I am working on an economic evaluation of a drug to promote wound healing on diabetic foot ulcers based on retrospective data (cost-consequences). I need to make a sensitivity analysis of efficacy and cost of the intervention. To an univariate analysis how to select the range of these variables? Which other alternatives can I use for the analysis?
Hi Darien,
Can you say what is the exact purpose of your sensitivity analysis?
For now, I will assume that you are doing this sensitivity analysis because it is customary, possibly because someone else told you to do so, and it might even be that this someone else does not know too well why you should do this.
If you do a one-way sensitivity analysis to investigate how analysis results change with changes in your input parameters, then a consistent percentage increase and decrease (e.g. + and – 10%) in each of your parameters would be good.
If you want to follow what is customary, then you should try to vary the values of your parameters to the upper and lower limit of a plausible range. If a parameter is estimated from a single data set and you have no further doubt that the data provide an unbiased estimate, then you can take the 95% confidence interval for that. If you do have those doubts, then you should choose an interval that is longer to reflect your uncertainty. For some parameters, you probably cannot estimate a confidence interval, then you should best ask people with expert knowledge about those parameters what they think are plausible intervals. And if you can’t find access to those experts, you will have to make them up yourself. Unit costs are generally treated a bit differently and there is a custom to simply vary them by + or – 20% or 30%.
I would like to not, though, that for a better impression of the uncertainty of your analysis results, it is better to do a multivariate sensitivity analysis but I imagine that you don’t have time for all that.
I hope that answers your question but please let me know if you have further information, questions, or comments.
Best regards,
Rob
Next to my rather 'pragmatic' reply, you may want to check some other sources, e.g. text books on health economic evaluation, or online at e.g. the ISPOR web site:
https://www.ispor.org/workpaper/Modeling_Methods/Model_Parameter_Estimation_and_Uncertainty-6.pdf
Thanks a lot Dr. Robert! Actually I am doing a Master's Degree in Health Economics. The aim of my sensitivity analysis is to determine the robustness of my economic evaluation. Certainly, it is required that I incorporate the variations of the variables of cost and consequences, although statistically I showed that other parameters influence the results and constitute sources of uncertainty. Prior to your response, I exchanged with some professors and they criticized the use of per cent intervals. Is it correct to use these intervals or do I have to go to media analysis and standard deviations for my plausible range in sensitivity analysis? Is the latter option valid even when the data does not have a normal distribution?
I will review your proposal to incorporate a multivariate analysis as a better option. Thanks for the paper!
If the sensitivity analysis is intended to show robustness, then I think you should use plausible ranges and not and arbitrary percentage up and down.
I would also say that a plausible range is generally wider than a standard deviation up and down.
I would start with trying to find or estimate 95% confidence intervals.
Is that what you mean by "per cent intervals"?
If so, then I assume that the critique you got about that is because there may be more uncertainty than described by the confidence interval. That only describes uncertainty due to random error in a specific analysis. They should have given you some clue why they think that there is more uncertainty and preferably also suggest a way to determine actual uncertainty. It is hard for me to tell without detailed knowledge about your input parameters and their function in your evaluation.
(I don't know what you mean by "media analysis".)
I wish you the best of success with getting your degree and I hope that you will still want to continue with this type of work after you're done!
By the way: I would also love to know a bit more about how health economics is applied in Cuba.
- Badges
- Science topic