I've been doing some homework on the PACE FOI-released data, regarding perceived Vs actual physical function, and being as GETSET shares the same issues, I'll post some charts here, four of them in all. Click on a chart to magnify it.Something that strikes me as strange about the SF-36 PF10 is that the possible scores are 0,5,10,15,20...100.
On the 10 questions you get 0 points for "limited a lot," 5 points for "limited a little," and 10 points for "not limited at all." Thus, the lower the score, the more limited you are.
But... is the interval between "not limited at all" and "limited a little" really the equivalent of the interval between "limited a little" and "limited a lot"? It doesn't seem like a very linear scale.
Since the difference between the GES and control groups was equivalent to one interval on one question out of the 10, it would be nice to know if that interval was between none and a little or a little and a lot.
Notes:-
- The charts show % improvement, not absolute improvement. So someone who did 10m at baseline and 15 at 52 weeks shows 50% improvement, same as someone who did 100m at baseline and 150 at 52 weeks.
- The charts are all about comparing a person's perceived percentage physical function improvement Vs actual % physical function improvement.
- Remember PACE interprets a "50% increase from baseline in SF-36 sub-scale score as a positive outcome".
- Each table, within each trial arm, show participants (X axis) from the lowest % increase in actual physical function, through to the highest % increase in actual physical function.
- For each participant, their perceived physical function % improvement is shown, as well as their actual % physical function improvement.
- The Y axis is % improvement. Negative values mean % PF decreased at 52 weeks compared to baseline. 0% means no change between baseline and 52 weeks.
- Perceived PF is as per the SF-36 PF sub-scale, and actual PF as per the 6 min walk test.
- In the data there are some participants who rated their PF at 0% at baseline, which gives divide-by-zero if trying to gauge % increase. I have got around this by treating their baseline as 1%, which still shows the expected sharp spike in % change, but keeps the values finite. I didn't want to exclude these participants, because they are a good indication of the disconnect between perceived and actual % PF improvement.
- There were quite a lot of drop-outs in the PACE data, with NULL results. These are excluded from the charts.
Last edited: