In practice, we can plot other matched measures (Φ(μ^i1),Φ(μ^i2)) to get the Bland-Altman graph on the probability scale. If the distribution of values on the probability scale is strongly distorted, we can perform a transformation and then represent the Bland-Altman graph of the transformed data if H0 is rejected, we conclude that there is a significant difference between the two measurement methods. Otherwise, we cannot reject the assumption that the two measurement methods are consistent, and the measures of the conformity of the method introduced in the next section would further assess the scope of the agreement. In the broadest sense, the term “measurement method” can refer to a medical device, an instrument, a questionnaire battery or a human judge. In this article, we use the term “evaluator” specifically for human judges. In the medical context, an evaluator may refer to a physician in a diagnostic process or to a clinical observer in a clinical trial. There are cases where the measurement of response does not necessarily require the evaluator to make a subjective assessment, para. B example when the patient`s blood pressure is read from a monitor. We use the term “recorder” specifically for the human judge, whose subjective evaluation is not required. “Evaluator reliability” is the term for communication between evaluators, while “method compliance” is the term for agreement between measurement methods. Throughout the literature, reviews of the agreement assessment generally deal with the conformity of the method and the reliability of the intervaluor in the sense that agreement indices can often be used for both method compliance and interletute reliability, although the entities against which the agreement is assessed are different. A detailed overview of the evaluation of the agreement can be found in the literature.4–7 The main approaches for continuous and categorical measures are summarized as follows.
Our methodology is versatile and based on the GLMM framework in the sense that it can be extended to different data structures in more complicated cases. In general, measurements can be continuous, binary, or ordinal, while this paper focuses on repeated binary measures. The theory of our approach requires hypothesis J → ∞, which is a common condition for GLMM-based approaches to ensure consistent asymptotic estimates of fixed effects and variance components for random effects. Although our approach theoretically aims to cover large-scale studies in which there are a large number of reviewers and subjects, the additional simulation study in the additional documents shows that our approach still works well with only a few reviewers in the dataset as long as there is a reasonably good agreement between the reviewers. In this article, we consider the evaluator effect to be random by intending to generalize the results of method matching to a large population of evaluators. On the other hand, if the evaluator effect is assumed to be fixed, there is a simple extension of the HYPOTHESIS TESTING PROCEDURE based on the GLMM. In the meantime, the newly developed Bland-Altman diagram and Cohen`s Kappa could be leveraged by a fixed set of evaluators for each evaluator. Alternatively, Cohen`s kappa, presented in the next section, could also provide a way to measure the extent of the method`s congruence. Footnote 1. In practice, if a data set consists of only a few appraisers, we could consider another scenario with a high value of Jσγ2/σαm2. If the variance of the evaluator σαm2 is relatively small, the value of Jσγ2/σαm2 could still be large even with only a few evaluators in the dataset.
A relatively low value of the evaluator`s variance usually implies a good match between evaluators. In the medical context, for example, training is needed to ensure that physicians make consistent diagnoses for the same patient. In this case, despite only a few evaluators, the amount μ ̄i1−μ ̄i2 is still appropriate to measure the difference between two measurement methods for each subject. The simulation results for this scenario are specified in the additional documents. In the simulation, we define the number of subjects, evaluators and points in time as I = 100, J = 30 and Ti = 5 for each i = 1, …, I. We will first demonstrate our approach in Section 3.1 using a simulated dataset for each configuration. We will also present the average values of parameter estimates based on 1000 replicas of Monte Carlo. In Section 3.2, we will compare our approach with approaches that do not take into account the evaluator`s effect. GlMM customization is implemented by PROC GLIMMIX in SAS 9.4.35 Symmetric tuning measures are not affected by the exchange of X and Y variables. These are useful in many other cases, for example. B when comparing observers, laboratories or other factors where neither is a natural comparator.
There are several measures based on the average of the proportions in which X corresponds to Y and Y corresponds to X. The Kulczynski, the Dice-Sørensen and the Ochiai are three of those measures that use means of arithmetic, harmonic and geometric proportions. Method comparison studies are essential for development in the medical and clinical field. These studies often compare a cheaper, faster, or less invasive method of measurement with a widely used method to see if they have a sufficient match for interchangeable use. In addition, in many clinical and medical evaluations, as opposed to simply reading measurements from devices, e.B reading body temperature from a thermometer, the measurement of the response is influenced not only by the meter, but also by the evaluator. For example, widespread inconsistencies among evaluators in psychological or cognitive assessment studies are often observed due to different characteristics such as the training and experience of evaluators, especially in large-scale evaluation studies where many evaluators are employed. This article proposes a model-based approach to evaluate the agreement of two measurement methods for binary measurements repeated in pairs in the scenario where the agreement between two measurement methods and the agreement between the evaluators must be studied simultaneously. On the basis of generalized linear mixed models (GLMM), the decision on the appropriateness of interchangeable use is made by testing the equality of fixed effects of the methods. Approaches to evaluate method matching, such as the Bland-Altman diagram and Cohen`s Kappa, are also being developed for repeated binary measures based on latent variables in GLMMs. We are evaluating our new model-based approach using simulation studies and a real-world clinical application in which patients are repeatedly screened for delirium using two validated screening methods. Simulation studies and analysis of real-world data show that the approach we propose can effectively assess the alignment of methods. .