What values do we need to keep in mind about the DALY? Part II

In the previous post in this series, I introduced the DALY and discussed two resolved issues: age-weighting and temporal discussion. In this post, I will look at the DALY and its role as a measure of disease burden that is transferable across different conditions and illnesses. I will also discuss the future of the DALY, its limitations as an indicator of resource allocation, and an additional method that could include community effects into the DALY.

The DALY as a transferrable measure of disease burden

The DALY is meant to transcend illnesses and causes. We should be able to study the circumstances in a region and estimate the total burden of disease, in DALYs, by aggregating the lives lost and disability inflicted by all diseases in that region. Furthermore, if, for example, diarrhoea is the foremost cause of DALYs lost in that region, and there are ways in which diarrhoea can be mitigated at a low cost, then we should naturally allocate resources to this cause. This is where the DALY is really powerful. It gives a very clear indication of where resources are put to their best use.

But what if I don’t agree with the disability weights used by the WHO? This is in fact a very real concern and something that has been debated since the conception of the DALY. How do we agree on disability weights for different health outcomes? In 1990 a panel of experts set all the disability weights. This was criticised and the 2010 update set disability weights using public surveys and expert consultation.

Still, suppose I have two interventions available. One distributes malaria nets and averts 0.2 DALYs per dollar. The other distributes antiretrovirals to people with HIV and averts 0.15 DALYs per dollar. Clearly, the net benefit of investing in the intervention that distributes malaria nets is higher.

However, I may reasonably argue that living with HIV/AIDS deserves a much higher disability weight than 0.547. This would change the cost-effectiveness estimate above. Maybe it would even make the DALYs averted higher than 0.2 per dollar. This example illustrates that we can never agree about the cost-effectiveness of different health interventions if we don’t agree on the relevant disability weights.

This point was stressed by Holden on the Give Well blog, who also cited a few other disagreements. Some people think certain conditions such as fistula worse than death, but many people disagree. Some people would rather bring a person to full health than save the life of an infant. Some people emphatically disagree.

Where does this leave us? Well, we shouldn’t forget that a lot of time and effort actually has gone into making the disability weights as representative as possible. I myself feel that I am in no position to question most of the disability weights worked out for the GBD 2010. So should we trust the experts and the public surveys? In most cases, I think we should. But this does not imply that we should give up the chance of critical review.

In fact, all the hard work put into the development of the disability weights and the DALY constitutes a tremendous resource. The hard work has, amongst other things, gone into assembling evidence on different health outcomes. Even if one doesn’t agree with the final disability weights, the information provided is still valuable.

For example, the latest update, GBD 2010, includes many non-communicable diseases that account for an increasing share of global DALYs. Diabetes, musculoskeletal disorders and mental and behavioural disorders is inflicting a much higher disease burden globally than in 1990, and will create new challenges to health care systems in the future. Only this result is very important, as it implies we should do more research into cost-effective interventions that combat these conditions. Since the scale of the GBD 2010 is so extensive, it has ample opportunity to find many issues that we need to put more focus on. In combination with efforts to make global health data more transparent, accessible and reliable, this can only be an asset to the cause of effective altruism.

Another concern I had with the DALY initially was that it would obscure important subtleties that vary across issues and regions. An analogy of this would be the credit rating of financial instruments prior to the US sub-prime crisis. The credit ratings didn’t account for the composition of these financial instruments. (Apologies for the analogy to finance). In the case of the DALY, it is equally important that we don’t forget that the DALY is not a measure of everything - it doesn’t measure the risk of contagion, for example. Averting 200 DALYs in community A may in fact be inferior to averting 150 DALYs in community B if community B is struck with a particularly contagious disease.

Naturally, no indicators account for all aspects of a situation. In many ways, it might not be a very serious concern, since rigorous analysis will always supplement the DALY with other methods and practises. Giving What We Can performs charity evaluations using a range of indicators and data - see for example our latest evaluation of Water, Sanitation & Hygiene. Nevertheless, one should take care and question cost-effectiveness statistics in the form of DALYs averted per dollar if they are referred to hastily. And don’t forget to think about the transferability of positive results – just because an intervention worked in one setting doesn’t mean it will work in another. A recent post covered some really interesting issues on the topic of results transferability[10].

The future of the DALY

We may reasonably expect DALY estimation to be an influential method in guiding resource allocation decisions in the future. But are there any other methodological problems that we need to consider?

When it comes to resource allocation there is one problem with the DALY that stands out. It was noted in this WHO bulletin from March 2014.

While we may determine disability weights for particular diseases and conditions, the welfare of the individuals suffering from these conditions will invariably depend on their own individual circumstances. For example, autism has a disability weight of 0.259. But we would expect the burden of autism to vary depending on income; country and place; and social attitudes. In fact, the burden of many conditions may vary so much as to make a universal disability weight meaningless.

A friend read my research for this post over my shoulder. He jokingly said: “What’s the disability weight for gluten intolerance?” (He is himself gluten intolerant but doesn’t suffer greatly from the condition since there are dietary alternatives available at a relatively low cost in Sweden). I immediately thought about how serious his gluten intolerance would be in a low-resource setting. Herein lies a challenge for the DALY. How do we account for circumstances such as income that changes the burden of a disease across individuals? How many more parameters do we include? Where to draw the line in this respect is a question for the future.

Another problem, which relates back to the social roles used to justify age-weighting, is community effects. Should we apply higher weights to contagious diseases that strike whole communities? Or are the disability weights attributed to such diseases high enough already? I think one could at least consider introducing some systematic (higher) weighting to diseases that are very contagious and thus disrupt social roles and groups.

For clarity, consider the following example. We have estimated the total disease burden for a large geographical area. Amongst the prevalent conditions are tuberculosis (disability weight: 0.331) and uncontrolled asthma (disability weight: 0.132). The total amount of YLDs (years of life lost due to disability = disability weight * total years lived with condition across population) caused by each condition in our geographical area is approximately the same. Remember that YLD is the second component of the DALY.

However, tuberculosis is prevalent in three communities only, whereas uncontrolled asthma is uniformly distributed across the entire population. Social roles in the small communities hit by tuberculosis are disrupting, since young children and young adults are particularly susceptible to tuberculosis. This systemic effect increases the burden of the disease both economically and socially.

Asthma, however, which is uniformly distributed, cannot be said to upset social roles in any systematic way. (I’m assuming that asthma is not generating any ostracization etc. - please take this as an idealised example). With reference to the burden of the diseases, it would then make sense to apply a weight to the YLDs in the communities struck by tuberculosis. This weight would not be applied to a lone victim of tuberculosis, but only to the three communities where the disease is disrupting social roles. Nor would the weight be applied to victims of asthma.

I think we must ask ourselves these systemic questions particularly if we want to use the DALY for resource allocation. To me, there is a value in a community, separate from the sum of the individuals, which should be relevant for the burden of a disease. This community effect should therefore also be relevant to cost-effectiveness analysis, and to the cause of effective altruism.

However, including such community effects would open up for the same kind of criticism that was levied against the age-weighting of the original DALY. For one, if we weight not only YLDs but DALYs according to community effects, then we are effectively going against the universalism of human life. Why? Because then the DALYs of 10 individuals living in a community would be weighted higher than the DALYs of 10 individuals living separately.

Community weighting could also introduce further problems - “should we give higher weights to certain communities which are of particular importance such as doctors?” “What is a community and where do you draw the line between communities and individuals?”. In fact, maybe it is easier to keep the DALY in its “pure” form and use other forms of analysis to assess community effects. However, this implies that reports such as the GBD 2010 might mislead future research and interventions by not acknowledging community effects. And community effects would at least capture my intuitive aversion of very contagious diseases.

Summary

The DALY is a fantastic effort to improve our understanding of the disease burden across causes and different regions of the world. But keep in mind:

• Your own personal values: do you give the authority to the people behind the GDB 2010 to set disability weights?
• The transferability of results: a success in one setting does not guarantee a success in another.
• Other systemic problems such as community effects.
• The disability weights are not adjusted to individual circumstances such as income and do not represent welfare: the burden of a disease may be much worse in one setting than another.