Validity of Administrative Coding for Nonarteritic Ischemic Optic Neuropathy.

BACKGROUND: Administrative claims have been used to study the incidence and outcomes of nonarteritic ischemic optic neuropathy (NAION), but the validity of International Classification of Diseases (ICD)-10 codes for identifying NAION has not been examined. METHODS: We identified patients at 3 academic centers who received ≥1 ICD-10 code for NAION in 2018. We abstracted the final diagnosis from clinical documentation and recorded the number of visits with an NAION diagnosis code. We calculated positive predictive value (PPV) for the overall sample and stratified by subspecialty and the number of diagnosis codes. For patients with ophthalmology or neuro-ophthalmology visit data, we recorded presenting symptoms, examination findings, and laboratory data and calculated PPV relative to case definitions of NAION that incorporated sudden onset of symptoms, optic disc edema, afferent pupillary defect, and other characteristics. RESULTS: Among 161 patients, PPV for ≥1 ICD-10 code was 74.5% (95% CI: 67.2%-80.7%). PPV was similar when restricted to patients who had visited an ophthalmologist (75.8%, 95% CI: 68.4%-82.0%) but increased to 86.8% when restricted to those who had visited neuro-ophthalmologists (95% CI: 79.2%-91.9%). Of 113 patients with >1 ICD-10 code and complete examination data, 37 (32.7%) had documented sudden onset, optic disc swelling, and an afferent pupillary defect (95% CI: 24.7%-42.0%). Of the 76 patients who did not meet these criteria, 54 (71.0%) still received a final clinical diagnosis of NAION; for most (41/54, 75.9%), this discrepancy was due to lack of documented optic disc edema. CONCLUSIONS: The validity of ICD-10 codes for NAION in administrative claims data is high, particularly when combined with provider specialty.

The partial derivative framework for substantive regression effects.

egression models are ubiquitous in the psychological sciences. The standard practice in reporting and interpreting regression models are to present and interpret coefficient estimates and the associated standard errors, confidence intervals and p-values. However, coefficient estimates have limited inferential utility if the outcome is modeled nonlinearly with respect to the substantively interpreted predictors. This is problematic in common modeling strategies, such as nonlinear predictor designs and/or generalized linear models. In the former, coefficients may correspond to product, power, log, and/or exponentially transformed units. In the latter, the relationship between the predictors and outcome are modeled via a function of the outcome, rather than the outcome in its original units. In both cases, the interpretation of the coefficients alone do not provide straightforward summaries of the data, and in fact may be misleading. We address these issues by developing a framework of regression effects by integrating two critical features. First, we explicitly model substantive variables in the units that provide the desired interpretation. Second, we use partial derivatives to summarize the relations between the substantive predictors and outcome variables to account for nonlinearities arising from modeling strategies. We show how to derive estimates and standard errors for quantities of interest in the interpretive units, as well as techniques to present the relationships between variables in meaningful ways. Finally, we provide demonstrations in both simulated and real data over a wide variety of models and estimation procedures. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Interpreting Interaction Effects in Generalized Linear Models of Nonlinear Probabilities and Counts.

Psychology research frequently involves the study of probabilities and counts. These are typically analyzed using generalized linear models (GLMs), which can produce these quantities via nonlinear transformation of model parameters. Interactions are central within many research applications of these models. To date, typical practice in evaluating interactions for probabilities or counts extends directly from linear approaches, in which evidence of an interaction effect is supported by using the product term coefficient between variables of interest. However, unlike linear models, interaction effects in GLMs describing probabilities and counts are not equal to product terms between predictor variables. Instead, interactions may be functions of the predictors of a model, requiring nontraditional approaches for interpreting these effects accurately. Here, we define interactions as change in a marginal effect of one variable as a function of change in another variable, and describe the use of partial derivatives and discrete differences for quantifying these effects. Using guidelines and simulated examples, we then use these approaches to describe how interaction effects should be estimated and interpreted for GLMs on probability and count scales. We conclude with an example using the Adolescent Brain Cognitive Development Study demonstrating how to correctly evaluate interaction effects in a logistic model.

Making sense of some odd ratios: A tutorial and improvements to present practices in reporting and visualizing quantities of interest for binary and count outcome models.

OBJECTIVE: Generalized linear models (GLMs) such as logistic and Poisson regression are among the most common statistical methods for modeling binary and count outcomes. Though single-coefficient tests (odds ratios, incidence rate ratios) are the most common way to test predictor-outcome relations in these models, they provide limited information on the magnitude and nature of relations with outcomes. We assert that this is largely because they do not describe direct relations with quantities of interest (QoIs) such as probabilities and counts. Shifting focus to QoIs makes several critical nuances of GLMs more apparent. METHOD: To bolster interpretability of these models, we provide a tutorial on logistic and Poisson regression and suggestions for enhancements to current reporting practices for predictor-outcome relations in GLMs. RESULTS: We first highlight differences in interpretation between traditional linear models and GLMs, and describe common misconceptions about GLMs. In particular, we highlight that link functions (a) introduce nonconstant relations between predictors and outcomes and (b) make predictor-QoI relations dependent on levels of other covariates. Each of these properties causes interpretation of GLM coefficients to diverge from interpretations of linear models. Next, we argue for a more central focus on QoIs (probabilities and counts). Finally, we propose and provide graphics and tables, with sample R code, for enhancing presentation and interpretation of QoIs. CONCLUSIONS: By improving present practices in the reporting of predictor-outcome relations in GLMs, we hope to maximize the amount of actionable information generated by statistical analyses and provide a tool for building a cumulative science of substance use disorders. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Compound Heterozygous Mutations in ZNF408 in a Patient with a Late Onset Pigmentary Retinopathy and Relatively Preserved Central Retina.

PURPOSE: To describe in detail the phenotype of a patient with compound heterozygous mutations in ZNF408 and an adult-onset pigmentary retinopathy rather than familial exudative vitreoretinopathy as expected with heterozygous mutations in this gene. METHODS: A 70-year-old male presented with a pigmentary retinopathy, which prompted a genetic evaluation that revealed two variants in trans in the ZNF408 gene. He underwent an ophthalmic examination, kinetic fields, electroretinography (ERG), spectral-domain optical coherence tomography (SD-OCT), fundus autofluorescence, wide-angle fluorescein angiography and near-infrared imaging. RESULTS: Visual acuity was 20/20 for both eyes. Fundus examination showed epiretinal membrane, vascular attenuation and peripheral bone spicule pigmentation in both eyes. Fluorescein angiography showed no vascular anomalies in both eyes. Fundus autofluorescence showed a preserved island of fundus autofluorescence centrally. Visual field by kinetic perimetry (V-4e stimulus) showed generalized constriction to 40 degrees of eccentricity and by an I-4e target showed generalized constriction to 10 degrees of eccentricity. ERG showed detectable but reduced cone-mediated responses. SD-OCT demonstrated preserved outer nuclear layer thickness centrally, which decreased with eccentricity. Static perimetry showed substantial rod and cone sensitivities centrally that declined with eccentricity. A next-generation sequencing panel revealed bi-allelic variants (p.Arg567Ter; c.1699C > T and p.Leu566His; c.1697 T > A) in the ZNF408 gene. CONCLUSIONS: ZNF408-associated retinal dystrophies can present with predominantly retinal findings and should be considered in the differential diagnosis of retinitis pigmentosa. Our study revealed a novel variant p.L566H, which to our knowledge has not previously been reported.

Detecting random responders with infrequency scales using an error-balancing threshold.

Infrequency scales are becoming a popular mode of data screening, due to their availability and ease of implementation. Recent research has indicated that the interpretation and functioning of infrequency items may not be as straightforward as had previously been thought (Curran & Hauser, 2015), yet there are no empirically based guidelines for implementing cutoffs using these items. In the present study, we compared two methods of detecting random responding with infrequency items: a zero-tolerance threshold versus a threshold that balances classification error rates. The results showed that a traditional zero-tolerance approach, on average, screens data that are less indicative of careless responding than those screened by the error-balancing approach. Thus, the de facto standard of applying a "zero-tolerance" approach when screening participants with infrequency scales may be too stringent, so that meaningful responses may also be removed from analyses. Recommendations and future directions are discussed.

Improving Present Practices in the Visual Display of Interactions.

Interaction plots are used frequently in psychology research to make inferences about moderation hypotheses. A common method of analyzing and displaying interactions is to create simple-slopes or marginal-effects plots using standard software programs. However, these plots omit features that are essential to both graphic integrity and statistical inference. For example, they often do not display all quantities of interest, omit information about uncertainty, or do not show the observed data underlying an interaction, and failure to include these features undermines the strength of the inferences that may be drawn from such displays. Here, we review the strengths and limitations of present practices in analyzing and visualizing interaction effects in psychology. We provide simulated examples of the conditions under which visual displays may lead to inappropriate inferences and introduce open-source software that provides optimized utilities for analyzing and visualizing interactions.

Random responses inflate statistical estimates in heavily skewed addictions data.

BACKGROUND: Some respondents may respond at random to self-report surveys, rather than responding conscientiously (Meade and Craig, 2012), and this has only recently come to the attention of researchers in the addictions field (Godinho et al., 2016). Almost no research in the published addictions literature has reported screening for random responses. We illustrate how random responses can bias statistical estimates using simulated and real data, and how this is especially problematic in skewed data, as is common with substance use outcomes. METHOD: We first tested the effects of varying amounts and types of random responses on covariance-based statistical estimates in distributions with varying amounts of skew. We replicated these findings in correlations from a real dataset (Add Health) by replacing varying amounts of real data with simulated random responses. RESULTS: Skew and the proportion of random responses influenced the amount and direction of bias. When the data were not skewed, uniformly random responses deflated estimates, while long-string random responses inflated estimates. As the distributions became more skewed, all types of random responses began to inflate estimates, even at very small proportions. We observed similar effects in the Add Health data. CONCLUSIONS: Failing to screen for random responses in survey data produces biased statistical estimates, and data with only 2.5% random responses can inflate covariance-based estimates (i.e., correlations, Cronbach's alpha, regression coefficients, factor loadings, etc.) when data are heavily skewed. Screening for random responses can substantially improve data quality, reliability and validity.

Is the Bifactor Model a Better Model or Is It Just Better at Modeling Implausible Responses? Application of Iteratively Reweighted Least Squares to the Rosenberg Self-Esteem Scale.

Although the structure of the Rosenberg Self-Esteem Scale (RSES) has been exhaustively evaluated, questions regarding dimensionality and direction of wording effects continue to be debated. To shed new light on these issues, we ask (a) for what percentage of individuals is a unidimensional model adequate, (b) what additional percentage of individuals can be modeled with multidimensional specifications, and (c) what percentage of individuals respond so inconsistently that they cannot be well modeled? To estimate these percentages, we applied iteratively reweighted least squares (IRLS) to examine the structure of the RSES in a large, publicly available data set. A distance measure, ds, reflecting a distance between a response pattern and an estimated model, was used for case weighting. We found that a bifactor model provided the best overall model fit, with one general factor and two wording-related group factors. However, on the basis of dr values, a distance measure based on individual residuals, we concluded that approximately 86% of cases were adequately modeled through a unidimensional structure, and only an additional 3% required a bifactor model. Roughly 11% of cases were judged as "unmodelable" due to their significant residuals in all models considered. Finally, analysis of ds revealed that some, but not all, of the superior fit of the bifactor model is owed to that model's ability to better accommodate implausible and possibly invalid response patterns, and not necessarily because it better accounts for the effects of direction of wording.