Retrospective validation of the postnatal Growth and Retinopathy of Prematurity (G-ROP) criteria in a Swiss cohort

Background Currently used screening criteria for retinopathy of prematurity (ROP) show high sensitivity for predicting treatment-requiring ROP but low specificity; over 90% of examined infants do not develop ROP that requires treatment (type 1 ROP). A novel weight gain-based prediction model was developed by the G-ROP study group to increase the specificity of the screening criteria and keep the number of ophthalmic examinations as low as possible. This retrospective cohort study aimed to externally validate the G-ROP screening criteria in a Swiss cohort. Methods Data from 645 preterm infants in ROP screening at Inselspital Bern between January 2015 and December 2019 were retrospectively retrieved from the screening log and analysed. The G-ROP screening criteria, consisting of 6 trigger parameters, were applied in infants with complete data. To determine the performance of the G-ROP prediction model for treatment-requiring ROP, sensitivity and specificity were calculated. Results Complete data were available for 322 infants who were included in the analysis. None of the excluded infants had developed type 1 ROP. By applying the 6 criteria in the G-ROP model, 214 infants were flagged to undergo screening: among these, 14 developed type 1 ROP, 9 developed type 2 ROP, and 43 developed milder stages of ROP. The sensitivity for predicting treatment-requiring ROP was 100% (CI, 0.79–1.00), and the specificity was 41% (CI, 0.35 –0.47). Implementing the novel G-ROP screening criteria would reduce the number of infants entering ROP screening by approximately one third. Conclusions The overall prevalence of treatment-requiring ROP was low (2.15%). Previously published performance parameters for the G-ROP algorithm were reproducible in this Swiss cohort. Importantly, all treatment-requiring infants were correctly identified. By using these novel criteria, the burden of screening examinations could be significantly reduced.

The Politics of Pain in Immigration Detention

In this paper I draw on qualitative material from the first complete data set of the ‘ Measure of the Quality of Life in Detention’ (MQLD) survey in the UK to reflect on its implication for understanding and challenging these sites. While similarities between immigration detention centres and prisons make it tempting to place the testimonies from people in detention within the framework of the ‘pains of imprisonment’, I propose an alternative reading of these first-hand accounts. Rather than approaching them as sociological statements of suffering, caused by the loss of liberty, I interpret them as political statements which, in turn, demand a political response. Immigration removal centres (IRCs), these people assert, are fundamentally at odds with key values of a liberal democracy. Those detained within them are not considered to be equal members of a shared community of value; rather, their incarceration marks them out symbolically and, quite practically, as outsiders to these ideas. The pain people describe illuminates the need for a new politics of detention.

Integrity protection method for trusted data of IOT nodes based on transfer learning

In order to overcome the problems of high data storage occupancy and long encryption time in traditional integrity protection methods for trusted data of IOT node, this paper proposes an integrity protection method for trusted data of IOT node based on transfer learning. Through the transfer learning algorithm, the data characteristics of the IOT node is obtained, the feature mapping function in the common characteristics of the node data is set to complete the classification of the complete data and incomplete data in the IOT nodes. The data of the IOT nodes is input into the data processing database to verify its security, eliminate the node data with low security, and integrate the security data and the complete data. On this basis, homomorphic encryption algorithm is used to encrypt the trusted data of IOT nodes, and embedded processor is added to the IOT to realize data integrity protection. The experimental results show that: after using the proposed method to protect the integrity of trusted data of IOT nodes, the data storage occupancy rate is only about 3.5%, the shortest time-consuming of trusted data encryption of IOT nodes is about 3 s, and the work efficiency is high.

Supplemental Material: Differences between soil and air temperatures: Implications for geological reconstructions of past climate

The complete data used in text Figures 7–11.

Supplemental Material: Differences between soil and air temperatures: Implications for geological reconstructions of past climate

The complete data used in text Figures 7–11.

Self–Training With Quantile Errors for Multivariate Missing Data Imputation for Regression Problems in Electronic Medical Records: Algorithm Development Study

Background When using machine learning in the real world, the missing value problem is the first problem encountered. Methods to impute this missing value include statistical methods such as mean, expectation-maximization, and multiple imputations by chained equations (MICE) as well as machine learning methods such as multilayer perceptron, k-nearest neighbor, and decision tree. Objective The objective of this study was to impute numeric medical data such as physical data and laboratory data. We aimed to effectively impute data using a progressive method called self-training in the medical field where training data are scarce. Methods In this paper, we propose a self-training method that gradually increases the available data. Models trained with complete data predict the missing values in incomplete data. Among the incomplete data, the data in which the missing value is validly predicted are incorporated into the complete data. Using the predicted value as the actual value is called pseudolabeling. This process is repeated until the condition is satisfied. The most important part of this process is how to evaluate the accuracy of pseudolabels. They can be evaluated by observing the effect of the pseudolabeled data on the performance of the model. Results In self-training using random forest (RF), mean squared error was up to 12% lower than pure RF, and the Pearson correlation coefficient was 0.1% higher. This difference was confirmed statistically. In the Friedman test performed on MICE and RF, self-training showed a P value between .003 and .02. A Wilcoxon signed-rank test performed on the mean imputation showed the lowest possible P value, 3.05e-5, in all situations. Conclusions Self-training showed significant results in comparing the predicted values and actual values, but it needs to be verified in an actual machine learning system. And self-training has the potential to improve performance according to the pseudolabel evaluation method, which will be the main subject of our future research.

Employee Performance Assessment at PG-TK Diamond Kids School Using the Analytical Hierarchy Process Method

The problem of work performance at the foundation has not been maximally measured, only a review of achievements based on an assessment that is considered not perspective because it is not based on accurate points from employees. Regarding solving complex problems based on employee assessments, foundation leaders need to be careful to collect more detailed, accurate, and complete data. Therefore, it is necessary to have a program that can analyze the work and achievements of employees at the foundation. So one of the ways to analyze the assessment is the Analytical Hierarchy Process.

The Impact of Burn Survivor Pre-injury Income and Payer Status on Health-Related Quality of Life

Introduction The costs required to provide acute care for patients with serious burn injuries are significant. In the US, these costs are often shared by patients. However, the impacts of pre-injury finances on health-related quality of life (HRQL) have been poorly characterized. We hypothesized that lower income and public payers would be associated with poorer HRQL. Methods Burn survivors with complete data for pre-injury personal income and payer status were extracted from the longitudinal Burn Model System National Database. HRQL outcomes were measured with VR-12 scores at 6, 12, and 24 months post-injury. VR-12 scores were evaluated using generalized linear models, adjusting for potential confounders (e.g., age, gender, self-identified race, burn injury severity). Results 453 participants had complete data for income and payer status. More than one third of BMS participants earned less than $25,000/year (36%), 24% earned $25,000-49,000/year, 23% earned $50,000-99,000/year, 11% earned $100,000-149,000/year, 3% earned $150,000-199,000/year, and 4% earned >$200,000/year. VR-12 mental component (MCS) and physical component summary (PCS) scores were highest for those who earned $150-199k/year (55.8 and 55.8), and lowest for those who earned <$25,000/year (49.0 and 46.4). After adjusting for demographics, payer, and burn severity, 12-month MCS and PCS and 24-month PCS scores were negatively associated with Medicare payer (p<0.05). Low income was not significantly associated with lower VR-12 scores. Conclusion There was a peaking relationship between HRQL and middle-class income, but this trend was not significant after adjusting for covariates. Public payers, particularly Medicare, were independently associated with poorer HRQL. The findings might be used to identify those at risk of financial toxicity for targeting assistance during rehabilitation.