Aanesthetic Agents for Day Surgery - NIHR Health Technology ...
Aanesthetic Agents for Day Surgery - NIHR Health Technology ...
Aanesthetic Agents for Day Surgery - NIHR Health Technology ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
42<br />
Economic evaluation methods<br />
similar to the bootstrap. Briggs and Gray 255<br />
offer specific guidance <strong>for</strong> judging whether skew<br />
in the data will have important implications <strong>for</strong><br />
the sampling distribution of the mean. They<br />
suggest that the skew in the sampling distribution<br />
of the mean (S m) will be a factor: S m = S S – √n S,<br />
where S S is the skewness of the original sample<br />
and n S is the number of observations in the<br />
original sample. When this rule was applied,<br />
the skewness in the samples was found to be<br />
sufficiently low to indicate normality <strong>for</strong> the<br />
sampling distribution of the mean. 255<br />
There<strong>for</strong>e, the t-test was employed in this study.<br />
The t-test (and analysis of variance (ANOVA) <strong>for</strong><br />
multiple group comparisons) was used <strong>for</strong> length<br />
of stay, cost, CV and net benefit variables.<br />
Sensitivity analysis<br />
Uncertainty: CESA RCT data<br />
Sensitivity analysis was required to supplement the<br />
statistical analysis, in order to assess the level of<br />
uncertainty in the data collected within the CESA<br />
RCT and the subsequent internal robustness of the<br />
results. Sensitivity analysis was used to evaluate<br />
uncertainty <strong>for</strong> two cases, as described below.<br />
Incremental cost-effectiveness ratios<br />
Statistical analysis is not appropriate <strong>for</strong> testing<br />
the robustness of ICERs. It is not possible to<br />
generate 95% CIs around ICERs because the<br />
ratio of two distributions does not necessarily<br />
have a finite mean or, there<strong>for</strong>e, a finite variance.<br />
256 The 95% CIs of the principal clinical<br />
outcome (PONV) were used to recalculate<br />
ICERs in order to assess the impact of uncertainty<br />
regarding clinical outcomes on ICERs. A simple<br />
deterministic sensitivity analysis was used to<br />
explore the impact of varying the incidence of<br />
PONV. The values <strong>for</strong> PONV incidence were<br />
varied between the limits of the 95% CIs. The<br />
low rate of PONV incidence <strong>for</strong> each arm was<br />
used simultaneously in the calculation of each<br />
ICER in the sensitivity analysis.<br />
This was followed by the generation of a<br />
bootstrap estimate of the ICER sampling<br />
distribution to identify the magnitude of<br />
uncertainty around the ICERs. This method<br />
allowed uncertainty around both the costs and<br />
effects to be taken into account. Bootstrapping<br />
with replacement was employed, utilising<br />
Microsoft’s Excel ® , using 1000 iterations.<br />
The 2.5% and 97.5% percentiles of the<br />
ICER distribution were obtained.<br />
Validity of the use of the Dion approximation<br />
of volatile consumption<br />
A substudy showed that the Dion algebraic<br />
approximation consistently underestimated<br />
the amount of volatile anaesthetic used (see<br />
appendix 18). Results from this study suggested<br />
that the actual amounts of isoflurane and sevoflurane<br />
were 6% and 27% higher, respectively,<br />
than estimated. There was a wide variation<br />
in these inflation factors. One-way sensitivity<br />
analysis was used was to recalculate the<br />
variable costs using inflation factors. The<br />
total variable cost <strong>for</strong> each group was calculated<br />
<strong>for</strong> each inflation level. ICERs were<br />
recalculated <strong>for</strong> the range of inflation<br />
factors, if appropriate.<br />
Uncertainty: differences between the<br />
CESA RCT and routine practice<br />
The trial protocol and comparators were defined<br />
to reflect routine practice as far as possible.<br />
However, it was recognised that anaesthesia<br />
practice was changing. Additional analyses were<br />
planned to combine the CESA RCT data with<br />
published evidence to explore the relative costs<br />
and effects of anaesthesia practice not included<br />
in the trial. The literature review (see chapter 2)<br />
and the national survey (see chapter 3) indicated<br />
that some anaesthetic agents and practices<br />
now in use were not measured in the trial.<br />
Decision analysis and probabilistic sensitivity<br />
analysis were used to assess the impact of<br />
the following on patient outcomes and<br />
total costs:<br />
• differences between the trial and routine<br />
practice in the use of prophylactic anti-emetics<br />
in adults (CESA MODEL)<br />
• differences between the trial and routine<br />
practice in the inhalational anaesthetics<br />
used in children (CESA MODEL).<br />
Additional data <strong>for</strong> these analyses were obtained<br />
from the literature review (see chapter 2) and<br />
national survey (see chapter 3). The literature<br />
review extracted and evaluated the clinical and<br />
economic evidence <strong>for</strong> the sensitivity analysis and<br />
modelling section. The data obtained from the<br />
survey are summarised in chapter 3.<br />
The analysis of uncertainty used the mean<br />
costs and variance calculated <strong>for</strong> the trial-based<br />
analysis above. To extrapolate the results of the<br />
trial to alternative anaesthetic agents and PONV<br />
prophylaxis practices, the results of the trial were<br />
synthesised with the data from the literature<br />
reviews and national survey.