Sam Watson’s journal round-up for 12th March 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.

[While the journal round-up is not on strike this week, academics and other university staff across the UK continue to be. Please support these staff members and the university sector that produces much of the great research we feature on this blog.]

How does household income affect child personality traits and behaviors? American Economic Review [RePEcPublished February 2018

The intergenerational transmission of poor social and health outcomes and its remediation has long been of concern to policy makers and economists alike. A popular hypothesis to explain this phenomenon is that of fetal origins: the nine months in the womb are perhaps the most important in determining a person’s health over their lifetime. We have featured numerous papers on this blog looking at the impact of in utero conditions on infant, child, and adult outcomes. This hypothesis though leaves a sense of pessimism since if this generational link is rooted in biology then it is not likely to be modifiable by any intervention. Studies of institutional interventions in schools and the health care system have shown that the health of  children from impoverished households can be improved. But what about the effects of simply improving the material conditions of those households? Would this have an effect? This study uses a longitudinal dataset of children in North Carolina, USA which oversampled children from Native American families who, in the middle of the period of observation, began to receive an unconditional cash transfer from the tribal government funded by casino revenues. A difference-in-difference-in-differences model is used with the relevant differences being: before v. after, younger cohorts v. older cohorts (older children’s households did not receive the cash while they were children), and Native American v non-Native American. An ‘event study analysis’ is also used, which takes into account time from the intervention. (This is the exact same method as another recently featured paper on this blog – perhaps sign of the growing popularity of such techniques). Average annual income increased by around $3,500 per year. Quite clear improvements in a range of psychological traits are estimated from the models including increases in conscientiousness and agreeableness, and declines in emotional and behavioural disorders. Potential mediating mechanisms for these changes are explored and uncertain evidence is shown indicating improved parental supervision and interaction and a reduction in parental mental health care seeking (they plot 90% confidence intervals which appear  ‘statistically significant’ where 95% confidence intervals clearly would not be – however, the lack of significance stars and p-values is refreshing). Such evidence should weigh heavily on policy makers’ minds when implementing reductions to social assistance programs and household income.

Adaptation or recovery after health shocks? Evidence using subjective and objective health measures. Health Economics. [PubMedPublished March 2018.

Hedonic adaptation is a well evidenced phenomenon in health economics and related fields. Individuals can get used to health conditions and adverse circumstances, such as amputation or blindness, and recover much of their pre-illness quality of life. This makes it hard for healthy people to judge the quality of life of these conditions and is one of the reasons for the divergence in preferences over health states depending on who you ask. This paper takes an interesting approach to looking at adaptation by asking whether the improvement in someone’s subjective assessment of their own life expectancy after a serious illness is reflective of actual recovery or is in fact due to the optimism brought on by adaptation. Typically, beliefs about life expectancy are found to accord well with actuarial assessments of life expectancy, but little is known about how this relates to serious illness. This study suggests that subjective assessments of mortality risk do drop with cancer, stroke, and myocardial infarction in line with changes to objective risk of death. However, these subjective assessments generally return to their pre-illness levels, which doesn’t reflect the continued increase in risk actually faced by these people. An explanation for this is hedonic adaptation – people perhaps end up feeling as well as they did before even if they are not. It’s hard to say though if there’s a survivorship bias in favour of the optimists.

The local influence of pioneer investigators on technology adoption: Evidence from new cancer drugs. Review of Economics and Statistics. [RePEcPublished March 2018.

Technology diffusion typically shows a strong spatial pattern. If you know someone who has adopted a new technology, you are more likely to do the same yourself. But what about in medicine – do doctors also adopt similar patterns of prescribing new drugs? In the UK, we might think such patterns are unlikely as doctors are not free to prescribe what they like since they are restricted generally to what the NHS will reimburse. New technologies have to be first approved on the basis of being demonstrably cost-effective. But in the United States doctors are freer to prescribe what they like. While this has benefits, it also leads to adoption of cost-ineffective interventions or persistence in prescribing sub-optimal treatments. If the diffusion of new treatments is based upon social and professional spatial networks then one might expect the epicentre to be where the drug was trialled, the PI may well also be the loudest cheerleader for the new drug should it be shown to be effective. Indeed if a ‘superstar’ researcher is involved with the development of a drug this may attract more attention to it still. The key finding from this study in the US is that patients treated in the hospital market where the first author of the paper reporting the results of the main clinical trial of a drug were 36% more likely to receive the drug than elsewhere in the first two years. This is generally beneficial to patients in those areas, particularly since the average survival benefit to those patients is larger than is attributable to the drug itself, which may suggest that doctors with local information are better at selecting which patients will benefit from a treatment. However, with some of the problems arising from reporting bias, p-values, and the like patients may also be getting a worse deal should the drug not be as good as claimed.


Sam Watson’s journal round-up for 1st May 2017

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.

Systematic review of health economic impact evaluations of risk prediction models: stop developing, start evaluating. Value in Health [PubMed] Published April 2017

Risk prediction models are pervasive in clinical medicine. For example, one 2012 review of type 2 diabetes (T2DM) models identified 16 studies with 25 models. There was not much difference between the models in ability to predict T2DM and models including biomarkers were slightly better. But, obviously no model is perfect, the T2DM risk prediction tools generally overestimated the risk of development of diabetes. One could see parallels here with screening. When subjected to cost-benefit analyses, many screening programs become somewhat controversial. False positives can cause harm to patients both psychologically and through further procedures they may be subjected to. Such concerns thus may also apply to risk prediction models. This review surveys the literature on health economic evaluations of risk prediction models. Forty studies examining 60 risk models were included. Compare this number with the total of T2DM models above and you will see how the authors might arrive at the conclusion that economic evaluations of risk prediction models are rare. Another key finding, and one I empathize with as I am currently reviewing economic evaluations in another area of heath economics, is that there is a large amount of methodological heterogeneity and quality differences between studies. This makes comparisons difficult if not impossible. This limits the utility of these findings to decision makers. A routine, standardised approach to economic evaluation is needed.

The fading American dream: trends in absolute income mobility since 1940. Science [PubMed] [RePEc] Published 28th April 2017

This one is not strictly health. But it’s findings may have important implications for how we understand the relationship between income and health, and the inter-generational transmission of health. And, it’s not everyday an economics paper gets into Science. Economic mobility is a key goal for many societies – children should earn more than their parents. One way of examining this quantitatively is the proportion of children who earn more than their parents. This paper shows that this can be estimated using (i) the marginal income distribution of children, (ii) the marginal income distribution of parents, and (iii) the joint distribution of child and parent income ranks. The key finding is that mobility has declined over the 20th Century. While around 90% of children were earning more than their parents in 1940, by 1980 this is only around 40%. The authors look at what would happen to these estimates if GDP growth were more equally distributed and find much of the decline in mobility would be reversed.

Economic consequences of legal and illegal drugs: the case of social costs in Belgium. International Journal of Drug Policy [PubMed] Published 23rd April 2017

Put ten economists in a room and you’ll get 11 different opinions. Or so the saying goes. But while there is division on a number of topics in economics, some issues find a strong consensus. Drug prohibition is one of those issues many economists agree on. As a policy is has high costs and reasonably little benefit, especially when harm reduction is the goal. David Nutt, whose work we’ve discussed before, is a prominent critic of the UK government’s policy on drugs. Just this week he has discussed how the recent increase in the use of and health problems due to ‘spice’ (synthetic cannabinoids) may well be attributable to the prohibition of natural cannabis. However, recreational drug use, whether illegal or legal, does bear a societal cost. This paper attempts to quantify both the indirect and direct costs of drug use in Belgium. They take a ‘cost of illness’ approach, a term I think is a little unsuitable for the topic – most drug use causes no harm so could hardly be called illness. They also refer to the drugs as ‘addictive substances’, which is also a stretch for what they consider. Costs are further divided into health care and crime costs. The headline finding is that the total cost is 4.6 billion Euros annually. Interestingly, for illegal drugs, law enforcement expenditure was higher than the health care costs. In my mind this further undermines a prohibition policy. However, I think this study reveals the difficulty of taking an objective stance on these matters. Recreational substance use is an ‘illness’ and ‘addictive’ and bears a cost to society – the word ‘benefit’ is mentioned only once.

New metrics for economic evaluation in the presence of heterogeneity: focusing on evaluating policy alternatives rather than treatment alternatives. Medical Decision Making [PubMed] Published 25th April 2017

Cost-effectiveness analyses (CEA) are a key aspect of the evaluation of medical technologies and pharmaceutical products. Typically, the main output of these analyses is an incremental cost-effectiveness ratio (ICER) or other summary measure of incremental costs and benefits. However, these ICERs typically use an average treatment effect and complete adoption. This is unlikely to be realistic, though, from a policy perspective. Both effectiveness and adoption rates may differ between sub-groups. This paper proposes a ‘policy’ framework that takes this heterogeneity into account. In essence, the paper advocates a weighted average ICER taking into account adoption rates and heterogeneous effectiveness. It takes this idea a step further and considers uncertainty about all the parameters. Conceptually, the framework is a straightforward extension of CEA, but the paper is clear and lucid and it certainly makes sense to evaluate technologies on the basis of how they will actually be used. Similar ideas have been used to take forward clinical trial design: with more information patients will make different treatment choices, for example. The trouble is, innovative and sensible ideas can be very slow to catch on.


Chris Sampson’s journal round-up for 27th June 2016

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.

A methodological review of US budget-impact models for new drugs. PharmacoEconomics [PubMed] Published 22nd June 2016

Budget-impact analysis is a necessary step in the decision-making process. In the UK, NICE make recommendations on the basis of cost-effectiveness (mainly) and facilitate regional budget-impact estimates using a costing template. Guidelines are available from a whole host of HTA agencies and other organisations. This study reviews the methods used in US-based studies of new drugs. The authors identified 7 key elements to consider in the design of budget-impact models: i) model structure, ii) population size and characteristics, iii) time horizon, iv) treatment mix, v) treatment costs, vi) disease-related costs and vii) uncertainty analysis. Papers identified in a literature review were divided into those for drugs for acute conditions (n=8) and chronic conditions (n=27) and studies that combined budget-impact and cost-effectiveness analyses for any kind of drug (n=10). Each paper is summarised in terms of the 7 key elements. The methods adopted by the reviewed studies were not consistent with recommendations. For example, many studies omitted adverse event costs and a 1-year time horizon was often adopted where it may not be sufficient. Combined budget-impact and cost-effectiveness models are not recommended, on the basis that this adds unnecessary complexity. Generally, the authors support the use of costing models with simple structures and advise the use of a cost-calculator approach wherever possible. A neat table is provided which sets out recommendations and common flaws in relation to the key elements.

Why do health economists promote technology adoption rather than the search for efficiency? A proposal for a change in our approach to economic evaluation in health care. Medical Decision Making [PubMed] Published 17th June 2016

It seems like the wrong question. Health economists don’t really decide what to research, research funding bodies do. It is difficult for a researcher to find the time to research something without any funding. So surely the blame lies with the NIHR et al? The paper starts by explaining why low-value care exists, before outlining two ways in which we health economists might appropriately realign economic evaluation towards the search for efficiency. First, ‘technology management’. This is the idea that evidence should be evaluated throughout a technology’s life-cycle. The authors discuss examples from diabetic retinopathy screening and gastrointestinal endoscopy. I think they are flawed examples as they don’t relate to disinvestment per se, but I’ll set that aside for now. The second idea is ‘pathway management’. This is akin to whole disease modelling. The authors present an illustrative example for the ways in which this might be used to ‘search for efficiency’. The authors then go on to discuss the promise and challenges associated with their suggestions and outline some things that we ought to be thinking about. Maybe research groups need reorganising along clinical lines. Certainly, we need to figure out how to deal with intellectual property associated with whole disease models. But it still seems like the wrong question to me, and that health economists don’t have that much sway. Broadly speaking, so long as we’re paid to evaluate technology adoption we will be evaluating technology adoption.

Using survival analysis to improve estimates of life year gains in policy evaluations. Medical Decision Making [PubMed] Published 16th June 2016

Evaluation of policies in terms of their cost-effectiveness is increasingly possible. Often, analyses of this kind extrapolate survival of both the intervention and the control group based on life expectancy estimates from the general population. It’s unlikely that people affected by a policy under evaluation will be completely representative of the wider population. Policies are often also evaluated on the basis of near-term mortality, despite the possibility for them to have longer-term impacts. This study explores the potential for using parametric survival models to extrapolate outcomes for policy evaluations, as is often done for clinical trials. As an example, the authors used their previously published evaluation of the Advancing Quality pay-for-performance programme. Three methods are compared: i) application of published life expectancy tariffs, ii) incorporation of short-term observed survival and iii) extrapolation using survival models. The third approach used two separate models: one for short-term post-hospitalisation survival and another for long-term survival that excluded the first 30 days after admission. For the evaluation of the AQ programme, the three methods found increases in life expectancy of i) 0.154, ii) 0.221 and iii) 0.380. This demonstrates the importance both of incorporating observed mortality rates using survival analysis and of using all available data.