Sam Watson’s journal round-up for 11th February 2019

Every Monday our authors provide a round-up of some of the most recently published peer reviewed articles from the field. We don’t cover everything, or even what’s most important – just a few papers that have interested the author. Visit our Resources page for links to more journals or follow the HealthEconBot. If you’d like to write one of our weekly journal round-ups, get in touch.

Contest models highlight inherent inefficiencies of scientific funding competitions. PLoS Biology [PubMed] Published 2nd January 2019

If you work in research you will have no doubt thought to yourself at one point that you spend more time applying to do research than actually doing it. You can spend weeks working on (what you believe to be) a strong proposal only for it to fail against other strong bids. That time could have been spent collecting and analysing data. Indeed, the opportunity cost of writing extensive proposals can be very high. The question arises as to whether there is another method of allocating research funding that reduces this waste and inefficiency. This paper compares the proposal competition to a partial lottery. In this lottery system, proposals are short, and among those that meet some qualifying standard those that are funded are selected at random. This system has the benefit of not taking up too much time but has the cost of reducing the average scientific value of the winning proposals. The authors compare the two approaches using an economic model of contests, which takes into account factors like proposal strength, public benefits, benefits to the scientist like reputation and prestige, and scientific value. Ultimately they conclude that, when the number of awards is smaller than the number of proposals worthy of funding, the proposal competition is inescapably inefficient. It means that researchers have to invest heavily to get a good project funded, and even if it is good enough it may still not get funded. The stiffer the competition the more researchers have to work to win the award. And what little evidence there is suggests that the format of the application makes little difference to the amount of time spent by researchers on writing it. The lottery mechanism only requires the researcher to propose something that is good enough to get into the lottery. Far less time would therefore be devoted to writing it and more time spent on actual science. I’m all for it!

Preventability of early versus late hospital readmissions in a national cohort of general medicine patients. Annals of Internal Medicine [PubMed] Published 5th June 2018

Hospital quality is hard to judge. We’ve discussed on this blog before the pitfalls of using measures such as adjusted mortality differences for this purpose. Just because a hospital has higher than expected mortality does not mean those death could have been prevented with higher quality care. More thorough methods assess errors and preventable harm in care. Case note review studies have suggested as little as 5% of deaths might be preventable in England and Wales. Another paper we have covered previously suggests then that the predictive value of standardised mortality ratios for preventable deaths may be less than 10%.

Another commonly used metric is readmission rates. Poor care can mean patients have to return to the hospital. But again, the question remains as to how preventable these readmissions are. Indeed, there may also be substantial differences between those patients who are readmitted shortly after discharge and those for whom it may take a longer time. This article explores the preventability of early and late readmissions in ten hospitals in the US. It uses case note review and a number of reviewers to evaluate preventability. The headline figures are that 36% of early readmissions are considered preventable compared to 23% of late readmissions. Moreover, it was considered that the early readmissions were most likely to have been preventable at the hospital whereas for late readmissions, an outpatient clinic or the home would have had more impact. All in all, another paper which provides evidence to suggest crude, or even adjusted rates, are not good indicators of hospital quality.

Visualisation in Bayesian workflow. Journal of the Royal Statistical Society: Series A (Statistics in Society) [RePEc] Published 15th January 2019

This article stems from a broader programme of work from these authors on good “Bayesian workflow”. That is to say, if we’re taking a Bayesian approach to analysing data, what steps ought we to be taking to ensure our analyses are as robust and reliable as possible? I’ve been following this work for a while as this type of pragmatic advice is invaluable. I’ve often read empirical papers where the authors have chosen, say, a logistic regression model with covariates x, y, and z and reported the outcomes, but at no point ever justified why this particular model might be any good at all for these data or the research objective. The key steps of the workflow include, first, exploratory data analysis to help set up a model, and second, performing model checks before estimating model parameters. This latter step is important: one can generate data from a model and set of prior distributions, and if the data that this model generates looks nothing like what we would expect the real data to look like, then clearly the model is not very good. Following this, we should check whether our inference algorithm is doing its job, for example, are the MCMC chains converging? We can also conduct posterior predictive model checks. These have had their criticisms in the literature for using the same data to both estimate and check the model which could lead to the model generalising poorly to new data. Indeed in a recent paper of my own, posterior predictive checks showed poor fit of a model to my data and that a more complex alternative was better fitting. But other model fit statistics, which penalise numbers of parameters, led to the alternative conclusions. So the simpler model was preferred on the grounds that the more complex model was overfitting the data. So I would argue posterior predictive model checks are a sensible test to perform but must be interpreted carefully as one step among many. Finally, we can compare models using tools like cross-validation.

This article discusses the use of visualisation to aid in this workflow. They use the running example of building a model to estimate exposure to small particulate matter from air pollution across the world. Plots are produced for each of the steps and show just how bad some models can be and how we can refine our model step by step to arrive at a convincing analysis. I agree wholeheartedly with the authors when they write, “Visualization is probably the most important tool in an applied statistician’s toolbox and is an important complement to quantitative statistical procedures.”

Credits

Sam Watson’s journal round-up for 9th July 2018

Every Monday our authors provide a round-up of some of the most recently published peer reviewed articles from the field. We don’t cover everything, or even what’s most important – just a few papers that have interested the author. Visit our Resources page for links to more journals or follow the HealthEconBot. If you’d like to write one of our weekly journal round-ups, get in touch.

Evaluating the 2014 sugar-sweetened beverage tax in Chile: an observational study in urban areas. PLoS Medicine [PubMedPublished 3rd July 2018

Sugar taxes are one of the public health policy options currently in vogue. Countries including Mexico, the UK, South Africa, and Sri Lanka all have sugar taxes. The aim of such levies is to reduce demand for the most sugary drinks, or if the tax is absorbed on the supply side, which is rare, to encourage producers to reduce the sugar content of their drinks. One may also view it as a form of Pigouvian taxation to internalise the public health costs associated with obesity. Chile has long had an ad valorem tax on soft drinks fixed at 13%, but in 2014 decided to pursue a sugar tax approach. Drinks with more than 6.25g/100ml saw their tax rate rise to 18% and the tax on those below this threshold dropped to 10%. To understand what effect this change had, we would want to know three key things along the causal pathway from tax policy to sugar consumption: did people know about the tax change, did prices change, and did consumption behaviour change. On this latter point, we can consider both the overall volume of soft drinks and whether people substituted low sugar for high sugar beverages. Using the Kantar Worldpanel, a household panel survey of purchasing behaviour, this paper examines these questions.

Everyone in Chile was affected by the tax so there is no control group. We must rely on time series variation to identify the effect of the tax. Sometimes, looking at plots of the data reveals a clear step-change when an intervention is introduced (e.g. the plot in this post), not so in this paper. We therefore rely heavily on the results of the model for our inferences, and I have a couple of small gripes with it. First, the model captures household fixed effects, but no consideration is given to dynamic effects. Some households may be more or less likely to buy drinks, but their decisions are also likely to be affected by how much they’ve recently bought. Similarly, the errors may be correlated over time. Ignoring dynamic effects can lead to large biases. Second, the authors choose among different functional form specifications of time using Akaike Information Criterion (AIC). While AIC and the Bayesian Information Criterion (BIC) are often thought to be interchangeable, they are not; AIC estimates predictive performance on future data, while BIC estimates goodness of fit to the data. Thus, I would think BIC would be more appropriate. Additional results show the estimates are very sensitive to the choice of functional form by an order of magnitude and in sign. The authors estimate a fairly substantial decrease of around 22% in the volume of high sugar drinks purchased, but find evidence that the price paid changed very little (~1.5%) and there was little change in other drinks. While the analysis is generally careful and well thought out, I am not wholly convinced by the authors’ conclusions that “Our main estimates suggest a significant, sizeable reduction in the volume of high-tax soft drinks purchased.”

A Bayesian framework for health economic evaluation in studies with missing data. Health Economics [PubMedPublished 3rd July 2018

Missing data is a ubiquitous problem. I’ve never used a data set where no observations were missing and I doubt I’m alone. Despite its pervasiveness, it’s often only afforded an acknowledgement in the discussion or perhaps, in more complete analyses, something like multiple imputation will be used. Indeed, the majority of trials in the top medical journals don’t handle it correctly, if at all. The majority of the methods used for missing data in practice assume the data are ‘missing at random’ (MAR). One interpretation is that this means that, conditional on the observable variables, the probability of data being missing is independent of unobserved factors influencing the outcome. Another interpretation is that the distribution of the potentially missing data does not depend on whether they are actually missing. This interpretation comes from factorising the joint distribution of the outcome $Y$ and an indicator of whether the datum is observed $R$, along with some covariates $X$, into a conditional and marginal model: $f(Y,R|X) = f(Y|R,X)f(R|X)$, a so-called pattern mixture model. This contrasts with the ‘selection model’ approach: $f(Y,R|X) = f(R|Y,X)f(Y|X)$.

This paper considers a Bayesian approach using the pattern mixture model for missing data for health economic evaluation. Specifically, the authors specify a multivariate normal model for the data with an additional term in the mean if it is missing, i.e. the model of $f(Y|R,X)$. A model is not specified for $f(R|X)$. If it were then you would typically allow for correlation between the errors in this model and the main outcomes model. But, one could view the additional term in the outcomes model as some function of the error from the observation model somewhat akin to a control function. Instead, this article uses expert elicitation methods to generate a prior distribution for the unobserved terms in the outcomes model. While this is certainly a legitimate way forward in my eyes, I do wonder how specification of a full observation model would affect the results. The approach of this article is useful and they show that it works, and I don’t want to detract from that but, given the lack of literature on missing data in this area, I am curious to compare approaches including selection models. You could even add shared parameter models as an alternative, all of which are feasible. Perhaps an idea for a follow-up study. As a final point, the models run in WinBUGS, but regular readers will know I think Stan is the future for estimating Bayesian models, especially in light of the problems with MCMC we’ve discussed previously. So equivalent Stan code would have been a bonus.

This is an economics blog. But focusing solely on economics papers in these round-ups would mean missing out on some papers from related fields that may provide insight into our own work. Thus I present to you a politics and sociology paper. It is not my field and I can’t give a reliable appraisal of the methods, but the results are of interest. In the global fight against non-communicable diseases, there is a range of policy tools available to governments, including the sugar tax of the paper at the top. The WHO recommends a large number. However, there is ongoing debate about whether trade rules and agreements are used to undermine this public health legislation. One agreement, the Technical Barriers to Trade (TBT) Agreement that World Trade Organization (WTO) members all sign, states that members may not impose ‘unnecessary trade costs’ or barriers to trade, especially if the intended aim of the measure can be achieved without doing so. For example, Philip Morris cited a bilateral trade agreement when it sued the Australian government for introducing plain packaging claiming it violated the terms of trade. Philip Morris eventually lost but not after substantial costs were incurred. In another example, the Thai government were deterred from introducing a traffic light warning system for food after threats of a trade dispute from the US, which cited WTO rules. However, there was no clear evidence on the extent to which trade disputes have undermined public health measures.

This article presents results from a new database of all TBT WTO challenges. Between 1995 and 2016, 93 challenges were raised concerning food, beverage, and tobacco products, the number per year growing over time. The most frequent challenges were over labelling products and then restricted ingredients. The paper presents four case studies, including Indonesia delaying food labelling of fat, sugar, and salt after challenge by several members including the EU, and many members including the EU again and the US objecting to the size and colour of a red STOP sign that Chile wanted to put on products containing high sugar, fat, and salt.

We have previously discussed the politics and political economy around public health policy relating to e-cigarettes, among other things. Understanding the political economy of public health and phenomena like government failure can be as important as understanding markets and market failure in designing effective interventions.

Credits