gms | German Medical Science

Clinical decision-support systems (CDSS) provide the potential to support experts’ decision-making based on the exploitation of big data. For example, a classification integrated in the CDSS can provide a statistical proposition which hearing device a patient would benefit from; or data-driven, unsupervised approaches can characterize patient groups available in the data, showing different profiles of audiological test outcomes and different prevalence. However, a main challenge is to base such a CDSS on international “big data”, since local clinical-audiological databases comprise different audiological tests and test conditions, data structure and formats, expertise of experts, or patient populations.

In previous work, e.g., [1], a CDSS for audiology was developed which is based on Common Audiological Functional Parameters (CAFPAs), an abstract parameter layer describing functional aspects of the auditory system, independent from the exact choice of audiological tests. Thereby, the CAFPAs allow integrating different databases, and provide an interpretable representation of the status of the auditory system. However, links from audiological tests to CAFPAs need to be established, and for different tests that are conducted for the same purpose, a common interpretation needs to be available. This is especially relevant for speech tests used for characterizing the listeners speech intelligibility performance. Speech tests are conducted in a variety of settings, with differences in the choice of speech material (words, every-day or Matrix-type sentences), language, or measurement procedure (adaptive vs. different fixed levels) leading to a different interpretation of the results.

Therefore, a model-based framework was developed that combines empirical data and model predictions based on the Speech Intelligibility Index (SII) and the Plomp model describing hearing loss by audibility (A) and distortion (D) component. This “hearing loss index” tool allows to describe the “effective hearing loss” of a patient as characterized by one speech test based on relationships and model predictions for different speech tests incorporated in the tool.

The development of this concept based on German databases comprising three typical German speech tests (Freiburg Monosyllabic Word Test, Goettingen Sentence Test, German Matrix Test) will be presented, complemented with first steps towards validation of the concept on French data.

In future work, the developed hearing loss index will be integrated in the CDSS, either directly or through its integration into CAFPAs. In addition, the development and application of data standards will be important towards integration of audiological databases.