Cost-of-illness studies are simultaneously misleading and helpful.
Let’s start with how they help. They add up the economic impacts associated with a particular illness or diagnosis (such as schizophrenia), usually for a single country for a single year. These economic impacts could include the costs of acute treatment, long-term support, impacts on employment and productivity, perhaps also impacts on family members. The overall figure usually totals many millions or billions of dollars, euros or pounds.
Advocacy bodies love cost-of-illness calculations because they grab headlines in the print and broadcast media, and are conveniently succinct for Twitter. They are therefore helpful to brandish when arguing the case for more policy attention and, especially, for treatment action for a particular condition. Pharma companies like them too, because they help set the scene for marketing campaigns.
As well as totalling up figures for a whole country, authors will often report cost per individual with the condition over, say, a 1-year period. For example, Jin and Mosweu (2017) calculated that total societal cost (direct healthcare, direct non-healthcare and productivity losses) per schizophrenia patient ranged from US$5,818 in Thailand to US$94,587 in Norway, based on original calculations by Phanthunane et al (2012) and Evensen et al (2015). Less commonly, cost-of-illness studies report costs from the point of incidence for the duration of an illness and/or for a period of years.
Sometimes these are reported as gross costs, and sometimes as excess costs over and above what would be expected for the general population.
The review paper by Maria Christensen and her co-authors looks at cost-of-illness studies “for mental disorders worldwide and to explore the distribution of the costs between disorders”.
The authors report a well-conducted systematic review in 2019, covering studies published since 1980. As well as cost-of-illness studies, they also included what they call ‘cost-analyses’ which may not aggregate figures up to a national total. They looked at all mental disorders but not dementia. They only included studies that looked at general populations, and where results were already reported in monetary units.
Their search strategy looks great, and the reporting of it is exemplary: everything is clear, either in the paper or in an online resource. Among many other things, each included study was given a quality rating. There is also a marvellous interactive website that summarises the findings (we are very jealous!).
Not surprisingly, the vast majority of the 143 studies included in this review were conducted in high-income countries, with just three from Africa and one from South America.
Looking at the mental health problems, the most frequently studied were ‘mood disorders’ (54 studies), ‘schizophrenia’ (40) and ‘neurotic disorders’ (28). Some studies reported estimates from multiple diagnostic groups or combinations. Almost all the studies used a prevalence-based approach (looking at all people with a particular condition in, say, a 1-year period), whilst fewer than 10% used an incidence-based approach (basically totalling up the costs from the point at which a condition starts or is diagnosed, perhaps over many years).
It is impossible to mention more than a tiny proportion of the results contained in this rich paper, but here are a few:
- A common finding across countries was that schizophrenia and intellectual disabilities tended to have the highest societal cost
- Costs were greater in high-income countries than elsewhere
- Indirect costs (particularly productivity losses because of illness or premature death) were high for all diagnostic groups
- For some conditions there were very few studies at all: for example, only two studies of ‘developmental disorders’ (which includes autism spectrum disorders) were found. This was the condition with the highest cost per individual
- Mood, neurotic and substance use disorders were less costly per individual, but obviously have much higher prevalence rates, and therefore their total impact at a national level is generally much greater.
As well as the summary results described above, the authors discuss broader trends and implications from their work:
- Mental disorders are very costly for societies
- There is a lot of variation between disorders, and between countries
- Indirect costs account for a substantial proportion (roughly half) of the total cost, yet often get missed in evaluative studies
- There are gaps in available evidence for some mental health problems; and there are enormous gaps for low- and middle-income countries.
Strengths and limitations
Maria Christensen and her co-authors emphasise the limitations of cost-of-illness calculations. As we noted earlier, these studies can mislead. One reason is because they only include what a research team has been able to measure: important economic impacts may get missed because data are not available, or impacts are not thought to be relevant.
A good example comes from the area of perinatal mental health. A study many years ago in England calculated the service-related costs of postnatal depression (Petrou et al 2002). This was a well-conducted study, but later work showed that more than two-thirds of the economic impact of postnatal depression stems from the adverse effects on the birth children (Bauer et al 2016).
Second, cost-of-illness figures do not represent any kind of ‘absolute truth’. Rather, they measure how much a healthcare system or a society is prepared to allocate to a particular patient group in terms of treatment and support. Costs will go up if more people are diagnosed and get more treatment. But costs will probably also go up if fewer people are diagnosed or get treatment, because there will be more crises, leading to extra demands on the healthcare system, as well as higher productivity losses from disrupted employment.
This can leave advocacy bodies in a quandary. Should they present the total cost as too high (‘Look at this enormous cost: surely we should be paying more attention to this particular patient group’) or instead as too low (‘Look at this level of spending: surely it can’t be enough to support this particular patient group’)?
Third – and this is where cost-of-illness studies can be most misleading – they do not and cannot signal the areas that need most policy or practice attention. Yes, we would probably all want our healthcare systems to pay more attention to health problems that are more prevalent and have bigger adverse impacts, including impacts measurable in economic terms. However, we would probably also want resources to be targeted so as to have their greatest impact, and this may not be in those areas with the biggest cost-of-illness figure.
Consequently, cost-of-illness studies do not necessarily help decision-makers to make decisions, despite what some advocacy bodies might argue. Those studies do not say whether the resources included in the calculations are appropriately deployed – or indeed needed at all – nor whether the treatments and services included in the total cost are actually any good. They could be useless or even harmful. In other words, cost-of-illness studies cannot signal ways to improve either the efficiency with which available resources are used, or the fairness of patterns of access, utilisation or impact.
Maria Christensen and colleagues clearly recognise and emphasise this last point. They recommend urgent use of cost-effectiveness analyses to provide more pertinent evidence to support decision-making. This does not make cost-of-illness studies irrelevant – and we confess that we have ourselves conducted some such studies, and we will again in the future – but they need to be seen for what they are and in their appropriate context. Where they can help is in highlighting the often enormous implementation challenge of moving from research evidence to better policy and practice (Knapp and Wong, 2020).
Implications for practice
For the reasons just set out, this paper doesn’t really have direct implications for practice. What it does do – and in a clear, comprehensive, well-conducted manner – is summarise cost evidence that can provide a backdrop to other research with more immediate relevance to practice decisions.
Statement of interests
MK co-authored a paper with Harvey Whiteford many years ago; otherwise neither MK nor GW has any other conflicts of interest.
Christensen M, Lim C, Saha S, Plana-Ripoll O, Cannon D, Presley F, Weye N, Momen N, Whiteford H, Iburg K, McGrath, J. (2020) The cost of mental disorders: A systematic review. Epidemiology and Psychiatric Sciences, 29, E161. doi:10.1017/S204579602000075X.
Bauer A, Knapp M, Parsonage M (2016) Lifetime costs of perinatal anxiety and depression. Journal of Affective Disorders 192 83-90.
Evensen S, Wisløff T, Ullevoldsæter Lystad J, Bull H, Ueland T, Falkum E (2016) Prevalence, employment rate, and cost of schizophrenia in a high-income welfare society: a population-based study using comprehensive health and welfare registers Schizophrenia Bulletin 42(2) 476–483.
Jin H, Mosweu I (2017) The societal cost of schizophrenia: a systematic review. Pharmacoeconomics 35 25-42.
Knapp M, Wong G (2020) Economics and mental health: the current scenario. World Psychiatry 19(1): 3-14.
Petrou S, Cooper P, Murray L, Davidson L (2002) Economic costs of post-natal depression in a high-risk British cohort. British Journal of Psychiatry 181(6) 505-512.
Phanthunane P, Whiteford H, Vos T, Bertram M (2012) Economic burden of schizophrenia: empirical analyses from a survey in Thailand. Journal of Mental Health Policy and Economics 15(1) 25–32.