Assessment of neonatal mortalities and stillbirths data quality in Offinso North District of Ghana

Background: Stillbirths and neonatal deaths when poorly documented or collated, negatively affect the quality of decision and interventions. This study sought to assess the quality of routine neonatal mortalities and stillbirth records in health facilities and propose interventions to improve the data quality gaps. Method: Descriptive cross-sectional study was employed. This study was carried out at three (3) purposively selected health facilities in Offinso North district. Stillbirths and neonatal deaths recorded in registers from 2015 to 2017, were recounted and compared with monthly aggregated data and District Health Information Management System 2 (DHIMS 2) data using a self-developed Excel Data Quality Assessment Tool (DQS). An observational checklist was used to collect primary data on completeness and availability. Accuracy ratio (verification factor), discrepancy rate, percentage availability and completeness of stillbirths and neonatal mortality data were computed using the DQS tool. Findings: The results showed high discrepancy rate of stillbirth data recorded in registers compared with monthly aggregated reports (12.5%), and monthly aggregated reports compared with DHIMS 2 (13.5%). Neonatal mortalities data were under-reported in monthly aggregated reports, but over-reported in DHIMS 2. Overall data completeness was about 84.6%, but only 68.5% of submitted reports were supervised by facility in-charges. Delivery and admission registers availability were 100% and 83.3% respectively. Conclusion: Quality of stillbirths and neonatal mortality data in the district is generally encouraging, but are not reliable for decision-making. Routine data quality audit is needed to reduce high discrepancies in stillbirth and neonatal mortality data in the district.

Assessing data accuracy in a large multi-institutional quality improvement registry: an update from the Pediatric Cardiac Critical Care Consortium (PC4)

Background: The Pediatric Cardiac Critical Care Consortium (PC4) is a multi-institutional quality improvement registry focused on the care delivered in the cardiac ICU for patients with CHD and acquired heart disease. To assess data quality, a rigorous procedure of data auditing has been in place since the inception of the consortium. Materials and methods: This report describes the data auditing process and quantifies the audit results for the initial 39 audits that took place after the transition from version one to version two of the registry’s database. Results: In total, 2219 total encounters were audited for an average of 57 encounters per site. The overall data accuracy rate across all sites was 99.4%, with a major discrepancy rate of 0.52%. A passing score is based on an overall accuracy of >97% (achieved by all sites) and a major discrepancy rate of <1.5% (achieved by 38 of 39 sites, with 35 of 39 sites having a major discrepancy rate of <1%). Fields with the highest discrepancy rates included arrhythmia type, cardiac arrest count, and current surgical status. Conclusions: The extensive PC4 auditing process, including initial and routinely scheduled follow-up audits of every participating site, demonstrates an extremely high level of accuracy across a broad array of audited fields and supports the continued use of consortium data to identify best practices in paediatric cardiac critical care.

The Intersection between Instructor Expectations and Student Interpretations of Academic Skills

Numerous studies exist on how and to what extent course instructors in higher education are embedding or directly teaching writing, learning and information literacy skills in their courses (Cilliers, 2012; Crosthwaite et al., 2006; Mager & Spronken-Smith, 2014). Yet, disparity within the literature demonstrates that there is no consistent approach to the scaffolded development of these necessary skills within courses, programs, disciplines, or across disciplines. This study sought to explore the skills expectations of instructors and whether students are capable of identifying or articulating the academic skills they are required to develop in to succeed in third-year undergraduate university courses. We discovered a discrepancy rate of approximately 63% between instructor and student responses when exploring differences in instructor expectations and student interpretations of academic skills indicated on course outlines. Data from this study suggests that instructors and students do not always share the same understanding of the skills required to complete course work and to be successful in assessments. With the support of learning, writing, and research specialists, instructors can embed academic skill development in the curriculum.

On Self-Distilling Graph Neural Network

Recently, the teacher-student knowledge distillation framework has demonstrated its potential in training Graph Neural Networks (GNNs). However, due to the difficulty of training over-parameterized GNN models, one may not easily obtain a satisfactory teacher model for distillation. Furthermore, the inefficient training process of teacher-student knowledge distillation also impedes its applications in GNN models. In this paper, we propose the first teacher-free knowledge distillation method for GNNs, termed GNN Self-Distillation (GNN-SD), that serves as a drop-in replacement of the standard training process. The method is built upon the proposed neighborhood discrepancy rate (NDR), which quantifies the non-smoothness of the embedded graph in an efficient way. Based on this metric, we propose the adaptive discrepancy retaining (ADR) regularizer to empower the transferability of knowledge that maintains high neighborhood discrepancy across GNN layers. We also summarize a generic GNN-SD framework that could be exploited to induce other distillation strategies. Experiments further prove the effectiveness and generalization of our approach, as it brings: 1) state-of-the-art GNN distillation performance with less training cost, 2) consistent and considerable performance enhancement for various popular backbones.

Could ante-mortem computed tomography be useful in forensic pathology of traumatic intracranial haemorrhage?

Background: Imaging techniques have proven valuable in forensic pathology practice, with computed tomography being preferred for forensic use. In the era of virtual autopsy and a low- to middle-income, resource-constrained country, a question arises as to whether ante-mortem computed tomography (ACT) could be cost-effective by reducing the number of invasive autopsies performed.Objective: The objective of this study was to assess the usefulness of ACT in forensic pathology by examining discrepancy rates between ACT scans and autopsy findings in cases of deceased individuals with traumatic intracranial haemorrhages and assess factors associated with discrepancies.Methods: Eighty-five cases of ACT and autopsy reports from 01 January 2014 to 31 December 2016 from the Polokwane Forensic Pathology Laboratory, South Africa, were analysed retrospectively. Using Cohen’s kappa statistics, measures of agreement and resultant discrepancy rates were determined. Also, the discrepancy patterns for each identified factor was also analysed.Results: The discrepancy rate between ACT and autopsy detection of haemorrhage was 24.71% while diagnostic categorisation of haemorrhage was 55.3%. Classification discrepancy was most observed in subarachnoid haemorrhages and least observed in extradural haemorrhages. A markedly reduced level of consciousness, hospital stay beyond two weeks and three or fewer years of doctors’ experience contributed to classification discrepancies.Conclusion: Ante-mortem computed tomography should be used only as an adjunct to autopsy findings. However, the low discrepancy rate seen for extradural haemorrhages implies that ACT may be useful in the forensic diagnosis of extradural haemorrhages.

Autopsy and premortem diagnostic discrepancy review in an Irish tertiary PICU.

Objective: 1) Review ante- and post-mortem diagnoses and assign a Goldman error classification. 2) Establish autopsy rates. Design: A retrospective analysis of autopsies performed on patients who died in Paediatric intensive care unit (PICU) between November 13th 2012 and October 31st 2018. We reviewed medical and autopsy data of all patients and Goldman classification of discrepancy between ante- and post-mortem diagnoses was assigned. Setting: Tertiary PICU. Patients: All patients that died in PICU within the designated timeframe. Interventions: Goldman error classification assignment. Measurements and main results: 396 deaths occurred in PICU from 8,329 (4.75%) admissions. 99 (25%) had an autopsy, 75 required by the coroner. All were included in the study. Fifty-three were male and 46 females. Fifty-three patients were transfers from external hospitals, 46 from our centre. Forty-one were neonates, 32 were <1 year of age, and 26 were >1 year of age. Median length of stay was 3 days. Eighteen were post cardiac surgery, and three post cardiac catheter procedure. Major diagnostic errors (Class I/II) were identified in 14 (14.1%), 2 (2%) Class I, and 12 (12.1%) were Class II errors. Class III and IV errors occurred in 28 (28.2%) patients. Complete concordance (Class V) occurred in 57 (57.5%) cases. Conclusion: The autopsy rate and the diagnostic discrepancy rate within our PICU is comparable to those previously reported. Our findings show the continuing value of autopsy in determining cause of death and providing greater diagnostic clarity. Given their value, post-mortem examinations, where indicated, should be considered part of a physician’s duty of care to families and future patients.

Endoscopic resections for superficial esophageal squamous cell epithelial neoplasia: focus on histological discrepancies between biopsy and resected specimens

Background Endoscopic resection has been used for high-grade intraepithelial neoplasia (HGIN) and superficial esophageal squamous cell carcinoma (ESCC) with limited risk of lymph node metastasis. However, some of these lesions cannot be accurately diagnosed based on forceps biopsy prior to treatment. In this study we aimed to investigate how to solve this histological discrepancy and avoid over- and under-treatment. Methods The medical records of patients with superficial esophageal squamous cell neoplasia who underwent endoscopic resection at our hospital from January 2012 to December 2019 were reviewed retrospectively. The histological discrepancy between the biopsy and resected specimens was calculated and its association with clinicopathological parameters was analyzed. Results A total of 137 lesions from 129 patients were included. The discrepancy rate between forceps biopsy and resected specimens was 45.3% (62/137). Histological discrepancy was associated with the histological category of the biopsy (p < 0.001). In addition, 17 of the 30 (56.7%) biopsies that was diagnosed as indefinite/negative for neoplasia or low-grade intraepithelial neoplasia were upgraded to HGIN or ESCC after resection. The upgrade was due to lesion size ≥ 10 mm (p = 0.002) and type B intrapapillary capillary loops (p < 0.001). Moreover, 34 of the 83 biopsies that were diagnosed with HGIN were upgraded to ESCC after resection, which was related to lesion size (p = 0.001), location (p = 0.018), and pink color sign (p = 0.002). Conclusions Histological discrepancy between forceps biopsy and resected specimens is common in clinical practice. Recognizing the risk factors for each histological category of biopsy may reduce these discrepancies and improve clinical management.

Discrepancy rates of preliminary and final reports for after-hours pediatric teleradiology interpretations

Background Children’s hospitals often do not have a high enough volume to justify providing radiologist staffing overnight, leading to hospitals employing teleradiology services to offer preliminary reports. There is limited literature related to discrepancies between preliminary teleradiology pediatric radiologists and final interpretations. Purpose The purpose of this study is to examine discrepancy rates for teleradiologists preliminarily interpreting pediatric exams at a children’s hospital. Material and Methods Eight thousand seven hundred seventy-eight consecutive preliminary reports issued by pediatric teleradiologists were reviewed. The hospital utilized a system in which local onsite radiologists rated the preliminary reports of teleradiologists following the interpretations as part of standard operating procedure. Discrepancies were also rated according to whether the discrepancy was actionable (judged to alter patient management by the final rater) or not. Rates were stratified by modality, preliminary teleradiologist reader, and final rater and compared to each using a normal approximation. The mean discrepancy rates were compared using a z test for proportions. Linear regression was applied to the effect of years of radiologist experience on the total and actionable discrepancy rates. Results The overall actionable discrepancy rate was 1.6%, similar to inter-observer discrepancy rates reported in other studies. There were no significant differences in the actionable discrepancy rates among teleradiologists. There was no correlation between years of experience and discrepancy rate for either the teleradiologists or the final raters. Conclusion Pediatric subspecialty teleradiologists issue reports that mirror discrepancy rates typical of radiologists who issue reports for emergent adult studies. Years of radiologist experience is not a predictor of discrepancy rate.

Interim evaluation of Google AI forecasting for COVID-19 compared with statistical forecasting by human intelligence in the first week

BackgroundSince June, Google (Alphabet Inc.) has provided forecasting for COVID-19 outbreak by artificial intelligence (AI) in the USA. In Japan, they provided similar services from November, 2020.ObjectWe compared Google AI forecasting with a statistical model by human intelligence.MethodWe regressed the number of patients whose onset date was day t on the number of patients whose past onset date was 14 days prior, with information about traditional surveillance data for common pediatric infectious diseases including influenza, and prescription surveillance 7 days prior. We predicted the number of onset patients for 7 days, prospectively. Finally, we compared the result with Googles AI-produced forecast. We used the discrepancy rate to evaluate the precision of prediction: the sum of absolute differences between data and prediction divided by the aggregate of data.ResultsWe found Google prediction significantly negative correlated with the actual observed data, but our model slightly correlated but not significant. Moreover, discrepancy rate of Google prediction was 27.7% for the first week. The discrepancy rate of our model was only 3.47%.Discussion and ConclusionResults show Googles prediction has negatively correlated and greater difference with the data than our results. Nevertheless, it is noteworthy that this result is tentative: the epidemic curve showing newly onset patients was not fixed.