Health Inequalities in Autism – Part 1

This is part of a series of posts on the challenges of accessing healthcare for people who are autistic. In this we will discuss the current evidence about health inequalities found in the autistic population, the impacts of inadequate access to healthcare, and what can be done to help people to access care at the right time.

“Autistic people die on average 16 years earlier than the general population”. Let that one sink in for a minute. Sixteen years. I read that the first time and thought about how much occurs in sixteen years. Looking at the start of life sixteen years means birth through to late teenage years, a time where so much development occurs. Then looking from the other extent, when someone is dying sixteen years can mean a lot.

When I worked as a doctor, one of my jobs was in palliative care where I was working with patients approaching the end of their life. Unsurprisingly, there were three different views on time in regards to death. There were people who wanted more time, who would always want more time with the things that were important to them. There were people who felt it was the right time, who had accepted their situation. Lastly, there were people who wanted to die now, who wanted it to be over now so that they could finally be peaceful. In reality though, most people wanted more time – they wanted more of the times when they were healthy and happy. Time when they could run after their children, have sleepy mornings in bed with their partner, and discover more of their interests.

Sixteen years more time of health has so much meaning to it, meaning for not just the individual and their loved ones but society as a whole through the impact of sick leave and unemployment. People matter, and so a number like this is simply put alarming.

This number gets worse when you add in people with autism and learning disabilities, where it becomes greater than thirty years. The confidential enquiry into premature deaths of people with learning disabilities had previously revealed a median mortality of thirteen to twenty years earlier than the general population, but adding in a diagnosis of autism increases this already concerning figure.

But where is this data coming from? Should we trust it? It’s time for one of my favourite games – critical analysis of a journal article! WOOOOO! (I’m going to try and explain my process in analysing this article but I may not explain it in the most simple way. If there is anything that I need to explain more please let me know in the comments and I’ll explain them better in future articles).

The main article everyone is talking about for this is by Hirvikoski, et al. titled “Premature mortality in autism spectrum disorder” published in 2016 in the British Journal of Psychiatry. The British Journal of Psychiatry had an impact factor of 7.06 in 2015 (this means that for every article published in it, on average 7 people would reference the article in another scientific publication. This is a pretty good value all things said and can be interpreted as the journal articles generally talking about interesting topics and being of a high enough standard to be referenced in others work).

The study is a matched case-control study. This means that they took two populations of people, one group with autism and one group without that are ‘matched’ for specific characteristics (in the case of this article: birth year, gender and country of residence during the time that the people in the other group were diagnosed with autism). From these two populations they then look back at all of the data see how many people have died in each group. From those people they then look at what they did of and report it.

This is where we bump into a problem that is difficult to solve. A case-control study has limitations due to confounding variables. This means that in all of this study we can talk about correlations (whether one factor is linked to another) but not about causation (how the factors are linked to each other). This is fine when looking at mortality rates between the two populations, as you can say if one population has a higher mortality rate than another. But you cannot draw any firm conclusions as to why this is – a way of looking at this is that this article talks about a link between epilepsy and autism, saying that more people who died with autism had epilepsy than the general population. However, you can’t say that autistic people die earlier because of epilepsy, as the study does not look at all of the factors that may show why autistic people die earlier (the confounding variables).

While this limits the study, it’s difficult to design a study to look at this in any other way. A method that could account for more of the confounding variables is a cohort study, where you take a population of people with autism and a population of people without and follow them up over many years to see what they die of and all of the other factors in their life leading up to death. However, this takes a lot of time and you would still need to have an idea of what the confounding variables could be before going into the study as you’d need to follow these up to see if they have any influence. Therefore, a case-control study is probably the best approach at this time for this study type. They also attempted to match the data for a few categories, which is a good thing to do. They could have potentially matched for more factors that would have provided some more information and removed some confounders (ex. socioeconomic background, ethnicity, educational level) but this may not have been possible with the dataset that they had.

The population of the study is of a sufficient size to give a lot of information, having 27,112 autistic participant and 2,672,185 non-autistic participants. This is important as it tells you that these factors are more likely to be replicable across the population as there are a larger number of participants. They further separated their autistic population into people with high functioning and low functioning autism (through the use of diagnostic coding distinguishing the level of ‘mental retardation’ in the individuals – which is really naff terminology, but is used a fair amount in research studies). Therefore, if someone had a comorbid learning disability of any sort then they would fall into the low functioning category (which isn’t true in reality, but the idea of what high functioning really means is a topic for another day).

Graph 1

Graph 1: Shows the number of people with ASD involved in the study. This further splits it down into their gender and whether they are classed as high or low functioning.

The other thing to note is that the population used is from Sweden, so may not be applicable to other countries. Wikipedia (‘a highly trustworthy source’ … says no one) states that the Swedish healthcare system is universal and mainly financed by the public sector with private healthcare being a rarity (like England). There is generally an underuse of primary care resources and the system is decentralised (ie. Doesn’t have the NHS, but the system is ran in each individual country. Thus, not like England). Therefore, it doesn’t totally apply to the UK, but there are enough similarities that you can get some information out of it.

Now for the fun part, the findings – meaning I get to make some graphs:

Graph 2

Graph 2: Shows the percentage of people in each group that died.

This is a good place to start – looking at this, more people with autism died than people without. Also looking at this, it is worse if you are female for mortality, but that exists with the controls as well. They report the odds ratio and confidence intervals for this and in those people with ASD are 2.56 times more likely to die than people without in this study. This is statistically significant as the confidence intervals at between 2.38 and 2.76, and so don’t overlap with 1 (the ‘point of null value’ where the numbers show that there isn’t a statistically significant difference in this group). So this shows that there is a difference. They report that individuals of the control group died at a mean age of 70.2 while the figure for the entire ASD group was 39.5 years and for the high-functioning group at 58.39 years – which shows where the reported figures come from.

We then get to see the cause-specific mortality – where they look at what people died from. Now this is where I need to stress the correlation-causation point again. This does not mean that people with autism die from this specific factor, it means that there is a link between a factor and the death of someone with autism. Causation is difficult to prove while correlation is really easy.

Graph 3

Graph 3: Shows the percentage cause-specific mortality in the control group and the population with autism when compared to the number of people who died.

Now this graph is messy, but shows a few bits that are important:

1) Their dataset reports the proportion of people in their total population who died from a specific cause, rather than just those already counted in the mortality group. This means that you couldn’t compare the causes of deaths between groups (and that it just looks like autistic people die more from everything than the general population). Therefore, I took that data and made it comparable between the different populations. Sadly this means I can’t discuss the confidence intervals as the ones they have produced aren’t really relevant to my question.

2) Common causes of death are represented in the control group – the most common causes of death across the population are generally from cancer (neoplasms) and cardiovascular/cerebrovascular disease (heart attacks and strokes), which is reflected in this data.

3) People with low functioning autism have a higher chance of dying from a neurological cause (by 7 times in comparison to the control group) – In this case this likely means epilepsy, as there are known higher rates of epilepsy in both autistic people and people with learning disability (the medically defined learning disability using IQ measurements – to keep this distinct from learning difficulty that can include dyslexia and dyspraxia, for which this doesn’t apply). But I cannot see the data completely so don’t know for sure.

4) People with low functioning autism have a higher chance of dying from congenital malformations – likely because congenital malformations are more likely to occur with increasing levels of severity of learning disability.

5) People with high functioning autism have a higher chance of dying from suicide – this one is pretty significant. From the data I have it appears that people with high functioning autism are 3.5 times more likely to commit suicide than the control group.

So from this I have a different opinion from the data that they have released. In their conclusions they say that the increase in mortality is across all causes. I’d say that this is true, but only because mortality is higher. When you look at the data from the causes relative to the number of people who have died rather than the population as a whole then it becomes (jn my opinion) more interesting as it shows the areas specific to the population where mortality is higher (which means you can target specific services to provide better healthcare and support) – these are neurological causes, congenital abnormalities and suicide. Ultimately, the article focusses on these anyway as they are the most distinctive features in both sets of data.

They also note that mortality is higher in females with ASD then with males. This is true, but women had a higher mortality in the control group too. This still had statistical significance, showing it was more prevalent than in the ASD group (1.99-2.51), but it was less significant an increase than for the male group (2.60-3.16). This basically just adds to show that women were more likely to die in this study then men regardless of whether they had autism, which is just a sad thing that is being looked into by other people.

A neat thing they do is talk about their strengths and weaknesses in the study and focus on the weakness of their selection from one database. They report that in Sweden the diagnostic label of autism was not used in the National Patient Register until 1987, so this limited their study. Furthermore, the first group of people to be given the diagnosis (from 1987-2001) were more likely to have a severe phenotype (characteristics) of the disorder (as they would have been diagnosed when they came into contact with clinical psychiatry services) and so influence the results. However, they tried to account for that as much as they possibly could (by mostly using participants from after 2001). They also note on the external validity of the study given its setting in Sweden (which I brought up earlier).

Overall the study is a pretty good study. I do think that they could have handled their data in a different way to make it a little bit easier to analyse, but it didn’t ultimately make a huge difference. They could have accounted for more confounding variables, but they may have been limited in this. Their study shows their conclusions well and highlights the need for more work.

So what does this mean?

1) Autistic people likely have, on average, a shorter life expectancy than people without.

2) High functioning people with autism are more likely to die from suicide

3) People with autism and a learning disability are more likely to die from neurological disorders and congenital abnormalities

4) More research needs to be done to find out what factors influence this mortality and generally to give more information as to why autistic people are dying at a younger age.

So, from this we have looked at the data and know some things about the problem. The next thing is to look at why these problems exist – why are higher functioning people with autism more likely to commit suicide? What are people with autism and a learning disability more likely to die from neurological disorders? This will be the focus of the next post.

Thank you for reading! As I said earlier, please let me know if I need to explain anything in a clearer manner. I’d be really interested to hear other people’s feedback and get an idea of anything that anyone else thinks about this.


Cusack, J. et al. (2016), Personal Tragedies, Public Crisis. Autistica.

Heslop, P. et al. (2013), Confidential Inquiry into premature deaths of people with learning disabilities (CIPOLD). University of Bristol.

Hirvikoski, T. Et al. (2016), Premature mortality in autistic spectrum disorder. The British Journal of Psychiatry. 208. 232-238.

Wikipedia (last accessed: 14th February 2018), Swedish Healthcare (URL:

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s