Accuracy and reproducibility of the ETDRS visual acuity chart: methodological issues

Comment on Hall et al. (2017), "How to Choose Between Measures of Tinnitus Loudness for Clinical Research? A Report on the Reliability and Validity of an Investigator-Administered Test and a Patient-Reported Measure Using Baseline Data Collected in a Phase IIa Drug Trial"

PURPOSE

The purpose of this letter, in response to Hall, Mehta, and Fackrell (2017), is to provide important knowledge about methodology and statistical issues in assessing the reliability and validity of an audiologist-administered tinnitus loudness matching test and a patient-reported tinnitus loudness rating.

METHOD

The author uses reference textbooks and published articles regarding scientific assessment of the validity and reliability of a clinical test to discuss the statistical test and the methodological approach in assessing validity and reliability in clinical research.

RESULTS

Depending on the type of the variable (qualitative or quantitative), well-known statistical tests can be applied to assess reliability and validity. The qualitative variables of sensitivity, specificity, positive predictive value, negative predictive value, false positive and false negative rates, likelihood ratio positive and likelihood ratio negative, as well as odds ratio (i.e., ratio of true to false results), are the most appropriate estimates to evaluate validity of a test compared to a gold standard. In the case of quantitative variables, depending on distribution of the variable, Pearson r or Spearman rho can be applied.

CONCLUSION

Diagnostic accuracy (validity) and diagnostic precision (reliability or agreement) are two completely different methodological issues. Depending on the type of the variable (qualitative or quantitative), well-known statistical tests can be applied to assess validity.

The minimum basic dataset for diagnoses of cerebrovascular disease: Methodological issue on reliability

Reliability (precision, repeatability, agreement) and validity (accuracy) are two completely different issues which should be assessed using appropriate tests. It is crucial to know that, reporting concordance rate; the selection error rate and the classification error rate are not the most appropriate estimates to assess reliability. Regarding reliability, for qualitative variables, weighted kappa should be used with caution. However, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), likelihood ratio positive and negative as well as diagnostic accuracy are estimates that are usually used to evaluate the validity of a test compared to a gold standard.