Brexit^{-1}

I've been asked to post about the EuroCIM (European Causal Inference Meeting), which will be held later this year in Florence. I very happily oblige, because: a) this is usually a very good conference; b) it is organised by nice and obviously very good people (well $-$ at least I like them!); c) at a time where everything UK seem to move away from anything Euro, it's actually very nice to see a conference formerly known as UKCIM going fully Euro!

EuroCIM: the European Causal Inference Meeting, April 2018, Florence

 We are pleased to announce that after five successful editions of the UK-CIM, *the first European Causal Inference Meeting (EuroCIM) will take place in Florence, Italy, in April 2018. *The meeting will be focused on “*Causal Inference in Health, Economic and Social Sciences*”. EuroCIM 2018 is organized by the Department of Statistics, Computer Science, Applications (DiSIA) and ARCO of the University of Florence, Italy. Conference dates are Wednesday April 11 to Friday April 13 2018, early bird January 17, Submission of Abstracts February 1. The conference will include keynote addresses from:

• Donald Rubin (Harvard University)

• Vanessa Didelez (University of Bremen) 

• Michael Hudgens (University of North Carolina)

Moreover, on April 10 2018 four workshops will be offered by Rhian Daniel (Cardiff University), Johannes Textor (Radboud University Medical Center), Fabrizia Mealli (University of Florence) and Guido Imbens (Stanford Graduate School of Business). The conference will also feature presentations and a poster section that will give researchers and practitioners the opportunity to show their work. For more info on the meeting, the fees, how to register and submit an abstract please visit: http://eurocim2018.arcolab.org/

Workshop on The Regression Discontinuity Design

As part of our bid to get an MRC grant (which we managed to do), we promised that, if successful, we'd also have a dissemination workshop, at the end of the project. Well, the project on the Regression Discontinuity Design (RDD) has now finished for a few months, but we're keeping our word and we have actually organised something that, as it happens, has probably turned into something slightly bigger (and better!) than intended...

As I was talking to Marcos (who's a co-director of our MSc programme, which I've mentioned for example here), we realised that the RDD is in fact a common interest of ours and so I jumped on his offer to do something together.

The plan is to have a full day on the 27th June at the Institute of Fiscal Studies in London, with the idea of mixing economists, statisticians and epidemiologists. We have a nice line up of speakers $-$ the original idea was more to show off the outputs of the project, but I think this works much better!

The registration is now open and free $-$ all the relevant information is here!

Face value

This is actually a not-so-recent paper, but I've only discovered now and I think it's very interesting. The underlying issue is about trying to do "causal inference" from observational data $-$ perhaps one could see this in a simpler way by considering the idea of "balancing" observational data, to mimic as far as possible an experimental setting (and so be able to estimate "causal" effects). [There's lot more on the philosophical aspects behind this problem, which I'm conveniently swiping under the carpet, here...]

Anyway, one of the most popular ways of dealing with this issue of unbalanced background covariates (or generally, confounding) is to use propensity score matching. But, while I think that the idea is somewhat neat and clearly important, what has always bothered me (among other things) is the fact that the resulting outcome model does assume that the estimate of the propensity score (PS) is "perfect" $-$ known with absolute precision, although the basic assumption is that "the PS model needs to be correct". But of course, there's no way of knowing perfectly that the PS model is correct...

So the idea of joining model selection and propagation of uncertainty through the outcome model is actually very interesting. I've only flipped through the paper and I did have some very preliminary ideas in mind on this, so I really want to have a proper look at this!

PhD opportunity!

Applications are invited for a PhD funding opportunity to conduct research in a branch of probability or statistics based in the UCL Department of Statistical Science, commencing in September 2017. This funding is provided by the Engineering and Physical Sciences Research Council (EPSRC).

The requirement for admission to the MPhil/PhD in Statistical Science is a 1st class or high upper 2nd class Bachelor’s degree, or a Master’s degree with merit or distinction, in Mathematics, Statistics, Computer Science, or a related quantitative discipline. Overseas qualifications of an equivalent standard are also acceptable. Further details can be found on the Departmental website. Applicants are expected to prepare an outline proposal of their work. We have some interesting project in our pipeline, including extensions of our work on survHE, or related to evidence synthesis and network meta-analysis, as well as the use of observational data for health economic evaluation.

The studentship will be four years in duration and covers tuition fees at the UK/EU rate plus a stipend of £16,785 per annum for eligible UK residents. EU nationals who have not been ordinarily resident in the UK for 3 years prior to the start of the studentship may still qualify for a fees only award. The studentship may only be awarded to applicants liable to pay tuition fees at the UK/EU rate (i.e. it cannot be used to part-cover overseas tuition fees). 


Further information, including details of how to apply, is available here.

My week at ISBA

I've spent the last few days in beautiful Sardinia for the ISBA world conference. The place is outstanding, really beautiful, although it's kind of weird that there is no real town along the cost for miles and miles. Leaving Cagliari and driving for over 50km, you only come across a massive oil refinery and the town of Pula. There are many resorts (some super high-end, like the one where the conference is, some more like very nice camping sites with tents or bungalows replaced by nice apartments), but no real town. I felt that was a bit weird as I wasn't used to things like that, in Italy $-$ but it may well be that I am just ignorant about my own country and there's tons of places like that...

As for the conference, my last ISBA was 10 years ago $-$ that was the last one of the original "Valencia meetings". Even then, it felt like a big conference, but that was nowhere near the level it has reached now (I think there are over 700 people here!). This means there are several parallel sessions and lots of heterogeneity in the topics. Also, the schedule is quite packed with talks from 9am to 7.30pm (with some breaks throughout) and then poster sessions after dinner. I was impressed by how well the organisation has worked: I've not seen a single session running late! 

My talk is later today; I'll be talking about our work on the RDD $-$ I've planned a rather high-level talk, showing some of the general themes we've developed and giving broad examples, rather than going to the details. Luckily, we do have a few papers on these so hopefully I'll be able to point people to the relevant references. 

More importantly, I'll have to rush off just after my talk (which isn't great) so that I won't miss my flight home $-$ I've tried to plan everything ahead, so I've put petrol in the car, checked out the hotel, etc. I'll give the talk with my fingers crossed, just in case...

A bunch of papers

The beginning of the new year has been particularly busy, as I'm working on several interesting projects. On the bright side, some of these are starting to give their fruits and, coincidentally, in the last few days we've had a few papers finalised (ie published, accepted for publication or submitted to the arxiv in an advanced status).

The first one has been published in Cost Effectiveness and Resource Allocation (the open access version is here). I've been involved in this paper with colleagues at UCL. The paper is an economic evaluation of an interesting and rather complex community trial conducted in Malawi, a country with particularly low life expectancy and high rates of HIV. In the paper, we did most of the economic analysis using BCEA.

The second one has also just been published in Pharmacoeconomics and is an "educational" piece that I co-wrote with several colleagues at UCL. I think this too was an interesting piece of work, in that we tried to focus on several statistical issues that are of concern in many economic evaluations $-$ the idea that health economics is in many ways inextricably related to statistics is of course one of my pieces de resistance (I guess I've showed off enough complicated words for a post...).

The third one is the RDD paper, which we had submitted ages ago to Statistics in Medicine. I had a brilliant experience with the Structural Zeros paper $-$ I submitted the first version in August and the paper was online by November. This time around, we had to struggle a lot more (apparently they couldn't find suitable reviewers, then the reviews arrived but took some time, then we responded to the comments $-$ long story short, it's been almost one year). Anyway, finally, they seem to have accepted the paper (which we previously arxived a similar version); we need a couple more changes and we should be good to go (I hope I'm not jinxing it!).

Finally, the last one is part of one of my PhD students (technically, I'm only second-supervising him). In fact, the paper develops a nice Bayesian non-parametric model to perform clustering and model selection simultaneously. We developed the model to handle a real clinical dataset, which records data on patients with lower urinary tract infection. I only knew little about Bayesian NP before working on this, so it was a nice opportunity. William has done a very good job in sorting this out and we have also submitted the full paper to Statistics in Medicine (hopefully, we'll get a quick turnaround!).

Best job ever

The job advert for the postdoc position in our MRC-funded project on the Regression Discontinuity Design is finally out.

Aidan has done a fantastic job in his little over a year in the position, but he's now moved to a lectureship in our department and so we need to find a suitable replacement. In fact, the new post has been extended and will be jointly funded by the project and the UCL department of Primary Care and Population Health $-$ who are collaborators on the RDD anyway.

The project is doing well and we do have a couple of interesting papers out $-$ here and here, for example. We're also currently working on some more extensions of the method, as well as the actual applications to the THIN dataset.


As they formally say, "Informal enquiries regarding the vacancy may be addressed to Dr Gianluca Baio, email: g.baio@ucl.ac.uk"...

B my J

As part of our work on the Regression Discontinuity Design for the British Journal of Medicine, we decided we should prepare a short, introductory research paper. We weren't holding our breath, as we thought that, while obviously interesting to clinicians, the topic may be a little too complex and technical for the BMJ audience. So we tried really hard to strip it out of the technicalities to highlight the substantial points $-$ which they liked! 

The paper was reviewed rather quickly and the comments were positive (although iI remember thinking that there was a sense of "you need more, but also much less" (which reminded me of Jeremy from Peep Show)... 

 

Anyway, they seem to have liked the revisions too and the paper is now out.

Causal Inference in Health, Economic and Social Sciences

The programme of the forthcoming UK Causal Inference Meeting "Causal Inference in Health, Economic and Social Sciences" is just out. The short conference will be at the end of the month (28th and 29th of April) at the University of Cambridge.

I indirectly feature as Aidan (who's part of our RDD team) is giving a presentation in one of the sessions. His talk is entitled "The Effect of Prior Beliefs on Causal Effect Estimators within a Bayesian Regression Discontinuity Design" and basically comes out as a follow up to our paper.

The idea is to implement the RDD within a full Bayesian framework and to actually assess carefully what's the gain of doing it this way. As is often the case, in some situations, there may be unwanted impact of the prior on the causal estimators, although generally there are advantages (both computationally and in terms of including extra sources of information $-$ that is crucial, especially when you want to go beyond statistical analysis, in my opinion).

Seven (a-day)

This week (among other things, including my Vespa breaking down twice in three days) I was busy taking part in an interview panel for a research associate position, together with colleagues in the Medical School at UCL.

One of the questions we were asking to the candidates was about commenting a new study (incidentally, by researchers at UCL) which using data from the Health Survey for England argued that the current "optimal" regime of consuming 5 portions of vegetables and fruit per day could (should) be in fact increased to at least 7, to reduce risk of death.

The point of the question was of course to get the candidates to recognise the possibility of confounding $-$ of course people consuming more veg & fruit might have a much lower risk of death to start with, due to different life-style, etc. (A few candidates got it right straight away, others less so).

But I think this has even more interestingly (eg, economic) implications in terms of the actual applicability of the health policy, in its current form as well as in terms of potential modifications, like this article in today's Guardian (which I thought was spot-on!) suggests. 

Man at work(-ish)

Perhaps one could argue that the obvious, manly activity to do at the weekend when you're home alone is to put and organise stuff in the garage. Well, I was home alone last weekend and my very own version of this was to arxiv the first paper coming out from our research on the regression discontinuity design (RDD) $-$ I know: probably *not* so manly. I did watch rugby and football, though.... 

The main of points of the papers are these:

1. How and why the RDD can be effectively applied to primary care data. The RDD works when there is some sort of external guideline that decides the allocation to some intervention $-$ drugs are often regulated so that patients with a certain profile should be given them (although, as we discuss, this is often a lot less clear cut...);
2. The implications of including genuine prior information in such an analysis. In our case study (prescription of statins), there's typically a lot of evidence coming from RCTs; and this may be the case in other areas where a recommendation exists to regulate prescriptions.

I think the plan is to explore next a few interesting (both methodological and substantial) matters, such as how this can be extended to non-continuous outcomes, or used to identify the "optimal" threshold for prescription, based on available primary care data (in addition to RCTs evidence).

The paper can be downloaded here.

More statins for everybody!

The National Institute for Health and Care Excellence (NICE) says in draft guidance which now goes out to consultation that the threshold for GPs to prescribe statins to their patients should halved from the current value of a 20% risk of cardiovascular disease. 

The current guideline has been in place for a few years now, but data from clinical practice seems to suggest that it is not strictly adhered to by GPs. For example, we've seen this in our Regression Discontinuity Design project (here's some preliminary analysis $-$ I believe we'll arxive a couple of papers on this shortly). 

The selection of the value of 20% 10-year risk as cut-off has been kind of controversial for some clinicians, since it was driven (also) by cost-effectiveness considerations. But at the time that the previous guideline was set, the prices of statins was much higher (it has since decreased due to the introduction of generics).

If after consultation, the guideline will be confirmed, this will probably imply a huge increase in the number of prescriptions filled in for statins.

New year, new cost-effectiveness thresholds?

Karl Claxton and colleagues at the University of York have recently published a working paper on Methods for the Estimation of the NICE Cost Effectiveness Threshold. Since a guideline was issued in 2004, NICE has used standard values of £20-30,000 per QALY as the official cost-effectiveness threshold. These are effectively equivalent to the cost per quality adjusted life year gained by investing in a new technology at the expenses of an already existing intervention. Decisions on reimbursements have been based on this decision rule (eg if the cost per QALY exceeded this range of thresholds, then the new intervention was not cost-effective).

This paper tries to produce an updated version of the threshold, based on empirical evidence (eg programme budgeting data for the English NHS). The methodology used is quite complex $-$ the full report is over 400 pages and I have only skimmed through it, reading with some care only some parts. Technically, the analysis is based on an instrumental variables approach within a structural equations setting and aims at simultaneously estimating the impact of the level of investment (and other variables) on health outcomes and the impact the overall budget constraint (and other variables) on the level of spending for a given health programme. Their main result is to suggest a slightly lower value to be used by NICE (£18,317 in some sort of baseline scenario). 

As I said, I only skimmed through the report, but I think it looks like a substantial piece of work. Nevertheless, I think there are some major limitation (which, to be fair, the authors acknowledge in the text. The Office for Health Economics has also produced a critique of this paper, which is available here).

The main one, seems to be the (lack of) availability of data for all the different programmes, to be used to translate the impact of expenditure into changes in quality of life). On the other hand, the paper tries to deal carefully with the issue of uncertainty propagation; for example, there is a whole section on the evaluation of structural and parametric uncertainty $-$ although this is not directly based on a full Bayesian model (which is kind of strange, given Karl is the main author on the paper...).

My talk at the LSHTM

Yesterday I gave a talk on our RDD project at the Centre for Statistical Methodology of the London School of Hygiene and Tropical Medicine. While presenting me, Karla (the organiser of the seminar) joked that I should go for a hat trick of presentations at the LSHTM, since only last month I gave another talk (on the structural zero problems in health economics $-$ on a related note, the paper, which I also discussed here, was actually accepted by Statistics in Medicine).

The main point of this talk was to try and point out various advantages of including genuine prior knowledge in the RDD framework, to try and get suitable estimates and make the assumptions underlying it more robust. I think we need to clarify a couple of points, but also I got good comments, so it was very helpful!

The slides of the talks are here.

Bad teacher(s)

This morning there has been some frenzy on the UK media (eg here or here) after the publication of a pamphlet by David Willetts, a junior minister for University and Science under the infamous coalition government.

The minister's point is that, comparatively to what used to be case in the past (notably in 1963 before changes in policy that led to increase in the number of university students), the proportion of time spent teaching by university lecturers is decreased in favour of the time that they spend otherwise.

Now, of course, this is not necessarily bad or good per se, but the minister says in his paper that "Looking back we will wonder how the higher education system was ever allowed to become so lopsided away from teaching." 

Well, one easy answer is of course to point out that apart from the huge increase in the number of students $-$ it would actually be interesting to have reasonable figures on the time spent teaching per-student, in comparison with pre-1963! $-$ the government(s) have switched the emphasis to research by decreasing the amount of funding available for universities and rewarding private initiative to obtain research money, eg from industry, or simply making the process of funding increasingly competitive!

Again, not necessarily a bad thing, but certainly not something to coolly swipe under the rug...

The (third) runway bride

I think I should disclaim the conflict of interest in this one (since Marta is one of the authors of the paper), but it was really, really cool to see her study on the impact on health of noise pollution close to airports in the newspapers today (for example here or here)! 

I thought that the Daily Mail would be also all over the news, while, interestingly, there's nothing on their homepage (although they do mention the article here).

I think the choice of pictures to accompany the articles is also quite interesting: The Guardian chose a rather romantic picture of an airplane taking off from Heathrow at dawn (or sunset $-$ I couldn't quite tell), while both the BBC and the Daily Mail had pictures of airplanes extremely close to properties or the ground (well, they were landing, after all...).

The original paper is linked here.

Parum PI

A couple of weeks ago, the MRC-funded research project on the Regression Discontinuity Design (of which I'm the Principal Investigator) has officially started, so I thought I wrote a few lines of update about it, after the couple of posts (here and here) referring to presentations we've given on (very preliminary) work we've done.

Unfortunately, I don't think I can quite claim to have the physique du role to be a Magnum PI [hence, and to show off that I did study Latin in school $-$ although you may argue that I could have easily used Google Translate... but I haven't: promised! $-$ the title of the post], if only for the fact that we're having a terrible September, weather-wise, here in London...

Let me be clear that I'm not complaining and of course I am very happy that we got the grant. But I must say that being PI is at the same time a very exciting and exhausting role (I'll pretend that it hasn't occurred to me that the project is just started). Today I put on my most Hawaiian shirt and spent most of the time trying to sort out a few admin things and sending emails. 

Hopefully, we'll shortly have something a bit more substantial to report about $-$ the signs are all there, luckily...

Biostatistics seminar

As part of the activities of the UCL Biostatistics Network, we organise regular seminars, to which we invite (usually relatively local $-$ for budget reasons only!) speakers. 

September is the start of the new term, and we'll kick off with what (in my opinion) is a very interesting topic. Alexina Mason (Imperial College) will speak about full Bayesian methods to deal with missing data. The seminar will be on September 18th at 4pm in the department of Statistical Science. 

Title: A general strategy for dealing with missing data using Bayesian methods
Abstract: Bayesian full probability modelling provides a flexible approach for analysing data with missing values, and offers an alternative to standard multiple imputation.  Plausible models allowing for missing responses and/or missing covariates can be built, which incorporate realistic assumptions about the missingness mechanism.  Additionally, the Bayesian approach lends itself naturally to sensitivity analysis, which is crucial when the missingness mechanism is unknown.  These strengths will be demonstrated by presenting a general strategy for a "statistically principled" investigation of data which contain missing values. The first part of this strategy entails constructing a "base model" by selecting an analysis model, then adding a sub-model to impute the missing covariates followed by a sub-model to allow informative missingness in the response.  The second part involves running a series of sensitivity analyses to check the robustness of the conclusions.  An antidepressant trial comparing the effects of three treatments will be used as an illustrative example throughout.  In particular, we will focus on missing responses assuming a non-ignorable missingness mechanism. 

JSM 2013

Next week I'll head to the Joint Statistical Meeting (the annual conference of the American Statistical Association), which funnily enough this year will be held in beautiful Montreal, Canada.

I've been once to Montreal for a couple of days and, despite the fact it was very, very cold (something like 3-4 degrees, despite the fact it was mid May!), we loved it. 

I'll give a talk on the Regression Discontinuity Design in a very interesting session organised by Fabrizia Mealli. Not for the first time this year, I'm doing reasonably well with preparing the slides (not finished yet, but getting there)...

Today I've flipped through the programme and it looks as though there are quite a few very interesting sessions. In fact, the programme looks more interesting than the one last year $-$ it wasn't too bad, but the few talks I really wanted to see where always clashing. For this year, there is only a (massive) clash on Tuesday afternoon, with a session on spatial stats (featuring Håvard), a session on causal inference (including Guido Imbens) and Judea Pearl's masterclass on the mathematics of causality all scheduled at the same time! I'm sure that organising the timetable for such a big event must be a nightmare, but this is just unfortunate...

Daily bias in the mail

David Spiegelhalter writes in his blog about this news headline on the Daily Mail (DM)'s website. According to the article, statins can weaken muscle and joints, raising the problem by up to 20%. As David points out, this claim is not justified by the evidence in the original article, since the DM is actually mixing up the meaning and interpretation of odds ratio and relative risk. 

So, in truth, the actual absolute difference in the risk of having muscle and joints problems is merely 2% (extra risk for those using statins). And even then, all sorts of additional problems arise (eg, again as pointed out by David, the fact that individuals taking statins are more likely to be under constant observation and information about their overall health status more likely to be recorded).

But what's also interesting, I think, is the confirmation bias in most of the comments to the article on the DM's website. Many of those who bothered: a) reading the article; and b) commenting (which of course in itself does not make for a very representative sample!), confirmed that they too had problems which by now they can assert with absolute confidence are due to consumption of statins!