OK, scrap this post, I've just been reminded that, for Societal Costs, the paper says that the differences between CBT, GET and SMC "were not statistically significant."
Exploring the paper, and some rough calculations.
(These are just my rough back-of-an-envelope calculations, and are not exact or adjusted figures, and there might be major mistakes.)
And, like earlier, I might have misunderstood major issues.
Below, I compare the difference in cost changes for SMC/CBT/GET for the 12 month pre-randomisation period (I get these figures by doubling the figures for the 6 month pre-randomisation period) with the 12 month post-randomisation period.
Societal Costs / Cost Benefit Analysis
My current understanding is that GET and CBT are considered to have a better cost benefit, based on 'societal costs' for GET and CBT decreasing sightly more than SMC:
(GET = extra £197 decrease in societal costs as extra to SMC £1914 decrease)
(CBT = extra £464 decrease in societal costs as extra to SMC £1914 decrease.)
At least, I think this might be the case, but I have no idea how exactly they worked it all out. The complexity is astounding.
Considering that these are mean figures, it's especially confusing because they are measuring the effects and costs on the 87% of participants who experienced no primary outcome effect for CBT/GET.
Here are my calculations:
Comparing changes in health care and societal costs from 12 month period before randomisation and 12 month period after randomisation - Difference between SMC and GET/CBT
Difference between SMC and GET (increases/decreases in means):
Extra health costs for GET (GET - SMC) 806 - (-116) = 922
(GET changes in health care costs are £922 more than SMC.)
(GET = £922 increase in health care costs, separate to SMC. SMC = £116 decrease in health care costs.)
Extra societal costs (GET - SMC) (-2111) - (-1914) = -197
(SMC societal costs are £197 more than GET.)
(GET = £197 decrease in societal costs, as extra to SMC. SMC = £1914 decrease in societal costs.)
Difference between SMC and CBT (increases/decreases in means):
Extra health care costs for CBT (CBT - SMC) 812 - (-116) = 928
(CBT changes in health care costs are £928 more than SMC.)
(CBT = £928 increase in health care costs, separate to SMC. SMC = £116 decrease in health care costs.)
Extra societal costs (CBT - SMC) (-2378) - (-1914) = -464
(SMC societal costs are £464 more than CBT.)
(CBT = £464 decrease in societal costs, as extra to SMC. SMC = £1914 decrease in societal costs.)
And these figures are interesting:
Health care costs per person improved for improvements in physical function.
SMC
Per person health savings (-116 x 148) / (148 x 58%) = £200 savings per person improved, during 12-month post-randomisation period.
GET
Per person health costs (922 x 140) / (140 x 12.6%) = £7317 health costs per person improved, separate to SMC savings, during 12-month post-randomisation period.
CBT
Per person health costs (928 x 145) / (145 x 13.4%) = £6925 health costs per person improved, separate to SMC savings, during 12-month post-randomisation period.
I might have got this wrong, but it does raise questions.
The extra spent on health care per person is about £7000 for CBT and GET, on top of spending/savings for SMC.
-----