The value of standards for health datasets in artificial intelligence-based applications.

Artificial intelligence as a medical device is increasingly being applied to healthcare for diagnosis, risk stratification and resource allocation. However, a growing body of evidence has highlighted the risk of algorithmic bias, which may perpetuate existing health inequity. This problem arises in part because of systemic inequalities in dataset curation, unequal opportunity to participate in research and inequalities of access. This study aims to explore existing standards, frameworks and best practices for ensuring adequate data diversity in health datasets. Exploring the body of existing literature and expert views is an important step towards the development of consensus-based guidelines. The study comprises two parts: a systematic review of existing standards, frameworks and best practices for healthcare datasets; and a survey and thematic analysis of stakeholder views of bias, health equity and best practices for artificial intelligence as a medical device. We found that the need for dataset diversity was well described in literature, and experts generally favored the development of a robust set of guidelines, but there were mixed views about how these could be implemented practically. The outputs of this study will be used to inform the development of standards for transparency of data diversity in health datasets (the STANDING Together initiative).

Artificial intelligence for radiological paediatric fracture assessment: a systematic review.

BACKGROUND: Majority of research and commercial efforts have focussed on use of artificial intelligence (AI) for fracture detection in adults, despite the greater long-term clinical and medicolegal implications of missed fractures in children. The objective of this study was to assess the available literature regarding diagnostic performance of AI tools for paediatric fracture assessment on imaging, and where available, how this compares with the performance of human readers. MATERIALS AND METHODS: MEDLINE, Embase and Cochrane Library databases were queried for studies published between 1 January 2011 and 2021 using terms related to 'fracture', 'artificial intelligence', 'imaging' and 'children'. Risk of bias was assessed using a modified QUADAS-2 tool. Descriptive statistics for diagnostic accuracies were collated. RESULTS: Nine eligible articles from 362 publications were included, with most (8/9) evaluating fracture detection on radiographs, with the elbow being the most common body part. Nearly all articles used data derived from a single institution, and used deep learning methodology with only a few (2/9) performing external validation. Accuracy rates generated by AI ranged from 88.8 to 97.9%. In two of the three articles where AI performance was compared to human readers, sensitivity rates for AI were marginally higher, but this was not statistically significant. CONCLUSIONS: Wide heterogeneity in the literature with limited information on algorithm performance on external datasets makes it difficult to understand how such tools may generalise to a wider paediatric population. Further research using a multicentric dataset with real-world evaluation would help to better understand the impact of these tools.

Implementation of prognostic machine learning algorithms in paediatric chronic respiratory conditions: a scoping review.

Machine learning (ML) holds great potential for predicting clinical outcomes in heterogeneous chronic respiratory diseases (CRD) affecting children, where timely individualised treatments offer opportunities for health optimisation. This paper identifies rate-limiting steps in ML prediction model development that impair clinical translation and discusses regulatory, clinical and ethical considerations for ML implementation. A scoping review of ML prediction models in paediatric CRDs was undertaken using the PRISMA extension scoping review guidelines. From 1209 results, 25 articles published between 2013 and 2021 were evaluated for features of a good clinical prediction model using the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines.Most of the studies were in asthma (80%), with few in cystic fibrosis (12%), bronchiolitis (4%) and childhood wheeze (4%). There were inconsistencies in model reporting and studies were limited by a lack of validation, and absence of equations or code for replication. Clinician involvement during ML model development is essential and diversity, equity and inclusion should be assessed at each step of the ML pipeline to ensure algorithms do not promote or amplify health disparities among marginalised groups. As ML prediction studies become more frequent, it is important that models are rigorously developed using published guidelines and take account of regulatory frameworks which depend on model complexity, patient safety, accountability and liability.

Review of study reporting guidelines for clinical studies using artificial intelligence in healthcare.

High-quality research is essential in guiding evidence-based care, and should be reported in a way that is reproducible, transparent and where appropriate, provide sufficient detail for inclusion in future meta-analyses. Reporting guidelines for various study designs have been widely used for clinical (and preclinical) studies, consisting of checklists with a minimum set of points for inclusion. With the recent rise in volume of research using artificial intelligence (AI), additional factors need to be evaluated, which do not neatly conform to traditional reporting guidelines (eg, details relating to technical algorithm development). In this review, reporting guidelines are highlighted to promote awareness of essential content required for studies evaluating AI interventions in healthcare. These include published and in progress extensions to well-known reporting guidelines such as Standard Protocol Items: Recommendations for Interventional Trials-AI (study protocols), Consolidated Standards of Reporting Trials-AI (randomised controlled trials), Standards for Reporting of Diagnostic Accuracy Studies-AI (diagnostic accuracy studies) and Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis-AI (prediction model studies). Additionally there are a number of guidelines that consider AI for health interventions more generally (eg, Checklist for Artificial Intelligence in Medical Imaging (CLAIM), minimum information (MI)-CLAIM, MI for Medical AI Reporting) or address a specific element such as the 'learning curve' (Developmental and Exploratory Clinical Investigation of Decision-AI) . Economic evaluation of AI health interventions is not currently addressed, and may benefit from extension to an existing guideline. In the face of a rapid influx of studies of AI health interventions, reporting guidelines help ensure that investigators and those appraising studies consider both the well-recognised elements of good study design and reporting, while also adequately addressing new challenges posed by AI-specific elements.

Development of a data utility framework to support effective health data curation.

OBJECTIVES: The value of healthcare data is being increasingly recognised, including the need to improve health dataset utility. There is no established mechanism for evaluating healthcare dataset utility making it difficult to evaluate the effectiveness of activities improving the data. To describe the method for generating and involving the user community in developing a proposed framework for evaluation and communication of healthcare dataset utility for given research areas. METHODS: Aninitial version of a matrix to review datasets across a range of dimensions wasdeveloped based on previous published findings regarding healthcare data. Thiswas used to initiate a design process through interviews and surveys with datausers representing a broad range of user types and use cases, to help develop afocused framework for characterising datasets. RESULTS: Following 21 interviews, 31 survey responses and testing on 43 datasets, five major categories and 13 subcategories were identified as useful for a dataset, including Data Model, Completeness and Linkage. Each sub-category was graded to facilitate rapid and reproducible evaluation of dataset utility for specific use-cases. Testing of applicability to >40 existing datasets demonstrated potential usefulness for subsequent evaluation in real-world practice. DISCUSSION: Theresearch has developed an evidenced-based initial approach for a framework tounderstand the utility of a healthcare dataset. It likely to require further refinementfollowing wider application and additional categories may be required. CONCLUSION: The process has resulted in a user-centred designed framework for objectively evaluating the likely utility of specific healthcare datasets, and therefore, should be of value both for potential users of health data, and for data custodians to identify the areas to provide the optimal value for data curation investment.

Phishing in healthcare organisations: threats, mitigation and approaches.

INTRODUCTION: Healthcare data have significant value as a potential target for hackers. Phishing is a method of exploitation for malicious reasons using targeted communications (email/messaging). This study reports on an internal evaluation targeting hospital staff and summarises peer-reviewed literature regarding phishing and healthcare. METHODS: An assessment was performed as part of cybersecurity activity during a designated test period using multiple credential harvesting approaches through staff email. We also searched the medical-related literature to identify relevant phishing-related publications. RESULTS: During the 1-month testing period, the organisation received 858 200 emails: 139 400 (16%) marketing, 18 871 (2%) identified as potential threats. Of 143 million internet transactions, around 5 million (3%) were suspected threats. 468 employee email addresses were identified from public data and targeted through phishing using a range of payloads including attachments and malicious links; however, no credentials were recovered or malicious files downloaded. Several hospital employees were, however, identified on social media profiles, including some tricked into accepting false friend requests. DISCUSSION: Healthcare organisations are increasingly moving to digital systems, but healthcare professionals have limited awareness of threats. Increasing emphasis on 'cyberhygiene' and information governance through mandatory training increases understanding of these risks. While no credentials were harvested in this study, since up to 5% of emails/internet traffic are suspicious, the need for robust firewalls, cybersecurity infrastructure, IT policies and, most importantly of all, staff training, is emphasised. CONCLUSION: Hospitals receive a significant volume of potentially malicious emails. While many staff appear to be aware of phishing and respond appropriately, ongoing education is required across the spectrum of cybersecurity, with specific emphasis around 'leakage' of information on social media.

Minimally invasive autopsy for fetuses and children based on a combination of post-mortem MRI and endoscopic examination: a feasibility study.

BACKGROUND: Less invasive perinatal and paediatric autopsy methods, such as imaging alongside targeted endoscopy and organ biopsy, may address declining consent rates for traditional autopsy, but their acceptability and accuracy are not known. OBJECTIVES: The aims of this study were to provide empirical data on the acceptability and likely uptake for different types of autopsy among key stakeholders (study 1); and to analyse existing autopsy data sources to provide estimates of the potential efficacy of less invasive autopsy (LIA) and its projected utility in clinical practice (study 2). REVIEW METHODS: Study 1: this was a mixed-methods study. Parents were involved in research design and interpretation of findings. Substudy 1: a cross-sectional survey of 859 parents who had experienced miscarriage, termination of pregnancy for fetal anomaly, stillbirth, infant or child death, and interviews with 20 responders. Substudy 2: interviews with 25 health professionals and four coroners. Substudy 3: interviews with 16 religious leaders and eight focus groups, with 76 members of the Muslim and Jewish community. Study 2: a retrospective analysis of national data in addition to detailed information from an existing in-house autopsy database of > 5000 clinical cases that had undergone standard autopsy to determine the proportion of cases by clinical indication group for which tissue sampling of specific internal organs significantly contributed to the diagnosis. RESULTS: Substudy 1: 91% of participants indicated that they would consent to some form of LIA, 54% would consent to standard autopsy, 74% to minimally invasive autopsy (MIA) and 77% to non-invasive autopsy (NIA). Substudy 2: participants viewed LIA as a positive development, but had concerns around the limitations of the technology and de-skilling the workforce. Cost implications, skills and training requirements were identified as implementation challenges. Substudy 3: religious leaders agreed that NIA was religiously permissible, but MIA was considered less acceptable. Community members indicated that they might consent to NIA if the body could be returned for burial within 24 hours. Study 2: in 5-10% of cases of sudden unexplained death in childhood and sudden unexplained death in infants, the final cause of death is determined by routine histological sampling of macroscopically normal organs, predominantly the heart and lungs, and in this group routine histological sampling therefore remains an important aspect of investigation. In contrast, routine histological examination of macroscopically normal organs rarely (< 0.5%) provides the cause of death in fetal cases, making LIA and NIA approaches potentially highly applicable. LIMITATIONS: A key limitation of the empirical research is that it is hypothetical. Further research is required to determine actual uptake. Furthermore, because of the retrospective nature of the autopsy data set, findings regarding the likely contribution of organ sampling to final diagnosis are based on extrapolation of findings from historical autopsies, and prospective data collection is required to validate the conclusions. CONCLUSIONS: LIA is viable and acceptable (except for unexplained deaths), and likely to increase uptake. Further health economic, performance and implementation studies are required to determine the optimal service configuration required to offer this as routine clinical care. FUNDING: The National Institute for Health Research Health Technology Assessment programme.

Autopsy (post-mortem) examination of babies and children who die is often necessary to help doctors or coroners find out the cause of death. It may also be useful for research. However, many bereaved parents dislike the idea of their child being cut and some religious communities prohibit the procedure. Over the past 30 years, consent rates for autopsies have declined. In order to address parental concerns and declining uptake, a number of less invasive options have been developed. These include X-ray and magnetic resonance imaging, by doing keyhole internal examination and needle organ biopsy. However, it is not known to what extent such methods are acceptable to parents, nor how accurate they are. We surveyed the attitudes of bereaved parents and religious group leaders to such less invasive methods. The less invasive option was considered acceptable and would be chosen by almost 1000 bereaved parents. Such an approach is also acceptable to those religious groups for whom standard autopsy examination is not. We also examined a database of > 5000 standard autopsies to determine the extent to which specific internal organ biopsy contributed to the diagnosis. In > 5000 standard autopsies, traditional organ biopsy rarely contributed to determination of the cause of death or the main diagnosis. Therefore, a more limited and targeted tissue sampling protocol could be introduced without significant reduction in the accuracy of final diagnosis. The specific approaches required will depend on individual circumstances and are likely to include a range, from targeted organ biopsy with an open incision, through incisionless image-guided needle biopsies, to non-invasive imaging-only techniques. Future studies may focus on how the NHS could implement offering less invasive approaches nationally, what the cost­benefit of such an approach could be and what the impact could be on real-world uptake if this were to be offered routinely.

Feasibility of Postmortem Imaging Assessment of Brain: Liver Volume Ratios with Pathological Validation.

INTRODUCTION: Organ volumes at postmortem magnetic resonance imaging (PMMR) should reflect autopsy organ weights, and thus brain:liver volume ratios on imaging could be a surrogate for weight volume ratios at autopsy to indicate fetal growth restriction (FGR). This study aims to determine whether imaging-based organ volume ratios can replace autopsy organ weight ratios. Materials and Meth ods: An unselected cohort of perinatal deaths underwent PMMR prior to autopsy. Semiautomated brain and liver volumes were compared to autopsy organ weights and ratios. Ratios were compared using Bland-Altman plots, and intra- and interobserver variability was assessed. RESULTS: A total 49 fetuses (25 male, 51%) at 17-42 weeks gestation were -assessed. There was a reasonable correlation between autopsy-derived brain:liver weight ratios (AB:LwR) and imaging-derived brain:liver volume ratios (IB:LvR; r = 0.8). The mean difference between AB:LwR and IB:LvR was +0.7 (95% limits of agreement range -1.5 to +2.9). In a small subset where FGR was present, the optimal IB:LvR ≥5.5 gave 83.3% sensitivity and 86.0% specificity for diagnosis. There was acceptable agreement within readers (mean difference in IB:LvRs 0.77 ± 2.21) and between readers -0.36 ± 0.68. CONCLUSION: IB:LvR provides a surrogate evaluation of AB:LwRs, and may be used as a marker of FGR where autopsy is declined.

Health professionals' and coroners' views on less invasive perinatal and paediatric autopsy: a qualitative study.

OBJECTIVE: To assess health professionals' and coroners' attitudes towards non-minimally and minimally invasive autopsy in the perinatal and paediatric setting. METHODS: A qualitative study using semistructured interviews. Data were analysed thematically. RESULTS: Twenty-five health professionals (including perinatal/paediatric pathologists and anatomical pathology technologists, obstetricians, fetal medicine consultants and bereavement midwives, intensive care consultants and family liaison nurses, a consultant neonatologist and a paediatric radiologist) and four coroners participated. Participants viewed less invasive methods of autopsy as a positive development in prenatal and paediatric care that could increase autopsy rates. Several procedural and psychological benefits were highlighted including improved diagnostic accuracy in some circumstances, potential for faster turnaround times, parental familiarity with imaging and laparoscopic approaches, and benefits to parents and faith groups who object to invasive approaches. Concerns around the limitations of the technology such not reaching the same levels of certainty as full autopsy, unsuitability of imaging in certain circumstances, the potential for missing a diagnosis (or misdiagnosis) and de-skilling the workforce were identified. Finally, a number of implementation issues were raised including skills and training requirements for pathologists and radiologists, access to scanning equipment, required computational infrastructure, need for a multidisciplinary approach to interpret results, cost implications, equity of access and acceptance from health professionals and hospital managers. CONCLUSION: Health professionals and coroners viewed less invasive autopsy as a positive development in perinatal and paediatric care. However, to inform implementation a detailed health economic analysis and further exploration of parental views, particularly in different religious groups, are required.

Assessment of body composition in Wistar rat offspring by DXA in relation to prenatal and postnatal nutritional manipulation.

BACKGROUND: This experimental study aims to investigate the impact of combinations of prenatal and postnatal food manipulation on body composition in rat offspring. METHODS: On day 12 of gestation, 100 timed pregnant rats were randomized into two nutritional groups: standard laboratory and 50% starved. Pups born to starved mothers were subdivided, based on birthweight (BiW), into fetal growth restricted (FGR) and non-FGR. Pups were born on day 21, cross-fostered, then left undisturbed lactating until the 26th postnatal day when they underwent dual-energy X-ray absorptiometry (DXA) examination. RESULTS: Prenatally control-fed animals had a significantly greater body weight at 26 d postnatally than the prenatally starved groups, irrespective of their postnatal diet (P < 0.001). Postnatal control diet was associated with significantly increased abdominal and total fat in non-FGR compared to FGR rats (P < 0.001). non-FGR/CONTROL rats showed higher values of abdominal fat than prenatally starved animals that were starved postnatally irrespective of their birth weight (P < 0.001). Postnatal control diet significantly increased total bone mineral content (BMC), head BMC, head area, abdominal BMC in non-FGR compared to FGR rats (P < 0.001). CONCLUSION: Interaction between prenatal and postnatal nutrition affects growth, abdominal adiposity, and bone accrual in Wistar rats' offspring at 26 d of life.

Routine perinatal and paediatric post-mortem radiography: detection rates and implications for practice.

BACKGROUND: Routine perinatal and paediatric post-mortem plain radiography allows for the diagnosis and assessment of skeletal dysplasias, fractures and other bony abnormalities. OBJECTIVE: The aim of this study was to review the diagnostic yield of this practice. MATERIALS AND METHODS: We identified 1,027 cases performed in a single institution over a 2½-year period, including babygrams (whole-body examinations) and full skeletal surveys. Images were reported prior to autopsy in all cases. Radiology findings were cross-referenced with the autopsy findings using an autopsy database. We scored each case from 0 to 4 according to the level of diagnostic usefulness. RESULTS: The overall abnormality rate was 126/1,027 (12.3%). There was a significantly higher rate of abnormality when a skeletal survey was performed (18%) rather than a babygram (10%; P < 0.01); 90% (665/739) of babygrams were normal. Of the 74 abnormal babygrams, we found 33 incidental non-contributory cases, 19 contributory, 20 diagnostic, and 2 false-positive cases. There were only 2 cases out of 739 (0.27%) in whom routine post-mortem imaging identified potentially significant abnormalities that would not have been detected if only selected imaging had been performed. A policy of performing selected, rather than routine, foetal post-mortem radiography could result in a significant cost saving. CONCLUSION: Routine post-mortem paediatric radiography in foetuses and neonates is neither diagnostically useful nor cost-effective. A more evidence-based, selective protocol should yield significant cost savings.

Immunohistochemical staining for diagnosis and prognostic assessment of hydatidiform moles: current evidence and future directions.

OBJECTIVE: To review the published data on studies examining immunohistochemical markers of hydatidiform moles for determination of diagnosis or prognosis, with regard to whether such investigations can provide clinically relevant information. STUDY DESIGN: Search of computerized literature databases to identify studies reporting on immunohistochemical findings in cases of hydatidiform mole followed by summarization and interpretation of the findings. RESULTS: Of 166 studies initially identified, 88 studies were included in the final analysis. Some markers, such as P57(KIP2), show distinct differences in expression between groups and are useful in clinical practice for the diagnosis of complete hydatidiform moles. Some markers appear to be associated with increased risk of progression to requiring chemotherapy, including increased expression of P53, EGFR, HER2, c-erbB-2 and telomerase and reduced expression of nm23. CONCLUSION: Despite technical issues complicating overall interpretation, such as small sample size and the uncertainty of classification of hydatidiform moles in many studies, convincing data is reported for several markers, which should guide future studies. Insufficient data exists to recommend changes in clinical management based on current evidence.

Diagnostic accuracy of post-mortem magnetic resonance imaging in fetuses, children and adults: a systematic review.

To determine, in a systematic review, the diagnostic accuracy, acceptability and cost-effectiveness of less invasive autopsy by post-mortem MR imaging, in fetuses, children and adults. We searched Medline, Embase, the Cochrane library and reference lists to identify all studies comparing post-mortem MR imaging with conventional autopsy, published between January 1990 and March 2009. 539 abstracts were identified; 15 papers met the inclusion criteria; data from 9 studies were extracted (total: 146 fetuses, 11 children and 24 adults). In accurately identifying the final cause of death or most clinically significant abnormality, post-mortem MR imaging had a sensitivity and specificity of 69% (95% CI-56%, 80%) and 95% (95% CI-88%, 98%) in fetuses, and 28% (95% CI-13%, 47%) and 64% (95% CI-23%, 94%) in children and adults, respectively; however the published data is limited to small, heterogenous and poorly designed studies. Insufficient data is available on acceptability and economic evaluation of post-mortem MR imaging. Well designed, large, prospective studies are required to evaluate the accuracy of post-mortem MR imaging, before it can be offered as a clinical tool.

Assessment of extrahepatic abdominal extension in primary malignant liver tumours of childhood.

Imaging is a crucial part of the evaluation of children with primary liver malignancies. In addition to staging the intrahepatic extent of the tumour, the 2005 revision of the PRETEXT system specifies certain criteria for assessment of extrahepatic disease. This article reviews the imaging findings of two of these criteria: direct extrahepatic extension of the tumour (E1, E1a, E2, E2a) and tumour rupture at diagnosis (H1).