A Regression Discontinuity Design Framework for Controlling Selection Bias in Evaluations of Differential Item Functioning

Differential item functioning (DIF) is often used to examine validity evidence of alternate form test accommodations. Unfortunately, traditional approaches for evaluating DIF are prone to selection bias. This article proposes a novel DIF framework that capitalizes on regression discontinuity design analysis to control for selection bias. A simulation study was performed to compare the new framework with traditional logistic regression, with respect to Type I error and power rates of the uniform DIF test statistics and bias and root mean square error of the corresponding effect size estimators. The new framework better controlled the Type I error rate and demonstrated minimal bias but suffered from low power and lack of precision. Implications for practice are discussed.

Implementing Serosurveys in India: Experiences, Lessons Learned, and Recommendations

Serological surveillance for vaccine-preventable diseases, such as measles and rubella, can provide direct measures of population immunity across age groups, identify gaps in immunity, and document changes in immunity over time. Rigorously conducted, representative household serosurveys provide high-quality estimates with minimal bias. However, they can be logistically challenging, expensive, and have higher refusal rates than vaccine coverage surveys. This article shares lessons learned through implementing nine measles and rubella household serosurveys in five districts in India—the challenges faced, the potential impact on results, and recommendations to facilitate the conduct of serosurveys. Specific lessons learned arose from challenges related to community mobilization owing to lack of cooperation in certain settings and populations, limitations of outdated census information, nonresponse due to refusal or unavailability during survey enumeration and enrollment, data collection issues, and specimen collection and handling issues. Although some experiences are specific to serosurveys in India, these lessons are generalizable to other household surveys, particularly vaccination coverage and serosurveys conducted in low- and middle-income settings.

Automated bone mineral density prediction and fracture risk assessment using plain radiographs via deep learning

Dual-energy X-ray absorptiometry (DXA) is underutilized to measure bone mineral density (BMD) and evaluate fracture risk. We present an automated tool to identify fractures, predict BMD, and evaluate fracture risk using plain radiographs. The tool performance is evaluated on 5164 and 18175 patients with pelvis/lumbar spine radiographs and Hologic DXA. The model is well calibrated with minimal bias in the hip (slope = 0.982, calibration-in-the-large = −0.003) and the lumbar spine BMD (slope = 0.978, calibration-in-the-large = 0.003). The area under the precision-recall curve and accuracy are 0.89 and 91.7% for hip osteoporosis, 0.89 and 86.2% for spine osteoporosis, 0.83 and 95.0% for high 10-year major fracture risk, and 0.96 and 90.0% for high hip fracture risk. The tool classifies 5206 (84.8%) patients with 95% positive or negative predictive value for osteoporosis, compared to 3008 DXA conducted at the same study period. This automated tool may help identify high-risk patients for osteoporosis.

APPetite: Validation of a smartphone app-based tool for the remote measure of free-living subjective appetite

This study determined the validity, reproducibility and usability of a smartphone app – APPetite – for the measure of free-living, subjective appetite. Validity was assessed compared with the criterion tool of pen-and-paper visual analogue scale (VAS) (n=22). Appetite was recorded using APPetite and VAS, one immediately after the other, upon waking and every hour thereafter for twelve hours. This was repeated the next day with the order of tool reversed. Agreement between tools was assessed using Bland-Altman analysis. Reproducibility and usability were assessed in a separate experiment (n=22) of two trials (APPetite vs. VAS), separated by seven days. Appetite was recorded in duplicate upon waking and every hour for twelve hours using APPetite or VAS. Agreement between duplicate measures was assessed using Bland-Altman analysis and coefficient of variation (CV) was compared between tools. Usability was assessed by comparing compliance and by qualitative evaluation. APPetite demonstrated good criterion validity with trivial bias of 1.65 units/mm·hr−1 between APPetite- and VAS-derived AUC appetite scores. Limits of agreement were within a maximum allowed difference of 10%. However, proportional bias was observed. APPetite demonstrated high reproducibility, with minimal bias (-0.578 units·hr−1) and no difference in CV between APPetite and VAS (1.29±1.42% vs 1.54±2.36%, p = 0.64). Compliance was high with APPetite (92.7±8.0%) and VAS (91.6±20.4%, p = 0.81). Ninety percent of participants preferred APPetite, citing greater accessibility, simplified process and easier/quicker use. While proportional bias precludes using APPetite and VAS interchangeably, APPetite appears a valid, reproducible and highly usable tool for measuring free-living appetite in young-to-middle-aged adults.

NetSolP: predicting protein solubility in E. coli using language models

Solubility and expression levels of proteins can be a limiting factor for large-scale studies and industrial production. By determining the solubility and expression directly from the protein sequence, the success rate of wet-lab experiments can be increased. In this study, we focus on predicting the solubility and usability for purification of proteins expressed in Escherichia coli directly from the sequence. Our model NetSolP is based on deep-learning protein language models called transformers and we show that it achieves state-of-the-art performance and improves extrapolation across datasets. As we find current methods are built on biased datasets, we curate existing datasets by using strict sequence-identity partitioning and ensure that there is minimal bias in the sequences.

Validation of Permafrost Active Layer Estimates from Airborne SAR Observations

In permafrost regions, active layer thickness (ALT) observations measure the effects of climate change and predict hydrologic and elemental cycling. Often, ALT is measured through direct ground-based measurements. Recently, synthetic aperture radar (SAR) measurements from airborne platforms have emerged as a method for observing seasonal thaw subsidence, soil moisture, and ALT in permafrost regions. This study validates airborne SAR-derived ALT estimates in three regions of Alaska, USA using calibrated ground penetrating radar (GPR) geophysical data. The remotely sensed ALT estimates matched the field observations within uncertainty for 79% of locations. The average uncertainty for the GPR-derived ALT validation dataset was 0.14 m while the average uncertainty for the SAR-derived ALT in pixels coincident with GPR data was 0.19 m. In the region near Utqiaġvik, the remotely sensed ALT appeared slightly larger than field observations while in the Yukon-Kuskokwim Delta region, the remotely sensed ALT appeared slightly smaller than field observations. In the northern foothills of the Brooks Range, near Toolik Lake, there was minimal bias between the field data and remotely sensed estimates. These findings suggest that airborne SAR-derived ALT estimates compare well with in situ probing and GPR, making SAR an effective tool to monitor permafrost measurements.

Chasing the Anchor: A Systematic Review and Meta-Analysis of Perceptual Anchoring Deficits in Developmental Dyslexia

Purpose We report the results of a systematic review and meta-analysis investigating the relationship between perceptual anchoring and dyslexia. Our goal was to assess the direction and degree of the effect between perceptual anchoring and reading ability in typical and atypical (i.e., dyslexic) readers. Method We performed a literature search of experiments explicitly assessing perceptual anchoring and reading ability using PsycInfo (Ovid, 1860–2020), MEDLINE (Ovid, 1860–2019), EMBASE (Ovid, 1883–2019), and PubMed for all available years up to June (2020). Our eligibility criteria consisted of English language articles, and, at minimum, one experimental group identified as dyslexic—either by reading assessment at the time or by previous diagnosis. We assessed for risk of bias using an adapted version of the Newcastle–Ottawa Scale. Eight studies were included in this review and meta-analysis ( n = 422 participants). Results The overall effect was negative, moderate, and statistically significant; g = −0.70, 95% confidence interval [−1.10, −0.29]: a negative effect size indicating less perceptual anchoring in dyslexic versus nondyslexic groups. Visual assessment of funnel plot and Egger's test suggest minimal bias but with significant heterogeneity; Q (7) = 17.03, prediction interval [−1.79, 0.40]. Conclusions Of the included studies, we find evidence for a moderate perceptual anchoring deficit in individuals with dyslexia. The primary limitation of the current review is the small number of included studies. The variability of effect sizes appears consistent with the inherent variability within subtypes of dyslexia.

Identifying and alleviating bias due to differential depletion of susceptible people in post-marketing evaluations of COVID-19 vaccines

Recent studies have provided key information about SARS-CoV-2 vaccines' efficacy and effectiveness (VE). One important question that remains is whether the protection conferred by vaccines wanes over time. However, estimates over time are subject to bias from differential depletion of susceptibles between vaccinated and unvaccinated groups. Here we examine the extent to which biases occur under different scenarios and assess whether serologic testing has the potential to correct this bias. By identifying non-vaccine antibodies, these tests could identify individuals with prior infection. We find that the main determinant of bias is the proportion of the population that has been infected since vaccination began, which is influenced by several factors. In scenarios with high baseline VE, differential depletion of susceptibles creates minimal bias in VE estimates. However, if VE is lower, the bias for leaky vaccines (that reduce individual probability of infection given contact) is larger and should be corrected if the mechanism is known to be leaky. Conducting analyses both unadjusted and adjusted for past infection could give lower and upper bounds for the true VE. Studies of VE should therefore enroll individuals regardless of prior infection history but also collect information, ideally through serologic testing, on this critical variable.

A comparison of liver fat fraction measurement on MRI at 3T and 1.5T

Purpose Volumetric liver fat fraction (VLFF) measurements were made using the HepaFat-Scan® technique at 1.5T and 3T to determine the agreement between the measurements obtained at the two fields. Methods Sixty patients with type 2 diabetes (67% male, mean age 50.92 ± 6.56yrs) and thirty healthy volunteers (50% male, mean age 48.63 ± 6.32yrs) were scanned on 1.5T Aera and 3T Skyra (Siemens, Erlangen, Germany) MRI scanners on the same day using the HepaFat-Scan® gradient echo protocol with modification of echo times for 3T (TEs 2.38, 4.76, 7.14 ms at 1.5T and 1.2, 2.4, 3.6 ms at 3T). The 3T analyses were performed independently of the 1.5T analyses by a different analyst, blinded from the 1.5T results. Data were analysed for agreement and bias using Bland-Altman methods and intraclass correlation coefficients (ICC). A second cohort of 17 participants underwent interstudy repeatability assessment of VLFF measured by HepaFat-Scan® at 3T. Results A small, but statistically significant mean bias of 0.48% was observed between 3T and 1.5T with 95% limits of agreement -2.2% to 3.2% VLFF. The ICC for agreement between field strengths was 0.983 (95% CI 0.972–0.989). In the repeatability cohort studied at 3T the repeatability coefficient was 4.2%. The ICC for agreement was 0.971 (95% CI 0.921–0.989). Conclusion There is minimal bias and excellent agreement between the measures of VLFF using the HepaFat-Scan® at 1.5 and 3T. The test retest repeatability coefficient at 3T is comparable to the 95% limits of agreement between 1.5T and 3T suggesting that measurements can be made interchangeably between field strengths.