gms | German Medical Science

62. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V. (GMDS)

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie

17.09. - 21.09.2017, Oldenburg

Distributional regression of self-reported hearing abilities in the HÖRSTAT study

Meeting Abstract

Suche in Medline nach

  • Fabian Sobotka - Carl von Ossietzky University Oldenburg, Oldenburg, Deutschland
  • Petra von Gablenz - Jade University of Applied Sciences, Oldenburg, Deutschland; Cluster of Excellence "Hearing4all", Oldenburg, Deutschland
  • Inga Holube - Jade University of Applied Sciences, Oldenburg, Deutschland; Cluster of Excellence "Hearing4all", Oldenburg, Deutschland

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie. 62. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e.V. (GMDS). Oldenburg, 17.-21.09.2017. Düsseldorf: German Medical Science GMS Publishing House; 2017. DocAbstr. 222

doi: 10.3205/17gmds049, urn:nbn:de:0183-17gmds0499

Veröffentlicht: 29. August 2017

© 2017 Sobotka et al.
Dieser Artikel ist ein Open-Access-Artikel und steht unter den Lizenzbedingungen der Creative Commons Attribution 4.0 License (Namensnennung). Lizenz-Angaben siehe



Introduction: Self-reports of hearing (dis)abilities play an important role in hearing rehabilitation. Among the large variety of questionnaires, the Speech, Spatial, and Qualities of Hearing Scale (SSQ) has become an internationally widely used measure. The subjects are asked to rate their speech understanding in adverse listening situations, their abilities in sound localization, distance judgements, and sound perception using a visualized scale ranging from 0 to 10. Higher scores represent superior abilities. As research mainly focused on adults with impaired hearing, results from adults with “normal” hearing have barely been considered. However, the range of scores from adults who do not seek the assistance of hearing aid acousticians and do not report hearing problems is of interest as a reference group in hearing rehabilitation. We aim to model the complete conditional distribution of the SSQ scores using semiparametric quantile and expectile regression [1].

Methods: The German SSQ17 short-form developed by Kießling et al. [2] was used in the cross-sectional, population-based study HÖRSTAT (n=1903) along with a standardized interview and comprehensive hearing examinations including pure-tone audiometry and speech understanding in noise [3]. As the SSQ scores are extremely positively skewed, the mean information is not conclusive. Instead of the arithmetic mean scores a dense set of quantiles with a semiparametric regression model including linear, categorical and smooth effects using P-splines with a transformed LASSO penalty is estimated. The latter can be easily combined with the linear programming estimator for quantile regression models. Age, gender, globally reported hearing problems, hearing loss, and social status are included in the model.

Results: Even adults who perform well in pure-tone audiometry and tests on speech understanding in noise do not report perfect hearing abilities in all situations described by the SSQ items. Preliminary results suggest that age does not have any notable effect on SSQ scores if gender, pure-tone hearing loss, SRT and perceived hearing difficulties is controlled for. Gender by contrast, has an effect of 0.5 points on the rating of abilities in the lower tail of the distribution of spatial hearing. An interaction of gender with self-perceived hearing problems can be found in the SSQ ratings of every subscale. Generally, heteroscedasticity of SSQ scores is more pronounced for the ear with the worse pure-tone thresholds than for the ear with better hearing.

Discussion: Though the distribution of SSQ scores observed in “normal” hearing populations cannot establish any normative cut-off values, it provides an orientation about the scores that reasonably or at best can be expected from hearing-impaired adults fitted with hearing aids. A mean regression would not provide a sufficient amount of information in this regard.

Die Autoren geben an, dass kein Interessenkonflikt besteht.

Die Autoren geben an, dass kein Ethikvotum erforderlich ist.


Sobotka F, Kneib T. Geoadditive Expectile Regression. Computational Statistics and Data Analysis. 2012;56(4):755-767.
Kießling J, Grugel L, Meister H, Meis M. Übertragung der Fragebögen SADL, ECHO und SSQ ins Deutsche und deren Evaluation. Zeitschrift für Audiologie. 2011;50(1):6-16.
Von Gablenz P, Holube I. Prävalenz von Schwerhörigkeit im Nordwesten Deutschlands. HNO. 2015;63(3):195-214.