Artificial intelligence and clinical decision support: clinicians' perspectives on trust, trustworthiness, and liability.

Artificial intelligence (AI) could revolutionise health care, potentially improving clinician decision making and patient safety, and reducing the impact of workforce shortages. However, policymakers and regulators have concerns over whether AI and clinical decision support systems (CDSSs) are trusted by stakeholders, and indeed whether they are worthy of trust. Yet, what is meant by trust and trustworthiness is often implicit, and it may not be clear who or what is being trusted. We address these lacunae, focusing largely on the perspective(s) of clinicians on trust and trustworthiness in AI and CDSSs. Empirical studies suggest that clinicians' concerns about their use include the accuracy of advice given and potential legal liability if harm to a patient occurs. Onora O'Neill's conceptualisation of trust and trustworthiness provides the framework for our analysis, generating a productive understanding of clinicians' reported trust issues. Through unpacking these concepts, we gain greater clarity over the meaning ascribed to them by stakeholders; delimit the extent to which stakeholders are talking at cross purposes; and promote the continued utility of trust and trustworthiness as useful concepts in current debates around the use of AI and CDSSs.

Insights from developing and evaluating the NHS blood choices transfusion app to support junior and middle-grade doctor decision making against guidelines.

OBJECTIVES: To: 1. Develop a CE-marked smartphone App to support doctors' concordance with transfusion guidelines in non-bleeding adult patients, emphasising informed consent and anaemia management. 2. Test App accuracy and potential to improve user decisions. BACKGROUND: Studies have shown inappropriate use of blood components and that most junior doctors own smartphones with medical apps. METHODS: A multidisciplinary team developed App screens and logic through an iterative process based on national guidelines. Thirty medical or surgical transfusion scenarios were developed based on national guidelines and each sent to Consultant Haematologist experts in Transfusion Medicine. To obtain a clinical consensus and exclude ambiguous scenarios, their independent decisions and associated certainty were compared. The consensus clinical decision was then compared with guidance from the App. To explore potential App impact on simulated user decisions, 26 junior doctors responded to five transfusion scenarios before and after access to the App. RESULTS: The Blood Choices App agreed with 91% (95% CI: 72%-99%) of expert decisions with a sensitivity of 100% (69% to 100%) and specificity of 85% (55%-98%). Excluding one malfunction scenario, the App had the potential to increase correct decisions by junior doctors from 83% (73%-90%) pre-App use to 96% (88%-99%) post (p-value 0.013), with 90% (67%-99%) saying they would use it in practice. CONCLUSIONS: Transfusion guidelines can be converted into an App with potential to improve guideline concordance. However, evaluating such Apps is essential to understand their limitations, detect malfunctions and prevent harm.

Correction: Online Guide for Electronic Health Evaluation Approaches: Systematic Scoping Review and Concept Mapping Study.

[This corrects the article DOI: .].

Online Guide for Electronic Health Evaluation Approaches: Systematic Scoping Review and Concept Mapping Study.

BACKGROUND: Despite the increase in use and high expectations of digital health solutions, scientific evidence about the effectiveness of electronic health (eHealth) and other aspects such as usability and accuracy is lagging behind. eHealth solutions are complex interventions, which require a wide array of evaluation approaches that are capable of answering the many different questions that arise during the consecutive study phases of eHealth development and implementation. However, evaluators seem to struggle in choosing suitable evaluation approaches in relation to a specific study phase. OBJECTIVE: The objective of this project was to provide a structured overview of the existing eHealth evaluation approaches, with the aim of assisting eHealth evaluators in selecting a suitable approach for evaluating their eHealth solution at a specific evaluation study phase. METHODS: Three consecutive steps were followed. Step 1 was a systematic scoping review, summarizing existing eHealth evaluation approaches. Step 2 was a concept mapping study asking eHealth researchers about approaches for evaluating eHealth. In step 3, the results of step 1 and 2 were used to develop an "eHealth evaluation cycle" and subsequently compose the online "eHealth methodology guide." RESULTS: The scoping review yielded 57 articles describing 50 unique evaluation approaches. The concept mapping study questioned 43 eHealth researchers, resulting in 48 unique approaches. After removing duplicates, 75 unique evaluation approaches remained. Thereafter, an "eHealth evaluation cycle" was developed, consisting of six evaluation study phases: conceptual and planning, design, development and usability, pilot (feasibility), effectiveness (impact), uptake (implementation), and all phases. Finally, the "eHealth methodology guide" was composed by assigning the 75 evaluation approaches to the specific study phases of the "eHealth evaluation cycle." CONCLUSIONS: Seventy-five unique evaluation approaches were found in the literature and suggested by eHealth researchers, which served as content for the online "eHealth methodology guide." By assisting evaluators in selecting a suitable evaluation approach in relation to a specific study phase of the "eHealth evaluation cycle," the guide aims to enhance the quality, safety, and successful long-term implementation of novel eHealth solutions.

Chemotherapy effectiveness in trial-underrepresented groups with early breast cancer: A retrospective cohort study.

BACKGROUND: Adjuvant chemotherapy in early stage breast cancer has been shown to reduce mortality in a large meta-analysis of over 100 randomised trials. However, these trials largely excluded patients aged 70 years and over or with higher levels of comorbidity. There is therefore uncertainty about whether the effectiveness of adjuvant chemotherapy generalises to these groups, hindering patient and clinician decision-making. This study utilises administrative healthcare data-real world data (RWD)-and econometric methods for causal analysis to estimate treatment effectiveness in these trial-underrepresented groups. METHODS AND FINDINGS: Women with early breast cancer aged 70 years and over and those under 70 years with a high level of comorbidity were identified and their records extracted from Scottish Cancer Registry (2001-2015) data linked to other routine health records. A high level of comorbidity was defined as scoring 1 or more on the Charlson comorbidity index, being in the top decile of inpatient stays, and/or having 5 or more visits to specific outpatient clinics, all within the 5 years preceding breast cancer diagnosis. Propensity score matching (PSM) and instrumental variable (IV) analysis, previously identified as feasible and valid in this setting, were used in conjunction with Cox regression to estimate hazard ratios for death from breast cancer and death from all causes. The analysis adjusts for age, clinical prognostic factors, and socioeconomic deprivation; the IV method may also adjust for unmeasured confounding factors. Cohorts of 9,653 and 7,965 were identified for women aged 70 years and over and those with high comorbidity, respectively. In the ≥70/high comorbidity cohorts, median follow-up was 5.17/6.53 years and there were 1,935/740 deaths from breast cancer. For women aged 70 years and over, the PSM-estimated HR was 0.73 (95% CI 0.64-0.95), while for women with high comorbidity it was 0.67 (95% CI 0.51-0.86). This translates to a mean predicted benefit in terms of overall survival at 10 years of approximately3% (percentage points) and 4%, respectively. A limitation of this analysis is that use of observational data means uncertainty remains both from sampling uncertainty and from potential bias from residual confounding. CONCLUSIONS: The results of this study, as RWD, should be interpreted with caution and in the context of existing and emerging randomised data. The relative effectiveness of adjuvant chemotherapy in reducing mortality in patients with early stage breast cancer appears to be generalisable to the selected trial-underrepresented groups.

Correction to: How can clinicians, specialty societies and others evaluate and improve the quality of apps for patient use?

Since the publication of this article [1] it has come to my attention that it contains an error in which the y-axis in Fig. 1 was inverted, thus incorrectly displaying a weak negative correlation rather than a weak positive one. This error was introduced as the order of the data on which Fig. 2 was based [2] was misread. The corrected version of Fig. 2 can be seen below, in which a weak positive correlation is now displayed. This does not change the general point, that app users and app stores appear to take little notice of the source of information on which apps are based. I apologise to readers for this error.

Preserving the Open Access Benefits Pioneered by the Journal of Medical Internet Research and Discouraging Fraudulent Journals.

The Journal of Medical Internet Research (JMIR) was an early pioneer of open access online publishing, and two decades later, some readers and authors may have forgotten the challenges of previous scientific publishing models. This commentary summarizes the many advantages of open access publishing for each of the main stakeholders in scientific publishing and reminds us that, like every innovation, there are disadvantages that we need to guard against, such as the problem of fraudulent journals. This paper then reviews the potential impact of some current initiatives, such as Plan S and JMIRx, concluding with some suggestions to help new open-access publishers ensure that the advantages of open access publishing outweigh the challenges.

Independent validation of the PREDICT breast cancer prognosis prediction tool in 45,789 patients using Scottish Cancer Registry data.

BACKGROUND: PREDICT is a widely used online prognostication and treatment benefit tool for patients with early stage breast cancer. The aim of this study was to conduct an independent validation exercise of the most up-to-date version of the PREDICT algorithm (version 2) using real-world outcomes from the Scottish population of women with breast cancer. METHODS: Patient data were obtained for all Scottish Cancer Registry (SCR) records with a diagnosis of primary invasive breast cancer diagnosed in the period between January 2001 and December 2015. Prognostic scores were calculated using the PREDICT version 2 algorithm. External validity was assessed by statistical analysis of discrimination and calibration. Discrimination was assessed by area under the receiver-operator curve (AUC). Calibration was assessed by comparing the predicted number of deaths to the observed number of deaths across relevant sub-groups. RESULTS: A total of 45,789 eligible cases were selected from 61,437 individual records. AUC statistics ranged from 0.74 to 0.77. Calibration results showed relatively close agreement between predicted and observed deaths. The 5-year complete follow-up sample reported some overestimation (11.5%), while the 10-year complete follow-up sample displayed more limited overestimation (1.7%). CONCLUSIONS: Validation results suggest that the PREDICT tool remains essentially relevant for contemporary patients with early stage breast cancer.

How can clinicians, specialty societies and others evaluate and improve the quality of apps for patient use?

BACKGROUND: Health-related apps have great potential to enhance health and prevent disease globally, but their quality currently varies too much for clinicians to feel confident about recommending them to patients. The major quality concerns are dubious app content, loss of privacy associated with widespread sharing of the patient data they capture, inaccurate advice or risk estimates and the paucity of impact studies. This may explain why current evidence about app use by people with health-related conditions is scanty and inconsistent. MAIN TEXT: There are many concerns about health-related apps designed for use by patients, such as poor regulation and implicit trust in technology. However, there are several actions that various stakeholders, including users, developers, health professionals and app distributors, can take to tackle these concerns and thus improve app quality. This article focuses on the use of checklists that can be applied to apps, novel evaluation methods and suggestions for how clinical specialty organisations can develop a low-cost curated app repository with explicit risk and quality criteria. CONCLUSIONS: Clinicians and professional societies must act now to ensure they are using good quality apps, support patients in choosing between available apps and improve the quality of apps under development. Funders must also invest in research to answer important questions about apps, such as how clinicians and patients decide which apps to use and which app factors are associated with effectiveness.

Breast cancer survivors' recollection of their illness and therapy seven years after enrolment into a randomised controlled clinical trial.

BACKGROUND: Little is known about the subjective experience of breast cancer survivors after primary treatment. However, these experiences are important because they shape their communication about their illness in everyday life, usage and acceptance of healthcare, and expectations of new generations of patients. The present study investigated this topic by combining qualitative and quantitative methods. METHODS: Breast cancer survivors in Bavaria, Germany were mailed a questionnaire up to seven years after enrolment into a randomised controlled clinical trial and start of their therapy. This enquired about their worst experiences during the breast cancer episode, positive aspects of the illness and any advice they would give to newly diagnosed patients. A category system for themes was systematically created and answers were categorised by two independent raters. Frequencies of key categories were then quantitatively analysed using descriptive statistics. In addition, local treating physicians gave their opinion on the response categories chosen by their patients. RESULTS: 133 (80%) of 166 eligible patients who survived up to seven years returned the questionnaire. The most prominent worst experience reported by survivors was psychological distress (i.e. anxiety, uncertainty; prevalence 38%) followed by chemotherapy (25%), and cancer diagnosis (18%). Positive aspects of the illness were reported by 48% with the most frequent including change in life priorities (50%) and social support (22%). The most frequent advice survivors gave was fighting spirit (i.e. think positive, never give up; prevalence 42%). Overall, physicians' estimates of the frequency of these responses corresponded well with survivors' answers. CONCLUSIONS: Although physicians' understanding of breast cancer patients was good, psychological distress and chemotherapy-related side effects were remembered as particularly burdensome by a substantial part of survivors. On the one hand, patients' quality of life needs to be assessed repeatedly during medical follow-up to identify such specific complaints also including specific recommendations to the physician for targeted psychosocial and medical support. On the other hand the advices and positive aspects of the disease, reported by the survivors, can be used to promote positive ways of coping with the illness.

App Usage Factor: A Simple Metric to Compare the Population Impact of Mobile Medical Apps.

BACKGROUND: One factor when assessing the quality of mobile apps is quantifying the impact of a given app on a population. There is currently no metric which can be used to compare the population impact of a mobile app across different health care disciplines. OBJECTIVE: The objective of this study is to create a novel metric to characterize the impact of a mobile app on a population. METHODS: We developed the simple novel metric, app usage factor (AUF), defined as the logarithm of the product of the number of active users of a mobile app with the median number of daily uses of the app. The behavior of this metric was modeled using simulated modeling in Python, a general-purpose programming language. Three simulations were conducted to explore the temporal and numerical stability of our metric and a simulated app ecosystem model using a simulated dataset of 20,000 apps. RESULTS: Simulations confirmed the metric was stable between predicted usage limits and remained stable at extremes of these limits. Analysis of a simulated dataset of 20,000 apps calculated an average value for the app usage factor of 4.90 (SD 0.78). A temporal simulation showed that the metric remained stable over time and suitable limits for its use were identified. CONCLUSIONS: A key component when assessing app risk and potential harm is understanding the potential population impact of each mobile app. Our metric has many potential uses for a wide range of stakeholders in the app ecosystem, including users, regulators, developers, and health care professionals. Furthermore, this metric forms part of the overall estimate of risk and potential for harm or benefit posed by a mobile medical app. We identify the merits and limitations of this metric, as well as potential avenues for future validation and research.

mHealth and mobile medical Apps: a framework to assess risk and promote safer use.

The use of mobile medical apps by clinicians and others has grown considerably since the introduction of mobile phones. Medical apps offer clinicians the ability to access medical knowledge and patient data at the point of care, but several studies have highlighted apps that could compromise patient safety and are potentially dangerous. This article identifies a range of different kinds of risks that medical apps can contribute to and important contextual variables that can modify these risks. We have also developed a simple generic risk framework that app users, developers, and other stakeholders can use to assess the likely risks posed by a specific app in a specific context. This should help app commissioners, developers, and users to manage risks and improve patient safety.

Which computable biomedical knowledge objects will be regulated? Results of a UK workshop discussing the regulation of knowledge libraries and software as a medical device.

Introduction: To understand when knowledge objects in a computable biomedical knowledge library are likely to be subject to regulation as a medical device in the United Kingdom. Methods: A briefing paper was circulated to a multi-disciplinary group of 25 including regulators, lawyers and others with insights into device regulation. A 1-day workshop was convened to discuss questions relating to our aim. A discussion paper was drafted by lead authors and circulated to other authors for their comments and contributions. Results: This article reports on those deliberations and describes how UK device regulators are likely to treat the different kinds of knowledge objects that may be stored in computable biomedical knowledge libraries. While our focus is the likely approach of UK regulators, our analogies and analysis will also be relevant to the approaches taken by regulators elsewhere. We include a table examining the implications for each of the four knowledge levels described by Boxwala in 2011 and propose an additional level. Conclusions: If a knowledge object is described as directly executable for a medical purpose to provide decision support, it will generally be in scope of UK regulation as "software as a medical device." However, if the knowledge object consists of an algorithm, a ruleset, pseudocode or some other representation that is not directly executable and whose developers make no claim that it can be used for a medical purpose, it is not likely to be subject to regulation. We expect similar reasoning to be applied by regulators in other countries.

Modelling clinical narrative as computable knowledge: The NICE computable implementation guidance project.

Introduction: Translating narrative clinical guidelines to computable knowledge is a long-standing challenge that has seen a diverse range of approaches. The UK National Institute for Health and Care Excellence (NICE) Content Advisory Board (CAB) aims ultimately to (1) guide clinical decision support and other software developers to increase traceability, fidelity and consistency in supporting clinical use of NICE recommendations, (2) guide local practice audit and intervention to reduce unwarranted variation, (3) provide feedback to NICE on how future recommendations should be developed. Objectives: The first phase of work was to explore a range of technical approaches to transition NICE toward the production of natively digital content. Methods: Following an initial 'collaborathon' in November 2022, the NICE Computable Implementation Guidance project (NCIG) was established. We held a series of workstream calls approximately fortnightly, focusing on (1) user stories and trigger events, (2) information model and definitions, (3) horizon-scanning and output format. A second collaborathon was held in March 2023 to consolidate progress across the workstreams and agree residual actions to complete. Results: While we initially focussed on technical implementation standards, we decided that an intermediate logical model was a more achievable first step in the journey from narrative to fully computable representation. NCIG adopted the WHO Digital Adaptation Kit (DAK) as a technology-agnostic method to model user scenarios, personae, processes and workflow, core data elements and decision-support logic. Further work will address indicators, such as prescribing compliance, and implementation in document templates for primary care patient record systems. Conclusions: The project has shown that the WHO DAK, with some modification, is a promising approach to build technology-neutral logical specifications of NICE recommendations. Implementation of concurrent computable modelling by multidisciplinary teams during guideline development poses methodological and cultural questions that are complex but tractable given suitable will and leadership.

Computer-generated reminders delivered on paper to healthcare professionals; effects on professional practice and health care outcomes.

BACKGROUND: Clinical practice does not always reflect best practice and evidence, partly because of unconscious acts of omission, information overload, or inaccessible information. Reminders may help clinicians overcome these problems by prompting the doctor to recall information that they already know or would be expected to know and by providing information or guidance in a more accessible and relevant format, at a particularly appropriate time. OBJECTIVES: To evaluate the effects of reminders automatically generated through a computerized system and delivered on paper to healthcare professionals on processes of care (related to healthcare professionals' practice) and outcomes of care (related to patients' health condition). SEARCH METHODS: For this update the EPOC Trials Search Co-ordinator searched the following databases between June 11-19, 2012: The Cochrane Central Register of Controlled Trials (CENTRAL) and Cochrane Library (Economics, Methods, and Health Technology Assessment sections), Issue 6, 2012; MEDLINE, OVID (1946- ), Daily Update, and In-process; EMBASE, Ovid (1947- ); CINAHL, EbscoHost (1980- ); EPOC Specialised Register, Reference Manager, and INSPEC, Engineering Village. The authors reviewed reference lists of related reviews and studies.  SELECTION CRITERIA: We included individual or cluster-randomized controlled trials (RCTs) and non-randomized controlled trials (NRCTs) that evaluated the impact of computer-generated reminders delivered on paper to healthcare professionals on processes and/or outcomes of care. DATA COLLECTION AND ANALYSIS: Review authors working in pairs independently screened studies for eligibility and abstracted data. We contacted authors to obtain important missing information for studies that were published within the last 10 years. For each study, we extracted the primary outcome when it was defined or calculated the median effect size across all reported outcomes. We then calculated the median absolute improvement and interquartile range (IQR) in process adherence across included studies using the primary outcome or median outcome as representative outcome. MAIN RESULTS: In the 32 included studies, computer-generated reminders delivered on paper to healthcare professionals achieved moderate improvement in professional practices, with a median improvement of processes of care of 7.0% (IQR: 3.9% to 16.4%). Implementing reminders alone improved care by 11.2% (IQR 6.5% to 19.6%) compared with usual care, while implementing reminders in addition to another intervention improved care by 4.0% only (IQR 3.0% to 6.0%) compared with the other intervention. The quality of evidence for these comparisons was rated as moderate according to the GRADE approach. Two reminder features were associated with larger effect sizes: providing space on the reminder for provider to enter a response (median 13.7% versus 4.3% for no response, P value = 0.01) and providing an explanation of the content or advice on the reminder (median 12.0% versus 4.2% for no explanation, P value = 0.02). Median improvement in processes of care also differed according to the behaviour the reminder targeted: for instance, reminders to vaccinate improved processes of care by 13.1% (IQR 12.2% to 20.7%) compared with other targeted behaviours. In the only study that had sufficient power to detect a clinically significant effect on outcomes of care, reminders were not associated with significant improvements. AUTHORS' CONCLUSIONS: There is moderate quality evidence that computer-generated reminders delivered on paper to healthcare professionals achieve moderate improvement in process of care. Two characteristics emerged as significant predictors of improvement: providing space on the reminder for a response from the clinician and providing an explanation of the reminder's content or advice. The heterogeneity of the reminder interventions included in this review also suggests that reminders can improve care in various settings under various conditions.

The Scottish Emergency Care Summary--an evaluation of a national shared record system aiming to improve patient care: technology report.

BACKGROUND: In Scotland, out-of-hours calls are all triaged by the National Health Service emergency service (NHS24) but the clinicians receiving calls have no direct access to patient records. OBJECTIVE: To improve the safety of patient care in unscheduled consultations when the usual primary care record is not available. TECHNOLOGY: The Emergency Care Summary (ECS) is a record system offering controlled access to medication and adverse reactions details for nearly every person registered with a general practice in Scotland. It holds a secure central copy of these parts of the general practitioner (GP) practice record and is updated automatically twice daily. It is accessible under specified unplanned clinical circumstances by clinicians working in out-of-hours organisations, NHS24 and accident and emergency departments if they have consent from the patient and a current legitimate relationship for that patient's care. APPLICATION: We describe the design of the security model, management of data quality, deployment, costs and clinical benefits of the ECS over four years nationwide in Scotland, to inform the debate on the safe and effective sharing of health data in other nations. EVALUATION: Forms were emailed to 300 NHS24 clinicians and 81% of the 113 respondents said that the ECS was helpful or very helpful and felt that it changed their clinical management in 20% of cases. CONCLUSION: The ECS is acceptable to patients and helpful for clinicians and is used routinely for unscheduled care when normal medical records are unavailable. Benefits include more efficient assessment and reduced drug interaction, adverse reaction and duplicate prescribing.

The provision and impact of online patient access to their electronic health records (EHR) and transactional services on the quality and safety of health care: systematic review protocol.

BACKGROUND: Innovators have piloted improvements in communication, changed patterns of practice and patient empowerment from online access to electronic health records (EHR). International studies of online services, such as prescription ordering, online appointment booking and secure communications with primary care, show good uptake of email consultations, accessing test results and booking appointments; when technologies and business process are in place. Online access and transactional services are due to be rolled out across England by 2015; this review seeks to explore the impact of online access to health records and other online services on the quality and safety of primary health care. OBJECTIVE: To assess the factors that may affect the provision of online patient access to their EHR and transactional services, and the impact of such access on the quality and safety of health care. METHOD: Two reviewers independently searched 11 international databases during the period 1999-2012. A range of papers including descriptive studies using qualitative or quantitative methods, hypothesis-testing studies and systematic reviews were included. A detailed eligibility criterion will be used to shape study inclusion. A team of experts will review these papers for eligibility, extract data using a customised extraction form and use the Grading of Recommendations Assessment, Development and Evaluation (GRADE) instrument to determine the quality of the evidence and the strengths of any recommendation. Data will then be descriptively summarised and thematically synthesised. Where feasible, we will perform a quantitative meta-analysis. Prospero (International Prospective Register of Systematic Reviews) registration number: crd42012003091.

Open source, open standards, and health care information systems.

Recognition of the improvements in patient safety, quality of patient care, and efficiency that health care information systems have the potential to bring has led to significant investment. Globally the sale of health care information systems now represents a multibillion dollar industry. As policy makers, health care professionals, and patients, we have a responsibility to maximize the return on this investment. To this end we analyze alternative licensing and software development models, as well as the role of standards. We describe how licensing affects development. We argue for the superiority of open source licensing to promote safer, more effective health care information systems. We claim that open source licensing in health care information systems is essential to rational procurement strategy.

Automation bias - a hidden issue for clinical decision support system use.

Automation bias - the tendency to over-rely on automation - has been studied in a variety of academic fields. Clinical Decision Support Systems aim to benefit the clinical decision making process. Although most research shows overall improved performance with use, there is often a failure to recognize the new errors that CDSS can introduce, and the healthcare field has a gap in this research. This paper outlines some of the most compelling theoretical factors in the literature involved in automation bias, and builds a simple model to be tested empirically. Ultimately, this will uncover the mechanisms by which this bias operates and help CDSS producers and healthcare practitioners optimize the medical decision making process.

An overview of the National COVID-19 Chest Imaging Database: data quality and cohort analysis.

BACKGROUND: The National COVID-19 Chest Imaging Database (NCCID) is a centralized database containing mainly chest X-rays and computed tomography scans from patients across the UK. The objective of the initiative is to support a better understanding of the coronavirus SARS-CoV-2 disease (COVID-19) and the development of machine learning technologies that will improve care for patients hospitalized with a severe COVID-19 infection. This article introduces the training dataset, including a snapshot analysis covering the completeness of clinical data, and availability of image data for the various use-cases (diagnosis, prognosis, longitudinal risk). An additional cohort analysis measures how well the NCCID represents the wider COVID-19-affected UK population in terms of geographic, demographic, and temporal coverage. FINDINGS: The NCCID offers high-quality DICOM images acquired across a variety of imaging machinery; multiple time points including historical images are available for a subset of patients. This volume and variety make the database well suited to development of diagnostic/prognostic models for COVID-associated respiratory conditions. Historical images and clinical data may aid long-term risk stratification, particularly as availability of comorbidity data increases through linkage to other resources. The cohort analysis revealed good alignment to general UK COVID-19 statistics for some categories, e.g., sex, whilst identifying areas for improvements to data collection methods, particularly geographic coverage. CONCLUSION: The NCCID is a growing resource that provides researchers with a large, high-quality database that can be leveraged both to support the response to the COVID-19 pandemic and as a test bed for building clinically viable medical imaging models.