The intervention group had greater mean increase in
HRQoL over time than controls as determined by overall
SF-36 (P = 0.0001). Increasing age (P = 0.0006) and living
alone (P = 0.01) were associated with smaller increases in
HRQoL. No other variables had a significant effect on
outcome and so were not included in the model.
Table 2 shows the baseline and follow-up least mean
square values for the overall SF-36 scores in addition to the
physical component scale (PCS) and mental component
scale (MCS) of the SF-36. There was a significant difference
in change in SF-36 PCS (P = 0.0002) and MCS
(P= 0.0003) in the intervention group compared with controls. Classification factors included in both of the models
were: gender, living arrangement and ethnicity
The intervention group had a lower mean starting SF-36
score and then a large mean increase over time, whereas the
control group showed a slight decrease in SF-36 over time.
We thus investigated the possibility of regression to the
mean (RTM) [20]. Residual diagnostic tests were undertaken
to determine the effect of outliers across the sample, and the
resultant changes in results when these outliers were
manipulated. Residuals were examined and sensitivity analyses without outliers were undertaken. There was no indication that outliers had a modifying effect on change in
SF-36 across the two groups.