What seems pointless about this research is that they are trying to use very squishy inputs (words the person uses) to predict squishy target outputs (depression as diagnosed by this scale). In a way it seems similar to the problem of trying to interpret the effects of unblinded interventions on subjective outcomes. GIGO.
I think research like this might potentially be interesting if one side of it was 'hard' data - e.g. language --> biological measurements, or biological measurements --> depression scale, etc.. But as it is it just seems much too squishy on both sides to get a finding that could actually be anchored in anything.
A related major problem is that reading and responding to a depression questionnaire is itself a use of language, so there is a lot of circularity built into the analysis. Particularly in this case, because...
...While the title says 'children with a chronic illness', the only illness studied is CFS/ME. This means that interpretation of the depression questionnaire is extremely confounded by the symptoms of the condition being studied, in a way it wouldn't be if they were trying to study, say, chronic visible skin conditions, or refractory asthma, etc..
I mean it takes about 5 minutes of thinking to see that this is a waste, regardless of how well-meaning, or how 'cool' the idea is. Unfortunately it appears that in this field 5 minutes of critical thinking is not to be found.
