ABSTRACT

Aim:

The aim of this study is translate the STARD (The STAndards for Reporting of Diagnostic Accuracy) criteria which were developed in 2003 by a group of epidemiologists and biostatisticians in order to establish a standard for planning and reporting the results of the diagnostic accuracy studies.

Materials and Design:

The STARD criteria that were published in 2003 have firstly been translated into Turkish language by the first four authors (RA, MAT, CA, and PS) independently and afterwards four translated versions have been combined with consensus of the whole group headed by YG. For assuring the language equivalency between the original and the translated STARD statements, the translated criteria was also checked by a professional interpreter and necessary corrections were made. The final version of the translated criteria is translated into English by another professional interpreter using reverse-translation method and found to be quite similar with the original. For checking the understandability and repeatability of the translated version of the STARD criteria, 2 pathologists were asked to use the translated STARD criteria and evaluate the quality of 20 diagnostic accuracy studies. The inter-rater agreement for each criterion was investigated by using consistency percentages and Cohen’s kappa statistic. Moreover, the agreement is also assessed by using Bland-Altman method and Within Group Correlation by using the total number of reported criteria.

Results:

The investigated 20 diagnostic accuracy studies were found to be variable with respect to reported criteria percentages. Some of the criteria were found to be reported in all studies whereas some other criteria were reported in none of the studies. Total number of reported criteria per article is found to be 15.9±3.2 (Rater 1) with a minimum of 10 and a maximum of 21. Agreement percentages per criteria were found to be 60% for one of the criterion, and 75% or above for the others. Kappa statistic could not be estimated for two items, was found to be not significant for 3 items and was above 0.32 for 20 items. The limits for agreement were found as -2.2-3.5, and it shows that the inter-rater agreement has an acceptable level along with SKK is calculated to be 0.945.

Conclusion:

It is expected that the translated STARD criteria will be a guideline for the researchers in the diagnostics accuracy studies as well as for the journal reviewers and editors. The use of Turkish translated version of STARD as a standard measure will increase the quality of diagnostic accuracy studies as well as the meta analysis studies.