Launch party

I think I've already mentioned this work here and here: after much tribulation, mostly due to the fact that we had to co-ordinate a relatively large number of papers in a single journal issue, we are very close to the publication of our work on the Stepped Wedge cluster-randomised in Trials. 

Interestingly, we're having a launch party (academic-style, of course, ie in the form of a symposium) to celebrate the special issue. This will be held at the London School of Hygiene and Tropical Medicine in September (details are below). 

I'm not sure we'll manage to get as many people as in the picture above, though...

What is a Stepped Wedge Trial?

The term Stepped Wedge Trial was coined by LSHTM’s Professor Peter Smith to describe a study of the impact of the roll out of Hepatitis B vaccination in the Gambia. The design has subsequently received significant and growing interest. The Stepped Wedge cluster-randomised trial is a research design in which clusters are randomly allocated to a time point at which they will introduce a new intervention. The design is used in a wide range of areas of public health, as well as other areas of public policy such as education and international development. It is an attractive design for a number of evaluation scenarios: notably the design is commonly recommended for pragmatic, implementation-science and translational evaluation studies. However, it also poses several challenges to researchers: guidance for many aspects of SWT are absent and there remain significant debate about the strengths and weaknesses of the design.

Symposium details

Researchers at the LSHTM Centre for Evaluation and UCL and written six papers on Stepped Wedge Trials (abstracts attached), due to be published at the end of July. We will host a symposium at LSHTM on 22 September, 2-5pm to discuss these papers and the remaining controversies. The detailed agenda will be released soon but will include plenty of time for lively debate and will be followed by a drinks reception.

The aims of the meeting are:

• to share current state-of-the-art understanding on the rationale, design, analysis and approaches to sample size determination for stepped wedge trials
• to foster discussion between, and provide technical support to, practitioners and methodologists considering, planning and undertaking Stepped Wedge Trials
• to present and launch a series of papers in the journal Trials on Stepped Wedge Trials

Questions the meeting will address include:

• When should you, and when shouldn’t you, do a Stepped Wedge Trial?
• What are the commonest, and what are the most robust, characteristics of Stepped Wedge Trials?
• How could you, and how should you, analyse data from a Stepped Wedge Trial?
• How should you determine the size of your stepped wedge trial?

Who is the symposium for?

Anyone interested in stepped wedge trials or study design more broadly. The meeting is open to internal and external people. Please feel free to invite people and to circulate this email to your networks.

Funding

It will cost us approximately £25 per person to put this event on, but the LSHTM Centre for Evaluation will cover the cost to maximise the number of people able to attend.

Registration

Please register for the event at https://stepped-wedge-symposium.eventbrite.co.uk.

Other exciting event that day

At lunchtime (1-2pm) on the same day there will be an exciting lecture entitled "De-worming reconsidered: re-analysis of an influential stepped-wedge trial with health and educational outcomes" by Alex Aiken and Calum Davey, also in the John Snow Lecture Theatre. Attendance at this lecture is encouraged for those interested in stepped wedge trials. It will be free to attend and no formal registration process is required.

Questions?

Please feel free to contact evaluation@lshtm.ac.uk or james.lewis@lshtm.ac.uk for any questions.

The mathematics of love

I can't remember how I came across this (I think I saw an article about it on Metro or something), but I got intrigued by Hannah Fry's work on "The mathematics of love". So I bought the book and read it $-$ it's a fairly easy read, so I got through it in just a couple of reading-sessions, while stuck on poorly ventilated trains...

Hannah is a lecturer at UCL, where she does serious work applying maths to understand complex phenomena such as outbreak investigations or the London riots. But (and I quite like this!) she has also done work in trying to explain something as complex and unexplainable as how people fall in love with maths $-$ at face value, this is of course ridiculous, but of course, she doesn't take things at face value and makes quite some interesting (and amusing) points (see the TED talk below).

What was also interesting is, I think, the characterisation she gives to her work $-$ which is to say from the perspective of a mathematician (which she is). That, I thought, was interesting as many of her points I would classify from the perspective of a statistician (which I am). I am not saying she's wrong, of course. But I don't think I'm either and it's just fascinating how you see the worlds with your own googles, sometimes... 

Our short course

We've opened officially registration for our short course on Bayesian methods in health economics (this is a link to last year's edition, with a little more information than the official webpage for this year's course). When we decided to do this, we agreed we'd have it rotating between UCL and the MRC Biostats Unit in Cambridge. This is where you can actually register.

The dates for this year are: Tuesday 24th - Thursday 26th November 2015 $-$ put that in your calendar!

The good, the bad and the ugly

This is me kind of whining $-$ although I do have some positive (and I think I'm right in whining).

First off: chapeau to the iHEA local organisers at Bocconi University in Milan! I think they've done an incredible job $-$ I think iHEA staff do help and get involved quite a lot in the organisation. But as far as I can tell, the whole thing went perfectly. In typical Italian style, the coffee and lunch breaks were really good, too (a nice change from previous iHEAs I went to).

Then the bad (here comes the whining)... As I mentioned here, this has been sort of a mixed conference experience, since I've not stayed in Milan for the whole time and have sort of commuted from Marta's parents' (they live on the lake). Anyway: yesterday I took the train to come back home after the afternoon sessions and I needed a ticket to the tram/metro to get to the train station. The first place I found on my way to the tram stop, told me that they had run out of tickets $-$ the man wasn't very friendly. "Go to the newsagent around the corner". So I went but as it turned out, they had run out too $-$ the man was even less friendly than the previous one. Third time lucky, I found another place who could sell me the €1.5 ticket, a bit further down the road $-$ by then I'd already walked a good 1/3 of the way.

Thanks to this, I just about manage to catch my train $-$ and here comes the ugly: firstly, the air conditioning wasn't working. Sitting next to me was an English family on holiday $-$ the little boy was wearing a Liverpool FC shirt which got so drenched with sweat I doubt he was able to take it off, last night... Then, when I got to my station and tried to get off the train, I simply couldn't because the door wasn't working. I did rush to the next carriage $-$ but that door didn't work either! Mid way my rush to the next-next carriage, the train left the station and I had to get off the next one $-$ luckily not very far from my stop.

Going to iHEA

iHEA's conference is kind of big deal in health economics: it's usually very big, with lots of sessions and lots of people participating. I have been to a few, both sides of the Atlantic and they are usually very good. This year it's in Milan (I think building on the Expo) and tomorrow I'm off to go. I'll be talking about the Structural Zero Costs model (not super-new $-$ I've mentioned this already here, here, here and here) in an interesting session on Tuesday.

It's always a bit weird, I think, going to conferences in Italy (or in London, for that matter) $-$ I think the spirit of being "away" at the conference kind of goes away. Still... I'm looking forward to some of the sessions!

Back to the future (or the day of the crises)

Yesterday was a very interesting day $-$ not sure if "interesting" is the best word to describe it, but for now I'll just use it...

We had our workshop on survival analysis in health economic evaluations, which instigated the first crisis of the day: Patricia Guyot, who was supposed to talk about digitising data from published studies, had to bail out $-$ this is somewhat similar to the Passport incident (except that that was entirely my fault, while this time, Patricia's train was delayed, which meant she missed her flight from Amsterdam). So, we had to reshuffle the order of the talks and make do with one less $-$ fortunately, both Nick and Chris had plenty of interesting things to say. In the end, the workshop was very good, I thought and I enjoyed it very much. The slides are available here.

On to the next crisis: when I got back to the office, there was a bunch of emails waiting for me to inform me that we may have a series of potential clashes with speakers not being able to give the talks they are supposed to, at the next ISBA World Meeting, at which our Section is organising a session (I will be talking about the RDD project). I think we managed to solve that crisis too $-$ luckily we had organised a session with 4 (instead of 3) speakers, so we could just lose one without impacting too much.

On to the next crisis: as I was finally riding my Vespa back home, I moved my phone from my jeans pocket to my jacket pocket. Or so I thought $-$ evidently it never made it to the jacket pocket and fell on the floor, without me noticing until I got home. So I've now been catapulted back to the 1980s with no phone or internet always with me... 

For example, I had to go to a meeting in a part of UCL where I'd never been, earlier today. But because I couldn't rely on a map on my phone, I had to print out a paper copy of all the references (name and phone number of the person I was supposed to meet, address, etc). Even though, spectacularly, I managed to lose the piece of paper with all this info in a matter of minutes (and I can't figure out how this may have happened!), I did get to the meeting on time.

Stress testing

Lately, we've been spending a lot of time "stress-testing" our method for the computation of the Expected Value of Partial Perfect Information (EVPPI $-$ I know: the terminology is a bit strange and possibly not-very helpful, as "perfect" information doesn't really exist, in statistical terms...).

I have mentioned this already here, here and here and the idea is to combine results from spatial statistics (and INLA) and Gaussian Process regression into the health economic problem. 

In a nutshell (I'll avoid all the technical details here), the "data" for our model consist on a vector of values $\mbox{NB}(\theta)$ (the "monetary net benefit", which determines the utility of a given health intervention) and a matrix of simulations from the (joint posterior) distributions of a set of relevant model parameters. The idea is that the multivariate parameter set can be split in a subset of "parameters of interest" ($\phi$), while the rest ($\psi$) are sort of "nuisance" or "unimportant" parameters. 

Following up on the work by Mark Strong et al, the relevant model can be written as
$$ \mbox{NB}(\theta) = g(\phi) + \varepsilon $$and the objective is to estimate the function $g(\phi)$, which is then used to compute the EVPPI. This saves up a huge computation time, in comparison to other methods.

In our model, we extended this framework and modelled
$$ \mbox{NB}(\theta) = H\beta + \omega + \varepsilon$$where $H\beta$ is a linear component depending on the simulations for all the "important" parameters, while $\omega$ is a spatially structured component, which accounts for the correlation among the important parameters. The big advantage is that using this formulation we are able to make inference based on INLA/SPDE, which is super-fast and can save up a lot of time even in comparison with the already fast "standard" Gaussian Process regression model.

Our first tests were giving very good results (as reported here). Then we've used a more complex model and found that while still being faster, our method was losing in accuracy for some specific parameters. This was a bummer, but it also meant that we had to go back and try and understand a bit better what was going on.

I'll make this sound very easy (when in fact Anna has spent a lot of time on this $-$ and wishing she never met me and started her PhD on this, I'm sure!), but eventually we figured out what the problem was. Firstly, in very complex situations (which are not that uncommon in real health economic evaluation problems), there may be quite a large correlation and non-linearity in the relationships among the parameters in $\phi$. This means that the combination of the standard linear predictor and the spatially structured effect cannot model properly the observed data, resulting in lower accuracy in the estimations. But, interestingly, extending $H\beta$ to include interactions among the relevant parameters can fix this problem, with only a small increase in computational cost.

Secondly (this is a bit more technical, but also quite interesting), the spatially structured component is based on constructing a mesh which describes the relationship between the parameters in a Euclidean space. If the "boundaries" of the resulting mesh are too close to the range of the observed points, then the estimation procedure will return several predictions at 0 $-$ in a rather vague sense, something like: the boundaries are areas where the estimated smooth curve is 0. But this means that in the computation of the EVPPI there is an artificially large number of 0 values, which produces an under-estimation of the "true" value.

We have fixed this by modifying the evppi function in the development version of BCEA and now the user can specify a non-linear part as well as fiddle with the INLA-specific parameters defining the mesh for the spatially structured component. The results seem to be much more accurate, still with some substantial computational savings. I'll try to put some R script to help people test the function (although I've also modified the help for evppi) to guide through the example involving the (simpler) Vaccine model