And the Band played on

Twenty years ago, I first read “And The Band Played On”, by the journalist Randy Shilts. The title refers to the band of the RMS Titanic, who played on — to calm the passengers — while the ship was sinking, and all lost their lives.

The book isn’t about the sinking of the Titanic; it’s about the HIV/AIDS epidemic. By the time the book was published in 1987, thousands of people had already died, and TV ads were warning that those deaths were just “the tip of the iceberg”.

A Tsunami of Dementia

Today, we’re facing another watery catastrophe: a “tsunami of dementia”.

When I joined the Sector Advisory Group for the National Dementia Care Pathway Framework, I was told — in no uncertain terms — that the scale of the problem made specialist Memory Services impractical, that dementia diagnosis and management must henceforth be the province of primary care.

I wasn’t sure about this, so I did some reading. At first I was impressed at the consistency of the statistics in all the documents I read. Then I realised that they were all quoting one source: the Dementia Economic Impact Report from Access Economics of Australia, commissioned by Alzheimers New Zealand in 2008.

Dementia in New Zealand

Key statistics

It states that, in 2008, there were 40,000 New Zealanders living with dementia and 12,000 new cases per year. It also states that people with dementia survive for an average of 7–10 years after diagnosis.

Problem is, these numbers literally don’t add up.

If you take the number of people living with a condition and divide by the number of new cases per year, you should get — roughly — the average survival time for the condition.

But 40,000 divided by 12,000 is nowhere near 7, let alone 10.

Of course, the population of people living with dementia isn’t fixed: it’s growing. The report estimates the increase at 1500 per year. But even that doesn’t balance the equation.

To reconcile the prevalence, incidence and survival times quoted in the report, the number of people in NZ living with dementia would have to be increasing not by 3% per year, but by 20% per year.

The number of people living with dementia would have to have doubled in the past five years… and I think we’d have noticed if that was the case.

Survival time

The survival time quoted in the report comes from a paper by Warner & Butler published in 2000… and which I couldn’t find anywhere, so I’ve no idea how that figure was derived. Still, that was 13 years ago, so there should be more recent research to corroborate the value.

In 2012, Rowntree et al reported an average survival time for people with Alzheimer type dementia of 11.3 years, but that’s only for Alzheimers and we’re interested in all types of dementia.

In 2008, Xie et al reported an average survival time of 4.5 years for all types of dementia.

In 2001, however, Wolfson et al pointed out a problem with studies like Rowntree’s and Xie’s.

They take their subjects from larger projects where people’s mental and/or physical health is being checked every two years or so.

People who are found to have developed dementia at one of these checkups are then tracked until they die, and their survival time is estimated as the time between their death and halfway between the checkup at which they were found to have dementia and the previous checkup.

length bias 1.024

Wolfson et al noted a “length bias": people who both develop dementia and die in the two years between checkups will not be counted, resulting in an overestimate of survival time.

Wolfson et al suggested a mathematical correction for this length bias, resulting in a survival time for all types of dementia of only 3.3 years

Which is a familiar figure (40,000 divided by 12,000) . Neither Xie et al or Rowntree et al used Wolfson’s correction, despite acknowledging the problem.

In 2010, Rait et al looked for people on GP’s lists who’d died with a diagnosis of dementia and tracked them back through their notes to the first mention of dementia.

The average survival time was 6.7 years for people in their 60s, falling to 1.9 years for people in their 90s.

This makes sense. Someone aged 65 can reasonably expect to live another twenty years; someone aged 85 can expect to live another five. So at all ages, dementia is a life-shortening condition.

Xie et al’s survival times for each age group were at least a year greater than Rait’s, supporting Wolfson’s idea of a “length bias”.

Rait et al didn’t give an overall survival time for dementia, but it should be a year or so lower than Xie et al’s, approximately 3–4 years.

This is a value which does fit with the prevalence and incidence estimates in the Economic Impact Report. But it is half the value used in the report. If the survival time in the Report was out by a factor of two, can we really trust the other values from the report?

Double checking


There are three ways of determining the prevalence of a condition: the simplest is to assess everyone and tot up the number of people who have the condition. This tends to be impractical, not least because of cost.

A cheaper approach is to assess everyone in a representative sample of the population and assume that the rate at which the condition occurs in the sample is the same as in the whole population.

The cheapest approach is to use prevalence rates already determined in another population, e.g. the UK or US. And this is what Access Economics did. They used a meta-analysis of meta-analyses of dementia prevalences in European countries from the 1940s to the 1990s to construct a model into which they could plug a breakdown of the NZ population and thereby estimate the prevalence of dementia in NZ.

They’d already used the same model in 2005 for a report on dementia in Australia. In 2010, Anstey et al questioned this model, noting that both the techniques of meta-analysis and the diagnosis of dementia had changed significantly over the half-century covered by the meta-analysis of meta-analyses.

In a 2011 update to the Australian report, Access Economics — now owned by Deloitte — acknowledged Anstey et al’s criticisms but argued that there was no better data available, so used the same model once again.

Six months later, in an update to the NZ report, they stated that they would incorporate some new data into the model, but to different degrees for different age groups, ranging from not at all though three parts old data to one part new data, up to one part old data to one part new data. Quite why they used these different ratios wasn’t entirely clear.

The World Health Organisation’s 10/66 Project has been working to measure current dementia prevalences throughout the world and, this year, Prince et al reported dementia prevalence as 5–7% of the vulnerable population, more in Latin America and less in sub-Saharan Africa, for reasons yet to be established.

So dementia could, at the time of the report, have affected 33–47,000 New Zealanders, a range of 700 people per Health Board.

How is that figure growing? As part of their criticism of Access Economics’ model, Anstey et al (2010) discussed a number of modern Australian data sets, one of which they set aside as potentially unreliable. This was the results of the National Surveys of Mental Health conducted in 1997 and in 2007.

At first glance, the figures do appear to be all over the place, but rearranging the data not by age group but by year of birth reveals an interesting pattern: dementia seems less prevalent amongst people born in the 1930s than in the 1920s.

Prevalence is a difficult measure with which to work: a more sensitive test for dementia will increase the prevalence, as more people will be found to have the condition.

In last month’s Lancet, Matthews et al published a repeat of a 1990s study of dementia prevalence in England using the same measures as were used in the 90s. In the original study, the percentage of people with dementia was found to increase with age, as we would now expect. In their recent repeat, the percentage of people with dementia still increased with age, but for almost every age group, the percentage of people with dementia was less than for people of the same age twenty years ago.

This means that the number of people living with dementia in England is much the same as it was twenty years ago, when an increasing and increasingly elderly population would have led us to expect a 30% increase.

Could this be true of NZ? We don’t know. But for now, not only can we not say for certain how fast the number of people living with dementia is increasing, we can’t even be sure that it is increasing.


What about incidence: the number of new cases every year?

As with prevalence, there are three ways of measuring incidence: monitor the whole population and note everyone who develops dementia; monitor a subset and apply the rate in the subset to the whole population; and use rates determined in other countries.

Access Economics again used data from the meta-analysis of meta-analyses of Europe to model the incidence of dementia in New Zealand (and Australia). In their 2012 update, they stick with this model. After all, it’s not likely that people are becoming less likely to develop dementia, is it?

In 2010, Rocca et al reviewed five studies which found trends towards fewer people of a given age developing dementia than in the past. None of these trends was statistically significant, however.

In 2012, Schrijvers et al again found a non-significant trend toward fewer new cases of dementia than in people of the same age twenty years ago. They did, however, find that people today have significantly greater brain volumes than people of the same age twenty years ago.

For now, we should probably stick with the WHO’s estimates, somewhere between 10 and 14,000 new cases of dementia per year.

So, we can’t be sure of the prevalence of dementia in NZ, we can’t be sure how the prevalence is changing — if, indeed, it is — and we can’t be sure of the incidence. Seems like all we can be sure of is the survival time.


Dementia Diagnosis

The Mini Mental State Examination

The Mini Mental State Exam was first published in 1975. Folstein, Folstein & McHugh warned that “the MMS[E] cannot be expected to replace a complete clinical appraisal” and yet it is the MMSE that is used in the vast majority of prevalence and incidence studies to “diagnose” dementia.

The WHO’s 10/66 project has laboured long and hard to create a valid cross-cultural, cross-language test battery for dementia but, at it’s heart, it’s really just the MMSE.

The MMSE was developed as a fast crosscheck for a psychiatrist’s subjective impression of a patient’s cognitive funtioning. As a means of diagnosing dementia, it’s not that great: it can miss up to 25% of cases and misdiagnose up to 40% of cases with a dementia they don’t have.

So imagine my surprise when I discovered that many GPs and physicians make dementia diagnoses without even using this flawed instrument to confirm their subjective impressions.

Dementia diagnosis is hard

Mitchell et all (2011) presented a meta-analysis of GP dementia diagnoses which was widely publicised. The finding that got all of the attention was that GPs miss a quarter of the dementia cases on their lists, and don’t offer a diagnosis of dementia in half of the cases of which they are aware. So if you actually have some form of progressive cognitive impairment there’s a 25% chance that your GP will never discover this and a 40% chance your GP is aware but isn’t telling you.

The finding that didn’t get so much attention was that, for every case of dementia GPs miss, they mistakenly diagnose two other people with dementia. So if you have been given a diagnosis of dementia by your GP, there’s a 40% chance you don’t actually have dementia. In NZ, that would be in excess of 20,000 people not only being treated for and using services for dementia they don’t have, but not being treated for whatever is their real problem (if any).

This is not a criticism of GPs. Dementia diagnosis is hard, even for the specialists. A recent study (TK) found that geriatricians diagnosed the wrong type of dementia in one-third of the cases they saw. The idea that dementia is a diagnosis that can be made by lone non-specialists, rather than a specialist multidisciplinary team, that cases which appear simple actually are simple, leads inevitably to these sorts of figures.

Survival time again

The reason I’m stressing this point is that our value for survival times was derived primarily from analysis of GP case records which, according to Mitchell et all, could be missing up to 25% of dementia cases and, of the people included, 40% may not actually have dementia.

So, really, we don’t really have any idea of the prevalence of, incidence of or survival time for dementia in New Zealand, now or in the future.


A tsunami of dementia?

When you’re facing a tsunami, it’s not that hard to figure out what to do: get far enough out to sea or to high enough ground and you can ride it out in safety. The trick, of course, is getting sufficient warning.

Are we facing a tsunami of dementia? I don’t think so.

I think we’re facing another iceberg. And the defining characteristic of an iceberg is not that what you can’t see is bigger than what you can see, the defining characteristic is that what you can see is no guide at all to the shape of what you can’t see.

So you can set what you think is a safe course and still be heading for disaster.

Access Economics have said, in every one of their reports, that we need hard data on dementia in NZ, especially given that Europe and Australia don’t have significant populations of Maori and Pacific people and we do. We know next to nothing about dementia in Maori, but then we know next to nothing about dementia in any New Zealanders.


There are three things we need to do.

First, we need to stop talking with certainty, as if we know what we’re talking about.

If I tell you that there are 50,000 people living with dementia in NZ and this number grows by 1,500 every year, you will — hopefully — say “we must do something about this.”

If I tell you that there’s probably at least 35,000, but maybe no more than 60,000, people living with dementia in NZ, and that this number might be increasing or it might not, and if it isn’t now then it might in the future but we’re no longer sure how or when, you’re more likely to say “shouldn’t we maybe find out a bit more about what’s happening?”. You’ll be correspondingly less likely to propose radical changes to our services to deal with a tsunami that may not be happening and may never happen.

Second, we need to have serious public debate about a dementia register: not a register of people with dementia who are willing to participate in research about dementia, but a register of all people with dementia, detailing how they were diagnosed and what happens subsequently. The scientific potential would be incredible: so too would be the potential for abuse of human rights. We need to start this debate now.

Third, all of the published research on dementia prevalence, incidence and survival times has flaws which are repeated again and again. We need a handbook or guidebook for researchers and funders of research to ensure that investment in NZ research won’t be squandered by repeating the same old mistakes again & again. These flaws are, in large part, due to methodologies which attempt to cut costs, but there’s no economy in cheap, meaningless research.

Calming the passengers

We don’t have to do these things, of course. We can carry on with our day-to-day responsibilities without thought of the course being set for us.

In which case we have to hope that we don’t one day find ourselves in the middle of a disaster, with nothing to do but play on, keeping everyone else calm as we go under.

I can’t play a note, and I never learned to swim. Please, let’s not be the band.