Several methods used to examine differential item functioning (DIF) in Patient-Reported Outcomes Measurement Information System (PROMIS®) measures are presented, including effect size estimation. A summary of factors that may affect DIF detection and challenges encountered in PROMIS DIF analyses, e.g., anchor item selection, is provided. An issue in PROMIS was the potential for inadequately modeled multidimensionality to result in false DIF detection. Section 1 is a presentation of the unidimensional models used by most PROMIS investigators for DIF detection, as well as their multidimensional expansions. Section 2 is an illustration that builds on previous unidimensional analyses of depression and anxiety short-forms to examine DIF detection using a multidimensional item response theory (MIRT) model. The Item Response Theory-Log-likelihood Ratio Test (IRT-LRT) method was used for a real data illustration with gender as the grouping variable. The IRT-LRT DIF detection method is a flexible approach to handle group differences in trait distributions, known as impact in the DIF literature, and was studied with both real data and in simulations to compare the performance of the IRT-LRT method within the unidimensional IRT (UIRT) and MIRT contexts. Additionally, different effect size measures were compared for the data presented in Section 2. A finding from the real data illustration was that using the IRT-LRT method within a MIRT context resulted in more flagged items as compared to using the IRT-LRT method within a UIRT context. The simulations provided some evidence that while unidimensional and multidimensional approaches were similar in terms of Type I error rates, power for DIF detection was greater for the multidimensional approach. Effect size measures presented in Section 1 and applied in Section 2 varied in terms of estimation methods, choice of density function, methods of equating, and anchor item selection. Despite these differences, there was considerable consistency in results, especially for the items showing the largest values. Future work is needed to examine DIF detection in the context of polytomous, multidimensional data. PROMIS standards included incorporation of effect size measures in determining salient DIF. Integrated methods for examining effect size measures in the context of IRT-based DIF detection procedures are still in early stages of development.

In response to the target article by Teresi et al. (2021), we explain why the article is useful and we also present a different approach. An alternative category of differential item functioning (DIF) is presented with a corresponding way of modeling DIF, based on random person and random item effects and explanatory covariates.

The psychometric process used to establish a relationship between the scores of two (or more) instruments is generically referred to as linking. When two instruments with the same content and statistical test specifications are linked, these instruments are said to be equated. Linking and equating procedures have long been used for practical benefit in educational testing. In recent years, health outcome researchers have increasingly applied linking techniques to patient-reported outcome (PRO) data. However, these applications have some noteworthy purposes and associated methodological questions. Purposes for linking health outcomes include the harmonization of data across studies or settings (enabling increased power in hypothesis testing), the aggregation of summed score data by means of score crosswalk tables, and score conversion in clinical settings where new instruments are introduced, but an interpretable connection to historical data is needed. When two PRO instruments are linked, assumptions for equating are typically not met and the extent to which those assumptions are violated becomes a decision point around how (and whether) to proceed with linking. We demonstrate multiple linking procedures—equipercentile, unidimensional IRT calibration, and calibrated projection—with the Patient-Reported Outcomes Measurement Information System Depression bank and the Patient Health Questionnaire-9. We validate this link across two samples and simulate different instrument correlation levels to provide guidance around which linking method is preferred. Finally, we discuss some remaining issues and directions for psychometric research in linking PRO instruments.

The PROsetta Stone Project, summarized in this issue by Schalet et al. (Psychometrika 86, 2021), is a major step forward in enabling comparability between different patient-reported outcomes measures. Schalet et al. clearly describe the psychometric methods used in the PROsetta Stone project and other projects from the Patient-Reported Outcomes Measurement Information System (PROMIS): linking based on unidimensional item response theory (IRT), equipercentile linking, and calibrated projection based on multidimensional IRT. Analyses in a validation data set and simulation studies provide strong support that the linking methods are robust when basic assumptions are fulfilled. The links already established will be of great value to the field, and the methodology described by Schalet et al. will hopefully inspire the next series of linking studies. Among potential improvements that should be considered by new studies are: (1) a thorough evaluation of the content of the measures to be linked to better guide the evaluation of measurement assumptions, (2) improvements in the design of linking studies such as selection of the optimal sample to provide data in the score ranges where linking precision is most critical and using counterbalanced designs to control for order effects. Finally, it may be useful to consider how the linking algorithms are used in subsequent data analyses. Analytic strategies based on plausible values or latent regression IRT models may be preferable to the simple transformation of scores from one patient at the time.

With decades of advance research and recent developments in the drug and medical device regulatory approval process, patient-reported outcomes (PROs) are becoming increasingly important in clinical trials. While clinical trial analyses typically treat scores from PROs as observed variables, the potential to use latent variable models when analyzing patient responses in clinical trial data presents novel opportunities for both psychometrics and regulatory science. An accessible overview of analyses commonly used to analyze longitudinal trial data and statistical models familiar in both psychometrics and biometrics, such as growth models, multilevel models, and latent variable models, is provided to call attention to connections and common themes among these models that have found use across many research areas. Additionally, examples using empirical data from a randomized clinical trial provide concrete demonstrations of the implementation of these models. The increasing availability of high-quality, psychometrically rigorous assessment instruments in clinical trials, of which the Patient-Reported Outcomes Measurement Information System (PROMIS®) is a prominent example, provides rare possibilities for psychometrics to help improve the statistical tools used in regulatory science.

The reliable change index has been used to evaluate the significance of individual change in health-related quality of life. We estimate reliable change for two measures (physical function and emotional distress) in the Patient-Reported Outcomes Measurement Information System (PROMIS®) 29-item health-related quality of life measure (PROMIS-29 v2.1). Using two waves of data collected 3 months apart in a longitudinal observational study of chronic low back pain and chronic neck pain patients receiving chiropractic care, and simulations, we compare estimates of reliable change from classical test theory fixed standard errors with item response theory standard errors from the graded response model. We find that unless true change in the PROMIS physical function and emotional distress scales is substantial, classical test theory estimates of significant individual change are much more optimistic than estimates of change based on item response theory.

Item response theory (IRT) models are non-linear latent variable models for discrete measures, whereas factor analysis (FA) is a latent variable model for continuous measures. In FA, the standard error (SE) of individuals’ scores is common for all individuals. In IRT, the SE depends on the individual’s score, and the SE function is to be provided. The empirical standard deviation of the scores across discrete ranges should also be computed to inform the extent to which IRT SEs overestimate or underestimate the variability of the scores. Within the target range of scores the test was designed to measure, one should expect IRT SEs to be smaller and more precise than FA SEs, and therefore preferable to assess clinical change. Outside the target range, IRT SEs may be too large and more imprecise than FA SEs, and FA more precise to assess change. As a result, whether FA or IRT characterize reliable change more accurately in a sample will depend on the proportion of individuals within or outside the IRT target score range. An application is provided to illustrate these concepts.

Item response theory (IRT) model applications extend well beyond cognitive ability testing, and various patient-reported outcomes (PRO) measures are among the more prominent examples. PRO (and like) constructs differ from cognitive ability constructs in many ways, and these differences have model fitting implications. With a few notable exceptions, however, most IRT applications to PRO constructs rely on traditional IRT models, such as the graded response model. We review some notable differences between cognitive and PRO constructs and how these differences can present challenges for traditional IRT model applications. We then apply two models (the traditional graded response model and an alternative log-logistic model) to depression measure data drawn from the Patient-Reported Outcomes Measurement Information System project. We do not claim that one model is “a better fit” or more “valid” than the other; rather, we show that the log-logistic model may be more consistent with the construct of depression as a unipolar phenomenon. Clearly, the graded response and log-logistic models can lead to different conclusions about the psychometrics of an instrument and the scaling of individual differences. We underscore, too, that, in general, explorations of which model may be more appropriate cannot be decided only by fit index comparisons; these decisions may require the integration of psychometrics with theory and research findings on the construct of interest.

This commentary is an attempt to present some additional alternatives to the suggestions made by Reise et al. (2021). IRT models as they are used for patient-reported outcome (PRO) scales may not be fully satisfactory when used with commonly made assumptions. The suggested change to an alternative parameterization is critically reflected with the intent to initiate discussion around more comprehensive alternatives that allow for more complex latent structures having the potential to be more appropriate for PRO scales as they are applied to diverse populations.

A Correction to this paper has been published: https://doi.org/10.1007/s11336-019-09692-3