Lot of people are trying to extend their lifespan. This is of less value if you get a disease. Especially one in which you suffer a lot, where the negative utility from getting sick is worse than positive utility from average day in your life. The negative utility could have a large absolute value, since bad is stronger than good. Therefore, you should try to minimize the probability of getting such an extreme disease.
What are the extreme diseases and how can we avoid getting them? The thing we are looking for is:
- High-suffering: enough to make the utility negative.
- Preventable: we have a prior probability p_1 of getting a disease and after preventive measures there is posterior probability p_2 of getting a disease, we are looking for diseases with high p_delta = p_1 – p_2.
- To be more specific, we are looking for highest suffering * p_delta, which is the expected suffering reduction of taking preventive measures.
Is there any data on which diseases are the worst? Disability-adjusted life year (DALY) measures the burden of disease. Let’s say you have a disease which takes away 10 years of life, that would be years of life lost (YLL) due to dying early. The same disease lasts for 10 years before you die and has a disability weight of 0.2, which gives you 10 * 0.2 = 2 years lost due to disability (YLD). DALY = YLL + YLD = 10 + 2 = 12.
The more years of life you lose, and the more severe the disease (higher disability weight), the higher the burden of disease – more DALY is bad, less is good. The disability weights go from 0 to 1, which means in the worst case, having some disease and living is the same as being dead. The obvious shortcoming of this is: what if you have a disease which makes your utility negative? Living longer with an extreme disease would bring less DALY, which means less burden, while it should increase DALY even more. The weights should be able to go above 1.
Looking at a post by Sindy Li we see DALY is not a measure of suffering or loss of quality of life, but a measure of lost health. It is estimated by presenting people with two hypothetical individuals in different health states (described briefly in lay language), and asking which person they regarded as healthier. (the pairwise comparison method) The accuracy of descriptions is questionable, as is the respondents’ ability to understand the implications of various health states. The other method is to ask people to think in the shoes of a decision maker who has to choose between policies that involve trade-off between severity of illness, the size of the health gain and the number of people helped. (the person trade-off method) This explicitly captures societal judgements (e.g. stigma of certain diseases), which we ideally don’t want to include. In general, with both methods, the non-health aspects (e.g. effect of income loss due to reduced productivity) are not taken into account – but they should be. Not enough dimensions of experience were taken into account.
Alternatives to DALY would be using QALY or EQ-5D, but I can’t find the table of that metrics for particular diseases online. There is a correlation between DALY and EQ-5D-5L which this study measured as PCC = 0.83. The EQ-5D asks questions about how the disease affects these 5 dimensions: mobility, self-care, usual activities, pain, anxiety/depression. If you have several conditions which are at maximum value on some dimension, one of them could be a lot higher on that dimension still, but that information is lost due to “ceiling effects”. If you can’t imagine what the strongest pain you could feel really feels like, that is equivalent to there being an implicit ceiling on your evaluation of pain where all pain strong as the strongest you can imagine gets the maximum evaluation for pain strength. This mechanism is active when evaluating mental disorders since a healthy person cannot, for example, imagine how it’s like to be schizophrenic.
Since the number of diseases considered in the DALY table is vastly smaller than the total number of diseases there probably there are some rare high-suffering diseases out there which did not get a mention. Of all of the diseases which did not get a mention, let us look to the one with highest prevalence (i.e. prior probability), the one which almost made it to the list. If the suffering of having that disease is high enough, and it is preventable enough, we may be missing our highest priority completely.
It would be useful to have data about:
- Better estimates on disability weights. No ideas how to get that, except conducting a better study than currently exists.
- How long do diseases last?
- What is the prior probability of getting each disease?
- How preventable are they?
One more problem is, diseases have various sub-types and they affect different people differently, so for each “high-level disease” (e.g. “cancer”, we are not going into sub-types here) we are actually dealing with a probability distribution of suffering. This is important if the amount of suffering goes up exponentially with the level of disability. Let’s break it up into three categories: light, medium, heavy, with light suffering 10, medium 40, heavy 160. A disease having 10% of being light, 80% being medium and 10% being heavy is not the same as having probabilties 20% light, 60% medium, 20% heavy. The first one gives 1 + 32 + 16 = 49 of expected suffering, the other one 2 + 24 + 32 = 58.
The less data we have the more we need to rely on priors. Having no evidence for intensity of suffering, we are dealing with a distribution of possible levels of suffering. This distribution is wider for mental disorders because they are further removed from our everyday experience, i.e. we have less information about how it is like to be in those states than to suffer pain.
Top 10 DALY diseases (actually, top 13 because the last 5 have the same weight) are:
- Schizophrenia, acute state
- Spinal cord lesion at neck level (untreated)
- Multiple sclerosis, severe
- Heroin and other opioid dependence
- Major depressive disorder, severe episode
- Traumatic brain injury, long-term consequences, severe (with or without treatment)
- Spinal cord lesion below neck level (untreated)
- Spinal cord lesion at neck level (treated)
- Chronic ischemic stroke severity level 5
- Acute ischemic stroke severity level 5
- Chronic hemorrhagic stroke severity level 5
- Acute hemorrhagic stroke severity level 5
- Schizophrenia residual state
Let’s simplify it:
- Schizophrenia, lifetime prevalence about 0.5% or 50 in 10,000.
- Spinal cord lesion, about 10 in 10,000.
- Multiple sclerosis, about 0.03% or 3 in 10,000.
- Heroin and opoids dependence, about 15 in 10,000 in a given year.
- Depression, about 670 per 10,000 in a given year.
- Stroke, harder to estimate but 795,000 in the US suffer a stroke each year, 75% of them being over 65, with half of them getting a disability, rough order of magnitude lifetime prevalence estimate: about 1% or 100 per 10,000. But what DALY estimate is talking about is “stroke severity level 5”, with no information about the probability distribution of severity, that would be 20 per 10,000.
Seems like schizophrenia and depression win here. They win despite being systematically underestimated. To recap, they are underestimated in three ways:
- Not enough dimensions of experience taken into account
- Ceiling effects on dimensions which are taken into account
- Wider probability distributions for suffering (on the assumption that with more disability suffering increases in a superlinear way)
Let’s take schizophrenia, risk factors which you can control are:
- Avoiding cannabis (increases risk twofold) and substance abuse in general
- Avoiding psychedelics. There is mixed evidence on psychedelics but there is a thing called LSD Psychosis and “taking of drugs might play some precipitating role in the onset of schizophrenia, bringing this disorder on more quickly” [source]. So we have a thing which puts people in mind-states which resemble schizophrenia, can cause psychosis, can trigger schizophrenia… Without good population studies it seems to me it probably can sometimes cause schizophrenia in people who would not otherwise experience it.
- Not sure if preventing autoimmune diseases is possible, but doing so would lower risk of schizophrenia.
Risk from depression can be lowered by:
- avoiding alcohol and other substance abuse
- eating healthy (omega-3, vitamin D and B complex look promising)
- regular exercise
- mindfulness meditation, which reduces risk from mental disorders in general.
Risks and negative effects of mental disease are underestimated (for more details you can see the already mentioned post by Sindy Li) and they come up on top despite being underestimated. In line with that, the main reason you should exercise regularly is for mental, not cardiovascular, health – in order to prevent mental disorders. Mindfulness meditation is important for health. Vitamin D, B complex and omega-3 supplements probably help. The other advice is the same old: eat your veggies, exercise, don’t take drugs, don’t smoke, don’t drink. Our grandmothers may have been wrong about lot of things but it turns out they got a lot of their health advice right.
More research is, of course, needed. Certainly more than a short blog post. What we should ideally have is a table with estimates of expected suffering reduction of taking preventive measures, by disease.