common characteristics of referrals, described how the triage tool shaped their decision making and shared how they involved the patient in the decision making. They shared too their frustrations with some GP decision making, which they admitted then influenced their future decision making. Finally, they described what motivated them to refer and discussed the lack of awareness and understanding of the scheme’s impact and aims. Conclusions This study provided valuable insight into the paramedic’s perspective of the GP referral scheme. Maximising understanding of the scheme, investigating the GP’s perspective in decision making and ensuring knowledge and accountability of paramedics, GPs and the public were identified as solutions to strengthen and increase referral rates and scheme success.

Purpose With the increasing demand on ambulance services, paramedics are tasked to arrange as much out of hospital care as possible, to develop integrated systems of care and work with hundreds of different providers – all in the 15 minutes allocated for assessment. A UK ambulance trust is navigating and leading much of this work as one of the first trusts to implement a general practitioner referral policy as an alternate to direct conveyance. The paper aims to discuss this issue. Design/methodology/approach Here the authors discuss the referral scheme, examine the limited evidence available and discuss what is needed to influence prospective success of implementing this scheme in other trusts. Findings Limited evidence for these schemes are described, however there is a clear gap in critical appraisal and methodologically rigorous evidence needed to implement these schemes in other ambulance schemes. Originality/value In order to facilitate collaboration of healthcare services and to minimize the burden of increasing numbers of patients, communication and discussion of alternate routes of care is crucial. This viewpoint piece is one of the first to emphasize the potential benefits of such schemes.

Mechanisms to facilitate rapid ambulance transport of diagnosed STEMI patients from the community and emergency departments (ED) settings directly to primary percutaneous coronary intervention (PPCI) facilities are well established within NHS Ambulance Services. Direct challenge of inter-hospital transfer requests for non-emergency percutaneous coronary intervention (PCI) patients by a regional NHS Ambulance Service identified disagreement between peripheral feeder hospitals and the NHS Ambulance Service on what level of ambulance transport is most appropriate. To reduce unnecessary peripheral feeder hospital requests for paramedic emergency service transfer and resource utilisation in non-emergency PCI patients and to assess the clinical safety of both non-emergency transport and multi-occupancy conveyance for this patient group. A process was established with a regional cardiothoracic centre to support pre-screening of non-emergency PCI patients for conveyance via non-emergency ambulance resources and multi-occupancy. This included centralisation of all non-emergency PCI ambulance transport booking practices and dissemination of learning materials on the process to all stakeholders. During the three-year period 3172 patients were identified as suitable for conveyance by both non-emergency ambulance transports. Of this, 36% (n=1767) were conveyed as part of a multi-occupancy journey and 56% (n=782) were conveyed by non-emergency resources. Overall, 69% (n=782) of all multi-occupancy conveyances were undertaken by non-emergency resources. Two clinical incidents were noted during this period, both of which were managed via clinical telephone advice. Non-emergency ambulances can be safely used to transport non-emergency PCI patients via multi-occupancy, following appropriate pre-screening by the receiving PCI unit. Further work is needed to understand the feasibility of this across other patient groups in the inter-hospital transfer scenario and its transferability to other NHS Ambulance Services.

Integrating and accessing data from the many existing formats is a key challenge for the Semantic Web. In particular, representing and understanding non-RDF/OWL data in terms of ontologies is important for reasoning about this data and its relationships to other data, and for extracting or accessing the data itself. This paper presents a declarative method for mapping and translating spreadsheet data to RDF/OWL. First, a declarative mapping language is presented. Then, for this syntax, the paper presents a declarative semantics in terms of matrix operations as abstractions over spreadsheets and the relations therein. The language supports arbitrary search and indexing functions over spreadsheet data and allows the generation of complex associations and aggregations in the RDF/OWL output. Imperative and functional implementations of the mapping and extraction technology are presented.

•We propose an ontology-mediated approach for environmental data translation.•We outline the principles underlying the design of a mediating ontology, and show the development of such an ontology.•We propose a declarative formalism for representing spreadsheet-to-ontology mappings, and give an algorithm for the mapping evaluation.•We propose a declarative formalism for representing ontology-to-XML mappings, and give an algorithm for the mapping evaluation.•We have developed an ontology-mediated spreadsheet-to-XML translation tool, and showed its effectiveness with real environmental observations data.

We propose an ontology-mediated approach for environmental data translation.We outline the principles underlying the design of a mediating ontology, and show the development of such an ontology.We ...

The Emergency Response Intelligence Capability (ERIC) tool, http://eric.csiro.au, automatically gathers data about emergency events from authoritative web sources, harmonises the information content and presents it on an interactive map. All data is recorded in a database which allows the changing status of emergency events to be identified and provides an archive for historical review. ERIC was developed for the Australian Government Department of Human Services Emergency Management team who is responsible for intelligence gathering and situation reporting during emergency events. Event information is combined with demographic data to profile the affected community. Identifying relevant community attributes, such as languages spoken or socioeconomic information, allows the department to tailor its response appropriately to better support the impacted community. An overview of ERIC is presented, including its use by the department and the difficulties overcome in establishing and maintaining a nationally consistent harmonised model of emergency event information. Preliminary results of republishing the emergency event information using the Australian Profile of the Common Alerting Protocol, an XML standard to facilitate the construction and exchange of emergency alert and warning messages, are also presented.

The Emergency Response Intelligence Capability (ERIC) is a web-based productivity tool developed by CSIRO for the Australian Government Department of Human Services (hereafter referred to as the department). ERIC automatically gathers 'live' web feed data about emergency events from authoritative Australian web sources, harmonises the information content, archives the content in a database, presents it in a map-based web application, and semi-automatically generates situation reports. This allows the department's Emergency Management Operations team to monitor events occurring around the country and provides fast and intuitive access to a wide collection of information, including demographic data from the Australian Bureau of Statistics (ABS). Information for a specific region under investigation can be extracted and collated semi-automatically to generate situation reports that include information synthesised from available datasets and augmented by user provided content. The current status of the ERIC project is provided with a focus on how it is being trialled by the department to improve the business process of creating and maintaining situation reports during emergency events.

Background English Ambulance Services are faced with annual increases in emergency demand. Addressing the demand for low acuity emergency calls relies upon the ability of ambulance clinicians to accurately identify the most appropriate destination or referral pathway. Given the risk of undertriage, the challenge is to develop processes that can safely determine patient dispositions, thereby increasing the number of patients receiving care closer to home.Aims The aim of the study was to evaluate the clinical utility and safety of triage support tools (Paramedic Pathfinders).Methods Two triage filters (Pathfinders) were developed (one medical, one trauma). These were applied by ambulance clinicians to 481 patients who had been transported to emergency departments (EDs). Preferred (gold standard) patient dispositions were established by senior medical practitioners using both ambulance and ED clinical records. The clinical utility of ambulance clinicians using Pathfinders was evaluated against this gold standard.Results The Medical Pathfinder was applied to 367 patients (76.3%) and the Trauma Pathfinder to 114 (23.7%). Agreement between ambulance clinician and gold standard was achieved in 387 cases (80.5%) giving the tools a combined sensitivity of 94.83% and specificity of 57.9%. 20.9% of medical patients and 30.7% of trauma patients who had been transported to hospital could have been safely cared for elsewhere.Conclusions Ambulance clinicians using Pathfinders have demonstrated acceptable levels of sensitivity in identifying patients who require ED care. The actual impact of the tools in clinical practice will be dependent on the provision of suitable alternatives to ED.