Diagnostic Accuracy of a Test Set to Classify Children with Auditory Processing Disorders (APD)

Abstract

To investigate the diagnostic performance of an testbattery. Sensitivity and specificity are measures to evaluate the validity of a test. These parameters were determined using ROC-curves for a battery of 10 diagnostic tests. The Youden Index, defined as the maximal effectiveness to determine the optimal cutpoint of diagnostic accuracy, was calculated as well. 91 children diagnosed with APD (51%) visiting 2(nd)-4(th) grade of a primary school; 87 2(nd)-4(th) graders with-out APD. A very good relation of sensitivity and false-positive-rate was found for the Mottier-Test, which measures auditory non-word repetition (Area under the Curve=AUC as a global statistic measure for validity: 0.96; p=0.000; 95%-CI: 0.93-0.99). The cut-off point to distinguish between children with and without APD was 17.5 raw score, according to the max. Youden Index 0.83 (sensitivity: 90.1%; specificity 93.1%; false-positive-rate: 6.9%). The HSET-Subtest "Imitation grammatischer Strukturformen" (measuring auditory short-term sentence memory) exhibited comparable high discriminative power (AUC: 0.94; p=0.000; 95%-CI: 0.90-0.98). At 21.5 raw score (max. Youden Index: 0.82), 84.7% of the children were classified correctly (false-positive-rate: 2.3%; specificity: 97.7%). Eight tests had a moderately diagnostic accuracy, two of them tended to lesser accuracy (phoneme analysis: AUC: 0.72; monaural temporal order judgment AUC: 0.75). Using certain tests of a defined test set for identification of APDs in primary school children according to the sensitivity improves the possibility to detect APDs. In order to precisely specify the APD and to decide which auditory dimension should be treated the entire combination of 10 diagnostic tests is indispensable.