New Perspectives for the Assessment of Depression: Development of an Item Bank and a Screening Instrument Applying Rasch Analysis and Structural Equation Modelling. / Neue Perspektiven der Depressionsdiagnostik : Entwicklung einer Itembank und eines Screeninginstruments mit Hilfe der Rasch-Analyse und Strukturgleichungsmodellierung. Der Medizinischen Fakultät der RWTH Aachen vorgelegte Dissertation zur Erlangung des akademischen Grades eines Doktors der Theoretischen Medizin

113 S. Gesamtzustand sehr gut --- Depressive disorders are among the most prevalent mental disorders and cause substantial harm and affliction as well as reduced quality of life and impaired psychosocial functioning. Apart from the clinical interview, questionnaires represent the most established and economic diagnostic instruments for depression. A multiplicity of questionnaires for depression has been developed across the past decades. However, several studies demonstrated substantial psychometric shortcomings of many depression questionnaires. Thus, further enhancements in this area appear to be mandatory. The present thesis comprises three studies that aim at further improving the quality of diagnostic methods for depression by making the benefits of modern test theoretical approaches, particularly the Rasch-Model, one of the Item-Response-Theory-Models, available for clinical practice. The first study aimed at providing an ample pool of depression items the basic requirement for item bank calibration. A thorough review of existing diagnostic tools for depression and relevant literature led to a comprehensive compilation of depression items. Based on the relevance judgments of 49 adept clinical experts 157 relevant items were identified. An additional major finding of Study One was that clinical psychologists tended to regard emotional/ cognitive items as more important than psychiatrists while the latter tended to judge somatic items as more relevant than clinical psychologists. However, across all experts items about somatic complaints were considered as not relevant. The second study was the core of the current thesis and based on the item pool of Study One aimed at developing a calibrated item bank for depression and evaluating its psychometric properties. Evaluation of Rasch Model fit, differential item functioning (DIF), dimensionality, local independence, item spread, item and person separation, and reliability resulted in a bank of 79 items that showed good psychometric properties. However, a minority of items showed DIF that must be dealt with in all future applications and utilizations of the bank. Thus, CAT applications on the basis of the item bank should incorporate separate item difficulty estimates for the respective groups (e.g., female vs. male test takers) and questionnaires derived from the bank should hold gender or age dependent norms. Study Three aimed at exemplifying how the Rasch calibrated item bank for depression may be utilized by the derivation of two parallel short but sensitive static depression questionnaires that should cover a sufficiently high spread of depression severities with high test accuracy. This major aim was achieved with the development of the Rasch-based Screening for Depression I and II (DESC-I/ -II), a short depression scale in two parallel forms comprising 10 items each. This instrument may provide considerable benefits in retest applications and for screening purposes in mentally as well as somatically ill patients. Future investigations should engage in validating item calibrations and Rasch model fit of the item bank in a larger sample and DIF analyses should be repeated. Furthermore, psychometric properties of DESC-I and -II should be further investigated, including e.g. evaluation of its retest reliability and sensitivity to change, especially in heterogeneously composed samples that realistically reflect the patient mixture encountered by the practitioner. Finally, the adaptive algorithm that accesses the calibrated item bank should be built and evaluated to allow for computer-adaptive test applications. Taken together, the results of the present thesis constitute a psychometrically sound basis for computer-adaptive testing and provide a promising new screening instrument for depression ready for use. If the major suggestions for future investigations on the basis of the current findings are further pursued, a substantial leap towards better quality and heightened conv.…