OBJECTIVE: Somatosensory amplification refers to a person's tendency to experience somatic sensations as inappropriately intense and involves hypervigilance concerning bodily sensations. We applied the Somatosensory Amplification Scale (SSAS) in an Internet sample of young adults (N=3031) to test whether the SSAS is Rasch scalable. METHODS: We applied mixture distribution extensions of the partial credit and rating scale models to identify possible subgroups that use the response set of the SSAS in different ways. RESULTS: A partial credit model, with two latent classes, showed a superior fit to all other models. Still, one of the SSAS items had to be removed because it showed severe underfit. Overall fit of the remaining items was acceptable, although the differentiation between at least two of the five item categories was questionable in both classes. Class 1 was characterized by a higher SSAS sum score, female gender, more somatic complaints, more anxiety, more psychosocial stress, and slightly higher depressiveness. Further exploratory analyses showed that the three mid categories of the SSAS can be collapsed without a large loss of information. CONCLUSIONS: Our results show that a shortened version of the SSAS is Rasch scalable but also reveal that there is a lot of room for further improvements of the scale. Based on our results, Item 1 should be removed from the scale and a reduction of the number of response categories is probably warranted.
OBJECTIVE: Somatosensory amplification refers to a person's tendency to experience somatic sensations as inappropriately intense and involves hypervigilance concerning bodily sensations. We applied the Somatosensory Amplification Scale (SSAS) in an Internet sample of young adults (N=3031) to test whether the SSAS is Rasch scalable. METHODS: We applied mixture distribution extensions of the partial credit and rating scale models to identify possible subgroups that use the response set of the SSAS in different ways. RESULTS: A partial credit model, with two latent classes, showed a superior fit to all other models. Still, one of the SSAS items had to be removed because it showed severe underfit. Overall fit of the remaining items was acceptable, although the differentiation between at least two of the five item categories was questionable in both classes. Class 1 was characterized by a higher SSAS sum score, female gender, more somatic complaints, more anxiety, more psychosocial stress, and slightly higher depressiveness. Further exploratory analyses showed that the three mid categories of the SSAS can be collapsed without a large loss of information. CONCLUSIONS: Our results show that a shortened version of the SSAS is Rasch scalable but also reveal that there is a lot of room for further improvements of the scale. Based on our results, Item 1 should be removed from the scale and a reduction of the number of response categories is probably warranted.