Limitations of Computerized Adaptive Testing For Anxiety

To the Editor: The Computerized Adaptive Testing–Anxiety Inventory (CAT-ANX) was introduced as a new “test for anxiety,” with suggested large-scale screening uses and planned commercial availability (1). This proposed use lacks sufficient validation. The pool of over 400 items has not been demonstrated to have construct validity or predictive validity. The resultant CAT-ANX was not validated against existing anxiety scales. Test-retest reliability was not demonstrated. The “anxiety” of CAT-ANX was not defined. It is not coterminous with generalized anxiety disorder, although the article focused on this disorder. Operationally, CAT-ANX draws from many disparate disorders to pool multiform symptoms under the label “anxiety.” Justification for using CAT-ANX dimensional results in diagnosis was attempted through “diagnostic screening” for generalized anxiety disorder, with only modest success. At the preferred threshold score, sensitivity in the full development sample was 0.67 and specificity was 0.87. In epidemiologic studies, where generalized anxiety disorder prevalence is 3%, 86% of positive screens then would be false-positive, while 99% of negative screens would be true-negative. The limitations of dimensional CAT-ANX measures are apparent in Table 3 in the article (1). For patient 1 with “mild anxiety,” generalized anxiety disorder probability was 0.458, not clearly ruling in or ruling out the diagnosis, while the 40th percentile ranking relative to patients with verified generalized anxiety disorder diagnoses was quite consistent with caseness. Patient 2 with “severe anxiety” was said to have 99% probability of generalized anxiety disorder. However, the item responses suggest severe panic disorder with agoraphobia and/or social anxiety disorder, not generalized anxiety disorder. DSM-5 diagnosis was not reported for either case. The data in the article thus provide no sound basis to say that screening with CAT-ANX will allow confident, rapid, and accurate positive identification of key clinical anxiety diagnoses.