More health economics updates

Quarto
R
Academia
health economics
Published

November 11, 2022

Hello everybody, I just wanted to take a break from the current series of posts about doing HTA in R and talk a bit about some interesting news related to the international HTA community and events.

More specifically, I have recently bumped into an update from the Dutch national health care institute or ZorgInstituut Nederland (ZIN) with respect to the provision of an update about HTA and cost-effectiveness reimbursement application guidelines within the country. You can find here all the relevant information (although the site is in Dutch the documents are in English so rest assured). I would like to talk about this update as I found it really interesting and I think it provides an incredibly welcome push from an institutional body towards the use of appropriate statistical software for reimbursement application procedures that companies are asked to comply in order for their proposed healthcare technologies to be evaluated.

One of the key points of the document is related to warn the audience about the intrinsic limitations of standard non-statistical software, e.g. Excel above all, for performing health economic evaluations which increasingly require the use of advanced statistical methods. Indeed, although Excel as certainly great advantages and merits when it comes down to spreadsheet purposes, it was also not specifically designed to perform statistical analyses or complex modelling tasks. In reality, due to its ease of access and popularity among private companies involved in HTA submissions, Excel is still currently one of the most used software for conducting HTA, at least in the private sector. However, as also the updated guidelines from ZIN highlight, the landscape of HTA modelling and statistical techniques is rapidly changing and the use of more appropriate software such as R becomes a necessary requirement (in the sense that these were developed with the objective of conducting advanced statistical analyses) in order to ensure a certain degree of consistency and quality in the statistical methods used within HTA submissions.

Of course, the guidelines also warn about the possible drawbacks of using R, such as the wider flexibility and free-source nature of the software, which may lead to complications that are difficult to spot if analysts are not familiar with the software (e.g. always using default commands relying on assumptions that are “hidden” instead of properly customise the command based on the current needs). In addition, due to its open-source framework, R packages and their functions are very susceptible to repeated updates and changes which may throw people off gard, especially if they do not keep up the pace with the different updates of the software or packages; similarly, this also implies that no overall and formal check from a responsible institution is done about the quality of the commands, package structures or even help files. This means that analysts should either be able to look into the code to check if everything is done correctly (which kind of eliminates the need to having a package) or blindly trust the author of the package on their commands. I know this can be quite unsettling for analysts used to fixed and stable commands in other software but I believe that, with the acquisition of familiarity with the software and standard (and typically more reliable) packages will allow analysts to acquire a much larger level of flexibility when it comes down to statistical modelling in HTA and will provide them with the necessary technical skills to modify or even create their own packages to solve their specific analysis problems.

The document then goes into more detail about the structure required for any reimbursement submission dossier in terms of data, tables, figures, code, etc … They also refer to the use of Markdown (I love it!) for structuring the R code provided with the submission with special care dedicated to the different aspects related to coding clearness (e.g. comments, packages used, testing methods, etc…). They also go all their way to provide a somewhat “comprehensive” list of recommended packages that ZIN suggests using for performing HTA modelling, among which even my own package appears, missingHE, urrah!!!

Overall, I find this news extremely positive in that I can see that in the Netherlands a strong interest is developing towards the use of R for ensuring a more coherent and quality wise more robust coding framework for conducting, assessing, reporting and comparing statistical analyses within HTA. I highly praise the effort that ZIN’s workers are putting into developing such framework and recognising how, by using a more transparent, appropriate and easy to check modelling technique, more fair economic evaluation assessments and recommendations for reimbursement decisions can be obtained. I really hope that even other national and international organisations and bodies involved in the process of HTA will follow ZIN’s example in adopting guidelines that take into account the importance of requiring a given level of quality for the statistical methodsused and a more transparent framework for their assessment.

Finally, and a bit unrelated to what I said so far, I would like to put here a recent tweet from Silvia Evers with regard to the next edition of lolaHESG, which this time will be held at EsCHER (Erasmus Centre for Health Economics Rotterdam) between 25-26th May.

Last edition I had a blast in presenting my work at this conference and meeting a bunch of amazing people involved in HTA in the Netherlands. I sincerely hope that I can replicate my experience in the upcoming year!