I had a read of the article, not every word, but I think enough to get the main message.
What they did: They asked people to respond to various questions at regular times throughout the day for 15 days, and they also got them to wear an actometer. The measures included how happy/sad/anxious the person felt, how much they were focusing on their symptoms/catastrophising and so on. The also got them to report on the level of fatigue they felt.
The idea was to see what factors were "perpetuating:" the self-reported fatigue.
Using factor analysis, they then reduced the number of measures, and came up with two main factors - activity (self-reported/objectively measured etc), and negative psycho-stuff (all the rest of the psycho stuff, but with happy/cheerful etc. obviously working in the reverse direction to the catastrophising and other bad stuff).
They kept the self-reported fatigue measures out of the factor analysis because that was the “outcome” measure.
They then looked at the correlation between the activity factor at a given timepoint and self-rated fatigue at the next timepoint. They did the same for the psycho-stuff factor. I think they were hoping to show that high scores on the negative psycho-stuff measure and/or low activity would predict greater fatigue at the next timepoint.
What they found: The associations - at least at the group level - were remarkably weak. Even the ones you'd expect would be strongly associated with each other weren’t very strongly associated - like you’d expect more activity at T1 to predict more activity at T2, because the activity measure would distinguish those with milder symptoms from those that are more severely affected. People were all over the place, with no strong consistent patterns of association.
But these guys, not wanting to produce a negative conclusion, talked up the idea of individuals varying respect to these associations. There is really no reason I can see for preferring this explanation over the other obvious one - that this is all just meaningless noise (a noise account would also predict the associations would vary all over the place form person to person, because, well, its noise).
My evaluation: Everything looks fairly cautious and carefully reported, except for the conclusions, which are well out of line with the incredibly negative results. It seems to be the conclusions where these researchers really create the lie. Just look at the title, and how it misrepresents the actual results - their pants are totally on fire.
.
