Sensitivity analysis
The costs for patients at the end of life are generally
high, and therefore we performed a sensitivity analysis of
the cost-effectiveness separately for patients alive and
deceased at 6 months. Furthermore, many observations
were missing for EQ-5D at 6-month follow-up. In the
cost-effectiveness analysis described above, these patients
were excluded. This is problematic since it means that
a substantial amount of information is lost, and the
results may thus be biased.19 We therefore performed
a sensitivity analysis with imputed values for missing data,
that is, a cost-effectiveness analysis in all 345 patients.
These values were imputed in a regression model, in
which we included the variables randomisation group,
age, sex, EQ-5D utility scores at baseline and at 6-month
follow-up and total costs. Five sets of data were created
(multiple imputation) and pooled results on EQ-5D
utility scores were used to estimate unadjusted and
adjusted gain in QALYs.