Health economics

Survival analysis in health economic evaluation Contains a suite of functions to streamline systematically the workflow involving survival analysis in health economic evaluation. survHE can fit a large range of survival models using both a frequentist approach (by calling the R package flexsurv) and a Bayesian perspective. For a selected range of models, both Integrated Nested Laplace Integration (via the R package INLA) and Hamiltonian Monte Carlo (HMC; via the R package rstan) are possible.

This tutorial is based on the Journal of Statistical Software paper describing the philosophy underlying survHE and its main functionality. All the technical details (including the statistical background) are described in the paper, while this tutorial is only meant to provide some annotated description of the main important functions provided in survHE. The example uses a fictional dataset — in fact the actual data are created by “digitising” an existing Kaplan Meier curve.

BCEAweb is a front-end that can be used to access many of the functionalities of the R package BCEA. It can be accessed here. Files The landing page of the web-app has a very short tutorial proving a very brief guide and pointing to files that can be downloaded to test its functions. These are Vaccine_spreadsheet.csv: a spreadsheet with simulated values for all the model parameters for the “Vaccine” fictional example, discussed in the BCEA Book.

I think it’s now official: after a couple of years in which Covid has ruined everything, we’re now returning with our summer school on Bayesian methods in health economics! Of course, so many bad things have happened that this can hardly count as a major problem, but I have really missed the good fun we’ve always had in the week-long course, in Florence.

I’ve made a major refactoring of (the development version of) survHE. I guess one of the main issues with the package (both from the point of view of the user and the maintainer) was that survHE is a big package and installation is a very lengthy process. And this is no surprise: the trade-off here is between the massive savings in computational time that are obtained by pre-compiling the Bayesian models available (through rstan) and the time it takes to get everything installed on your machine… And, from the developer’s point of view, often submission to CRAN has been a pain, because some of the files that get installed are very large and, again because of the nature of the package, there’s quite an intricate structure of “dependencies”, which makes the package very heavy.

This week has been a fantastic one. I went to Cambridge to give the 18th Armitage Lecture; I have to say, when I was invited back in the summer, my first reaction was of course that I was very, very pleased and honoured. Looking at the list of previous Lecturers was pretty daunting, though — which in a way only increased the honour, while also increasing slightly the pressure… 😉

More changes to survHE. Earlier this week, I gave a talk at the Virtual Ispor US conference (our panel was W11). In preparation for the panel (and other things I was doing anyway), I got to talk to colleagues and thought a bit more of how survival modelling in HTA is kind of weird and “special”… In particular, I realised how survHE doesn’t give by default any real information about the hazard and the cumulative hazard functions — this isn’t entirely true, because once the survival curve has been estimated (which is the whole point of survHE!

Philip Cooney has alerted me to a slight inconsistency in survHE. Basically, he was trying to work in a setting where the data contain no censoring, using the HMC/rstan models generated by survHE. All works OK when using some of the models implemented, but failed when trying to fit the Gamma, Generalised Gamma and Generalised F. The reason for this is that, for simplicity (and actually, our empirical tests show that computational run-time is much reduced) these models construct the log-likelihood by separating the set of individuals who are fully observed from the one of individuals who are censored.

A couple of colleagues have alerted us to some issues with the BCEA function evppi, under MS Windows. Basically, it seemed as if under MS Windows and with R 4.0.0 (which is a recent, though not the most recent release of R) the evppi function would break, when trying to make the analysis based on INLA (and the work discussed here).

Earlier this week, while I was working on my second-last lecture of the year (I’ve been relatively good and the vast majority of the new material I had to prepare to make everthing accessible — and I also took the chance to give my slides a little makeover, mostly learning and enjoying xaringan — but this one I kind of forgot I had to do it, so I needed to kind of do it quickly this week… But as usual, I digress…), I discovered a little bug in survHE.