OBJECTIVES: Symptom PRO questionnaires should measure the most relevant symptoms for most patients in the target population. These symptoms are typically identified by plotting the prevalence and saliency of symptoms reported in patient interviews. It is more difficult to derive clear conclusions with this inductive approach when the sample size is small or when there is heterogeneity in symptom presentation or severity. MDS and CA are robust analytical methods that can be used in these scenarios. These techniques are sample-size independent and based on patterns of response (dis)similarities and association measures. MDS and CA group symptoms according to their overall impact across patients instead of their individual prevalence/saliency. Patient and symptoms outliers can also be identified.

METHODS: MDS and CA were applied to a small data set (n=24) of interviews with patients with rare eosinophilic gastrointestinal diseases (EGIDs). The data set recorded 41 heterogeneous symptoms. Data was converted into binary format (0/1). Jaccard Distances and measures of association between patients and symptoms were calculated. MDS and CA were weighted by average disturbance ratings and frequency of endorsement to identify clusters of salient symptoms for scale definition.

RESULTS: MDS and CA singled out abdominal pain, nausea and bloating as plausible candidates for items forming a cardinal symptom scale, given their prevalence, importance, strength of reciprocal association and centrality to patient experience. They also identified an additional five salient symptoms as the basis for items forming a PRO instrument. They validated the exclusion of 33 symptoms as being peripheral to the typical experience of EGID patients.

CONCLUSIONS: MDS and CA show considerable potential for the definition of symptom-based PRO measures and identification of potential scales for scoring algorithms of diseases with heterogeneous configurations of symptoms.