Measurement Bias Research Articles

Forced Measurement of Astronomical Sources at Low Signal-to-noise

We propose a modified moment matching algorithm to avoid catastrophic failures for sources with a low signal to noise ratio. The proposed modifications include a method to eliminate nonphysical negative pixel values and a forced single iteration with an initial guess derived from coadd measurements when iterative methods are unstable. We correct for all biases in measurements introduced by the method. We find that the proposed modifications allow the algorithm to avoid catastrophic failures in nearly 100% of the cases, especially at low SNR. Additionally, with a reasonable guess from coadd measurements, the algorithm measures the flux, centroid, size, shape, and ellipticity with a bias statistically consistent with zero. We show that the proposed method allows us to measure sources 7 times fainter than traditional methods when applied to images obtained from WIYN-ODI. We also present a scheme to find uncertainties in measurements when using the new method to measure astronomical sources.

Data-Driven Diagnosis Of Measurement Biases For Filtered Rayleigh Scattering

This work focuses on the data mining of a filtered Rayleigh scattering (FRS) database to diagnose the source(s) of measurement biases experienced in these applied measurements in harsh environments. Qualitative statistical analysis is performed to assess modeled-to-measured signal disagreements. This qualitative analysis shows that the signal model converges to measured signals as the non-dimensional thermodynamic parameter, termed the Y-parameter for Rayleigh scattering analysis, decreases. To perform a quantitative assessment, these FRS datasets were processed for Y-parameter and compared to known values derived from reference measurements of the corresponding flow fields. Residuals between the FRS results and the reference values were plotted as a histogram, which showed that there is a region of "least-bias" corresponding to Y-parameter less than 1. Root cause assessment of bias sources suggests that the key contributing factor is uncertainty in the modeled Rayleigh-Brillouin scattering signal for air when Y-parameter is greater than 1.

A New Criterion for Determining a Cutoff Value Based on the Biases of Incidence Proportions in the Presence of Non-differential Outcome Misclassifications.

When conducting database studies, researchers sometimes use an algorithm known as "case definition," "outcome definition," or "computable phenotype" to identify the outcome of interest. Generally, algorithms are created by combining multiple variables and codes, and we need to select the most appropriate one to apply to the database study. Validation studies compare algorithms with the gold standard and calculate indicators such as sensitivity and specificity to assess their validities. As the indicators are calculated for each algorithm, selecting an algorithm is equivalent to choosing a pair of sensitivity and specificity. Therefore, receiver operating characteristic curves can be utilized, and two intuitive criteria are commonly used. However, neither was conceived to reduce the biases of effect measures (e.g., risk difference and risk ratio), which are important in database studies. In this study, we evaluated two existing criteria from perspectives of the biases and found that one of them, called the Youden index always minimizes the bias of the risk difference regardless of the true incidence proportions under nondifferential outcome misclassifications. However, both criteria may lead to inaccurate estimates of absolute risks, and such property is undesirable in decision-making. Therefore, we propose a new criterion based on minimizing the sum of the squared biases of absolute risks to estimate them more accurately. Subsequently, we apply all criteria to the data from the actual validation study on postsurgical infections and present the results of a sensitivity analysis to examine the robustness of the assumption our proposed criterion requires.

P-477 Associations between maternal early pregnancy depression and longitudinal fetal growth

Abstract Study question Does maternal depression during early pregnancy affect fetal growth? Summary answer Maternal depression during early pregnancy is associated with increased fetal biometrics, higher birth weight, and also an elevated risk of macrosomia. What is known already Prenatal depression, a prevalent psychological condition occurring during pregnancy, affects approximately 7% to 25% of pregnant women. According to the developmental origins of health and disease theory, intrauterine adverse exposure, such as malnutrition and prenatal depression, may exert long-term effects by altering placental function and epigenetic modification. Growing evidence from epidemiologic studies has shown that maternal depression during mid-to-late stages of pregnancy may be associated with reduced fetal growth. However, the effect of depression may vary depending on the timing of intrauterine exposure. For early pregnancy depression, there is a limited number of studies examining its association with fetal growth. Study design, size, duration Prospective evaluation of the associations of maternal depression during early pregnancy with fetal growth and neonatal outcomes in 23,465 mothers. The study was conducted at a hospital-based center in Shanghai, China. The mothers were enrolled between March 2019 to April 2022. Participants/materials, setting, methods The Patient Health Questionnaire (PHQ-9) was used to screen for prenatal depression before 14 weeks’ gestational weeks. Fetal ultrasound parameters, including abdominal circumference(AC), head circumference(HC), femur length(FL),and humerus length(HL), along with birth outcomes were obtained from medical records. Estimated fetal weight(EFW) was calculated using the Hadlock formula. Differences in fetal growth trajectory of different maternal depressive statuses during three periods (16-23, 24-31, and 32-41 gestational weeks) were compared using a multilevel model with fractional polynomials. Main results and the role of chance In this prospective cohort study, maternal depressive symptoms affected 20402(13%) pregnant women during early pregnancy. Overall, compared to women without depressive symptoms, those with depressive symptoms during early pregnancy had higher longitudinal fetal trajectories, with an estimated increase in EFW (β = 0.33; 95% CI, 0.06-0.61), AC (β = 0.16; 95% CI, 0.04-0.29), HL (β = 0.14; 95% CI, 0.05-0.23), FL (β = 0.14; 95% CI, 0.03-0.25) and HL(β = 0.16; 95% CI, 0.04-0.27). Stratified by different gestational periods, the increments in HC (β = 0.16; 95% CI, 0.06-0.26), HL (β = 0.24; 95% CI, 0.11-0.38) and EFW (β = 0.60; 95% CI, 025-0.96) of women with depressive symptoms commenced after 32 weeks of gestational age and FL started after 24 gestational weeks (β = 0.24; 95% CI, 0.06-0.42 [24-31 weeks] and β = 0.19; 95% CI, 0.07-0.31 [32-41 weeks]). Notably, increases in AC among women with depressive symptoms were observed during 16-23 gestational weeks (β = 0.21; 95% CI, 0.04-0.39). What’s more, offspring born to mothers with early pregnancy depression had a significantly higher birth weight of 14.13g (95% CI, 1.33-27.81 g) and an increased risk of macrosomia (odds ratio, 1.21; 95% CI, 1.02-1.43). Limitations, reasons for caution A self-reported scale was used to assess the mental status of pariticipants, potentially introducing measurement bias. What’s more, long-term effects of early pregnancy depression on offspring were not explored due to the limited duration of follow-up thus far. Wider implications of the findings It may be necessary to reconsider the timing of prenatal depression screening and implement preventive measures during early pregnancy to mitigate excessive fetal growth. Further studies are warranted to validate the findings and supplement more biological evidence. Trial registration number not applicable

Efficacy and safety of acupuncture for cognitive impairment in Alzheimer's disease: a systematic review and meta-analysis.

To systematically evaluate the efficacy of acupuncture in the treatment of cognitive impairment in Alzheimer's disease (AD) by meta-analysis, in order to provide evidence-based evidence for the application of acupuncture therapy in the clinical process of AD. From the establishment of the database to December 31, 2022, China Biomedical Literature Database (CBM), China National Knowledge Network (CNKI), VIP database, WanFang Database, Pubmed, Embase and Cochrane Library Database were systematically searched. To collect published randomized controlled clinical trials (RCTS) of acupuncture in the treatment of cognitive impairment in AD. The subjects in the intervention group were given acupuncture alone or combined with other treatments the same as the control group; the control group received conventional Western medicine treatment. The main outcome indicators of the study were cognitive function assessment of subjects, including: Simple Mental State Examination Scale (MMSE), Assessment of daily Living Ability Scale (ADL), Alzheimer's Disease Cognitive Function Assessment Scale (ADAS-Cog), TCM syndrome score (SDSD), Montreal Cognitive Test (MoCA), Secondary outcome indicators were the occurrence of adverse reactions. Literature screening, data extraction, and quality evaluation of the included literature were performed independently by two researchers, according to bias risk assessment tools recommended in the Cochrane manual. Data were analyzed by RevMan5.3 software. Dichotomous variables were represented by risk ratio (OR) and 95% CI, and continuity variables were represented by mean difference (MD) and 95% CI. For heterogeneity analysis, when P > 0.1 and I 2 ≤ 50%, fixed effect model was applied. When P ≤ 0.1 and I 2 > 50%, the random effects model is applied. A total of 1,172 eligible subjects were included in 18 RCTS, including 595 in the intervention group and 577 in the control group. The results of meta-analysis are as follows: acupuncture intervention group improved MMSE [MD = 1.67, 95% CI (0.94, 2.41), P < 0.00001], ADL [MD = -1.18, 95% CI (-3.09, 0.72), P = 0.22], ADAS-Cog [MD = 3.31, 95% CI (5.84, 0.78), P = 0.01], SDSD [MD = 2.40, 95% CI (3.53, 1.26), P < 0.0001], MoCA [MD = 4.80, 95% CI (3.74, 5.86), P = 0.04] were better than the control group. No serious adverse reactions related to acupuncture were observed in the intervention group, and the incidence and severity of adverse reactions were lower than those in the control group, with statistical significance [OR = 0.17, 95% CI (0.04, 0.67), P = 0.01]. Existing data show that acupuncture therapy has certain advantages in improving cognitive dysfunction and improving self-care ability of patients with Alzheimer's disease. However, due to the small number of RCTS and cases evaluating the efficacy of acupuncture, and the possibility of measurement bias and selectivity bias in included studies, it is still unable to conduct high-intensity demonstration on its effectiveness. Further large-scale, high-quality randomized, double-blind controlled trials are needed to evaluate its efficacy. https://inplasy.com/inplasy-2021-12-0125/, identifier: INPLASY2021120125.

Measurement Biases in Ocean Temperature Profiles from Marine Mammal Dataloggers

Abstract The study focuses on biases in ocean temperature profiles obtained by means of Satellite Relay Data Loggers (SRDL recorders) and time–depth recorder (TDR) attached to marine mammals. Quasi-collocated profiles from Argo floats and from ship-based conductivity–temperature–depth (CTD) profilers are used as reference. SRDL temperature biases depend on the sensor type and vary with depth. For the most numerous group of Valeport 3 (VP3) and conductivity–temperature–fluorescence (CTF) sensors, the bias is negative except for the layer 100–200 m. The vertical bias structure suggests a link to the upper-ocean thermal structure within the upper 200-m layer. Accounting for a time lag which might remain in the postprocessed data reduces the bias variability throughout the water column. Below 200-m depth, the bias remains negative with the overall mean of −0.027° ± 0.07°C. The suggested depth and thermal corrections for biases in SRDL data are within the uncertainty limits declared by the manufacturer. TDR recorders exhibit a different bias pattern, showing the predominantly positive bias of 0.08°–0.14°C below 100 m primarily due to the systematic error in pressure. Significance Statement The purpose of this work is to improve the consistency of the data from the specific instrumentation type used to measure ocean water temperature, namely, the data from miniature temperature sensors attached to marine mammals. As mammals dive during their route to and from their feeding areas, these sensors measure water temperature and dataloggers send the measured temperature data to oceanographic data centers via satellites as soon as the mammals return to the sea surface. We have shown that these data exhibit small systematic instrumental errors and suggested the respective corrections. Taking these corrections into account is important for the assessment of the ocean climate change.

The Variation in the Galaxy-wide Initial Mass Function for Low-mass Stars: Modeling and Observational Insights

The stellar initial mass function (IMF) characterizes the mass distribution of newly formed stars in various cosmic environments, serving as a fundamental assumption in astrophysical research. Recent findings challenge the prevalent notion of a universal and static IMF, proposing instead that the IMF’s shape is contingent upon the star formation environment. In this study, we analyze the galaxy-wide variation in the IMF for low-mass stars in both dwarf and massive galaxies with diverse observational methods. Despite systematic discrepancies between different approaches, an IMF model with a metallicity-dependent slope for the low-mass stars aligns with the majority of observations, indicating a high degree of uniformity in the star formation processes across the Universe. We also emphasize the need for a more comprehensive understanding of the variation in the low-mass IMF, considering measurement biases and factors beyond metallicity.

Forecast of Joint Analysis of Cosmic Shear and Supernovae Magnification from the CSST and LSST

Cosmic shear and cosmic magnification reflect the same gravitational lensing field. Each of these two probes are affected by different systematics. We study the auto- and cross-correlations of cosmic shear from the China Space Survey Telescope and cosmic magnification of supernovae from the Large Synoptic Survey Telescope. We want to determine to what extent, by adding the magnification data, we can remove the systematic bias in cosmic shear measurements. We generate mock shear/magnification maps based on the correlation between different tomographic bins. After obtaining the corrected power spectra, we adopt the Markov Chain Monte Carlo technique to fit the theoretical models and investigate the constraints on the cosmological and nuisance parameters. We find that, with only the cosmic shear data, there are 1σ biases in the σ 8 and intrinsic alignment model parameters. By adding the magnification data, we are able to remove these biases perfectly.

Pharmacological Enhancement of Dopamine Neurotransmission Does Not Affect Illusory Pattern Perception.

Psychotic symptoms and delusional beliefs have been linked to dopamine transmission in both healthy and clinical samples and are assumed to result at least in part from perceiving illusory patterns in noise. However, the existing literature on the role of dopamine in detecting patterns in noise is inconclusive. To address this issue, we assessed the effect of manipulating dopaminergic neurotransmission on illusory pattern perception in healthy individuals (n = 48, n = 19 female) in a double-blind placebo-controlled within-subjects design (see preregistration at https://osf.io/a4k9j/). We predicted individuals on versus off ʟ-DOPA to be more likely to perceive illusory patterns, specifically objects in images containing only noise. Using a signal detection model, however, we found no credible evidence that ʟ-DOPA compared with placebo increased false alarm rates. Further, ʟ-DOPA did not reliably modulate measures of accuracy, discrimination sensitivity, and response bias. In all cases, Bayesian statistics revealed strong evidence in favor of the null hypothesis. The task design followed previous work on illusory pattern perception and comprised a limited number of items per condition. The results therefore need to be interpreted with caution, as power was limited. Future studies should address illusory pattern perception using more items and take into account potential dose-dependent effects and differential effects in healthy versus clinical samples.

Effects of housing stability and contemporary mortgage lending bias on breast cancer stage at diagnosis among older women in the United States.

Interventions aimed at upstream factors contributing to late-stage diagnoses could reduce disparities and improve breast cancer outcomes. This study examines the association between measures of housing stability and contemporary mortgage lending bias on breast cancer stage at diagnosis among older women in the United States. We studied 67,588 women aged 66-90 from the SEER-Medicare linked database (2010-2015). The primary outcome was breast cancer stage at diagnosis. Multinomial regression models adjusted for individual and neighborhood socio-economic factors were performed using a three-category outcome (stage 0, early stage, and late stage). Key census tract-level independent variables were residence in the same house as the previous year, owner-occupied homes, and an index of contemporary mortgage lending bias. In models adjusted for individual factors, higher levels of mortgage lending bias were associated with later stage diagnosis (RR = 1.10, 95% CI 1.02-1.20; RR = 1.31, 95% CI 1.16-1.49; RR = 1.41, 95% CI 1.24-1.60 for least to high, respectively). In models adjusted for individual and neighborhood socio-economic factors, moderate and high levels of mortgage lending bias were associated with later stage diagnosis (RR = 1.16, 95% CI 1.02-1.33 for moderate and RR = 1.18, 95% CI 1.02-1.37 for high). Owner occupancy and tenure were not associated with later stage diagnosis in adjusted models. Contemporary mortgage lending bias demonstrated a significant gradient relationship with later stage at diagnosis of breast cancer. Policy interventions aimed at reducing place-based mortgage disinvestment and its impacts on local resources and opportunities should be considered as part of an overall strategy to decrease late-stage breast cancer diagnosis and improve prognosis.

Supplemental Material for Uncovering Measurement Bias: Differential Item Functioning by Income in the Couples Satisfaction Index

Supplemental Material for Uncovering Measurement Bias: Differential Item Functioning by Income in the Couples Satisfaction Index

Maternal Mortality and Gender Differences in Adult Mortality in Early New England

Abstract: Published estimates of adult mortality and life expectancy in early America are based on poor data, faulty methods, and small numbers of cases. Estimates for women are especially difficult to make and are often inaccurate. This study relies on the remarkably complete and reliable demographic data available for the graduates of Yale College and their wives to examine gender differences in mortality and life expectancy. Because these data are nearly complete, mortality differences can be estimated without significant measurement bias. The results indicate that women married to Yale graduates suffered higher mortality during their peak childbearing years relative to their husbands, but lower mortality afterwards. Much of the difference in mortality under the age of fifty appears to be related to maternal mortality, but gender differences in mortality from other causes such as tuberculosis also appears to be important. This study also examines perceptions of the risk of death in childbearing among white women in early America and suggests that while religion provided an interpretive framework that may have overemphasized the risk, the cumulative risk of death from maternal causes was nontrivial and rightly feared by pregnant women, who took active measures to mitigate their fears and eventually to reduce their fertility.

OA01.03 Pack-Year Smoking History: An Inadequate and Biased Measure to Determine Lung Cancer Screening Eligibility

OA01.03 Pack-Year Smoking History: An Inadequate and Biased Measure to Determine Lung Cancer Screening Eligibility

Rapid and Precise Computation of Fractional Flow Reserve from Routine Two-Dimensional Coronary Angiograms Based on Fluid Mechanics: The Pilot FFR2D Study.

Objective: To present a novel pipeline for rapid and precise computation of fractional flow reserve from an analysis of routine two-dimensional coronary angiograms based on fluid mechanics equations (FFR2D). Material and methods: This was a pilot analytical study that was designed to assess the diagnostic performance of FFR2D versus the gold standard of FFR (threshold ≤ 0.80) measured with a pressure wire for the physiological assessment of intermediate coronary artery stenoses. In a single academic center, consecutive patients referred for diagnostic coronary angiography and potential revascularization between 1 September 2020 and 1 September 2022 were screened for eligibility. Routine two-dimensional angiograms at optimal viewing angles with minimal overlap and/or foreshortening were segmented semi-automatically to derive the vascular geometry of intermediate coronary lesions, and nonlinear pressure-flow mathematical relationships were applied to compute FFR2D. Results: Some 88 consecutive patients with a single intermediate coronary artery lesion were analyzed (LAD n = 74, RCA n = 9 and LCX n = 5; percent diameter stenosis of 45.7 ± 11.0%). The computed FFR2D was on average 0.821 ± 0.048 and correlated well with invasive FFR (r = 0.68, p < 0.001). There was very good agreement between FFR2D and invasive-wire FFR with minimal measurement bias (mean difference: 0.000 ± 0.048). The overall accuracy of FFR2D for diagnosing a critical epicardial artery stenosis was 90.9% (80 cases classified correctly out of 88 in total). FFR2D identified 24 true positives, 56 true negatives, 4 false positives, and 4 false negatives and predicted FFR ≤ 0.80 with a sensitivity of 85.7%, specificity of 93.3%, positive likelihood ratio of 13.0, and negative likelihood ratio of 0.15. FFR2D had a significantly better discriminatory capacity (area under the ROC curve: 0.95 [95% CI: 0.91-0.99]) compared to 50%DS on 2D-QCA (area under the ROC curve: 0.70 [95% CI: 0.59-0.82]; p = 0.0001) in predicting wire FFR ≤ 0.80. The median time of image analysis was 2 min and the median time of computation of the FFR2D results was 0.1 s. Conclusion: FFR2D may rapidly derive a precise image-based metric of fractional flow reserve with high diagnostic accuracy based on a single two-dimensional coronary angiogram.

Reliable, rapid, and remote measurement of metacognitive bias

Metacognitive biases have been repeatedly associated with transdiagnostic psychiatric dimensions of ‘anxious-depression’ and ‘compulsivity and intrusive thought’, cross-sectionally. To progress our understanding of the underlying neurocognitive mechanisms, new methods are required to measure metacognition remotely, within individuals over time. We developed a gamified smartphone task designed to measure visuo-perceptual metacognitive (confidence) bias and investigated its psychometric properties across two studies (N = 3410 unpaid citizen scientists, N = 52 paid participants). We assessed convergent validity, split-half and test–retest reliability, and identified the minimum number of trials required to capture its clinical correlates. Convergent validity of metacognitive bias was moderate (r(50) = 0.64, p < 0.001) and it demonstrated excellent split-half reliability (r(50) = 0.91, p < 0.001). Anxious-depression was associated with decreased confidence (β = − 0.23, SE = 0.02, p < 0.001), while compulsivity and intrusive thought was associated with greater confidence (β = 0.07, SE = 0.02, p < 0.001). The associations between metacognitive biases and transdiagnostic psychiatry dimensions are evident in as few as 40 trials. Metacognitive biases in decision-making are stable within and across sessions, exhibiting very high test–retest reliability for the 100-trial (ICC = 0.86, N = 110) and 40-trial (ICC = 0.86, N = 120) versions of Meta Mind. Hybrid ‘self-report cognition’ tasks may be one way to bridge the recently discussed reliability gap in computational psychiatry.

Acquisition Efficiency and Technical Repeatability of Dual-Frequency 3D Vector MR Elastography of the Liver.

MR elastography (MRE) at 60 Hz is widely used for staging liver fibrosis. MRE with lower frequencies may provide inflammation biomarkers. To establish a practical simultaneous dual-frequency liver MRE protocol at both 30 Hz and 60 Hz during a single examination and validate the occurrence of second harmonic waves at 30 Hz. Retrospective. One hundred six patients (48 females, age: 50.0 ± 13.4 years) were divided as follows: Cohort One (15 patients with chronic liver disease [CLD] and 25 healthy volunteers) with simultaneous dual-frequency MRE. Cohort Two (66 patients with CLD) with second harmonic MRE. 3-T, single- or dual-frequency MRE at 30 Hz and 60 Hz. Liver stiffness (LS) in both cohorts was evaluated with manually placed volumetric ROIs by two independent analyzers. Image quality was assessed by three independent readers on a 4-point scale (0-3: none/failed, fair, moderate, excellent) based on the depth of wave propagation with 1/3 incremental penetration. The success rate was derived from the percentage of nonzero quality scores. Measurement agreement, bias, and repeatability of LS were assessed using intraclass correlation coefficients (ICCs), Bland-Altman plots, and repeatability coefficient (RC). Mann-Whitney U tests were used to evaluate the differences in image quality between different methods. A P-value <0.05 was considered statistically significant. Success rate was 97.5% in Cohort One and 91% success rate for the second harmonic MRE in Cohort Two. The second harmonic and conventional MRE showed excellent agreement in LS (all ICCs >0.90). The quality scores for the second harmonic wave images were lower than those from the conventional MRE (Z = -4.523). Compared with conventional and second harmonic methods, simultaneous dual-frequency had better image quality, high success rate and the advantage of intrinsic co-registration, while the second harmonic method can be an alternative if custom waveform is not available. 3 TECHNICAL EFFICACY: Stage 1.

Bias in the prenatal lung measurements in fetal congenital diaphragmatic hernia with intrauterine growth restriction.

The failure of a fetus to develop to its full potential due to maternal or placental factors is known as intrauterine growth restriction (IUGR). Fetal head growth is usually preserved in that situation producing a potential discordance between head and body size. Our goal is to discover if IUGR has an impact on the prenatal ultrasound measurements taken to assess pulmonary development in congenital diaphragmatic hernia (CDH). A retrospective chart review (IRB#2017-6361) was performed on all prenatally diagnosed CDH patients from 2007 to 2016. Patient demographics, fetal and neonatal anthropometric measurements, and fetal lung parameters were the main subjects of the data that were gathered. Fetal growth was assessed by the curves based on US data by Olsen etal. and by Peleg etal. Of 147 CDH patients, 19 (12.9 %) patients were diagnosed with IUGR before the 30th gestational week while there were 20 (13.6 %) patients after the 30th gestational week. Patients with IUGR and the observed-to-expected lung-to-head ratio (O/E LHR) less than 25 % had better survival rates both to discharge and date compared to non IUGR group (p=0.226, OR 2.25 95 % CI 0.60-1.08 and p=0.175, OR 2.40 95 % CI 0.66-1.17, respectively). Moreover, the ECMO need of the patients who had IUGR and O/E LHR less than 25 % was significantly less than the patients without IUGR (38.5 vs. 80.0 %, p=0.005). This study confirms that the intrauterine measurements to predict pulmonary hypoplasia in CDH patients are misleading in the presence of IUGR and cause an overestimation.

Assessing racial bias in healthcare predictive models: Practical lessons from an empirical evaluation of 30-day hospital readmission models

ObjectiveDespite increased availability of methodologies to identify algorithmic bias, the operationalization of bias evaluation for healthcare predictive models is still limited. Therefore, this study proposes a process for bias evaluation through an empirical assessment of common hospital readmission models. The process includes selecting bias measures, interpretation, determining disparity impact and potential mitigations. MethodsThis retrospective analysis evaluated racial bias of four common models predicting 30-day unplanned readmission (i.e., LACE Index, HOSPITAL Score, and the CMS readmission measure applied as is and retrained). The models were assessed using 2.4 million adult inpatient discharges in Maryland from 2016 to 2019. Fairness metrics that are model-agnostic, easy to compute, and interpretable were implemented and apprised to select the most appropriate bias measures. The impact of changing model’s risk thresholds on these measures was further assessed to guide the selection of optimal thresholds to control and mitigate bias. ResultsFour bias measures were selected for the predictive task: zero-one-loss difference, false negative rate (FNR) parity, false positive rate (FPR) parity, and generalized entropy index. Based on these measures, the HOSPITAL score and the retrained CMS measure demonstrated the lowest racial bias. White patients showed a higher FNR while Black patients resulted in a higher FPR and zero-one-loss. As the models’ risk threshold changed, trade-offs between models’ fairness and overall performance were observed, and the assessment showed all models’ default thresholds were reasonable for balancing accuracy and bias. ConclusionsThis study proposes an Applied Framework to Assess Fairness of Predictive Models (AFAFPM) and demonstrates the process using 30-day hospital readmission model as the example. It suggests the feasibility of applying algorithmic bias assessment to determine optimized risk thresholds so that predictive models can be used more equitably and accurately. It is evident that a combination of qualitative and quantitative methods and a multidisciplinary team are necessary to identify, understand and respond to algorithm bias in real-world healthcare settings. Users should also apply multiple bias measures to ensure a more comprehensive, tailored, and balanced view. The results of bias measures, however, must be interpreted with caution and consider the larger operational, clinical, and policy context.

Re-thinking all-cause COVID-19 hospitalizations as a surrogate measure for severe illness in observational surveillance studies

All-cause COVID-19 hospitalization ≤ 30 days of infection is a common outcome for severe illness in observational/surveillance studies. Milder COVID-19 disease and COVID-19-specific measurements calls for an evaluation of this endpoint. This was a descriptive, retrospective cohort study of adults ≥ 18 who were established in primary care at Veteran Health Administration (VHA) facilities. The outcome was hospitalization within 30 days of a laboratory-confirmed, symptomatic SARS-CoV-2 infection. Between December 15, 2021 and May 1, 2022, a simple random sample of all VA facilities, excluding Puerto Rico or Philippines, was drawn to identify these hospitalized cases and determine whether hospitalization was due to COVID-19-specific causes. A chart review was conducted to record the inpatient clinical team’s diagnosis and whether the inpatient team classified the diagnosis as COVID-19 related or not. These data were used to classify hospitalizations as either due to COVID-19-specific causes (direct manifestations of SARS-CoV-2 infection) or non-COVID-19-specific hospitalizations (incidental SARS-CoV-2 infection), A simple random sample of 9966 (12.3%) all-cause hospitalizations (95% CI: 12.1%, 12.5%) was used to select 300 representative patients. Of these, 226/300 (75.3%) were determined to be COVID-19-specific. COVID-19 pneumonia was most common (147/226, 65.0%). The highest proportion of COVID-19-specific hospitalizations occurred among unvaccinated (85.0%), followed by vaccinated but not boosted (73.7%) and boosted (59.4%) (p < 0.001). The proportion of non-COVID-19-specific hospitalizations was higher in the later period (15–30 days: 55.0%) than the early (0–15 days: 22.5%) (p = 0.003). This study supports the outcome of COVID-19-specific hospitalization instead of all-cause hospitalization in observational studies. The earlier outcome period (0–15 days) was less susceptible to potential measurement bias.

Measurement of visual function in infantile nystagmus: a systematic review

Background/aimsRecent work has called into question the ability of visual acuity (VA) to accurately represent changes in visual function in infantile nystagmus (IN). This systematic review investigated factors affecting visual...