Background
Dementia is a principal cause of death in England and Wales [
1] and the United States [
2], and it is estimated that a third of people over the age of 60 die with the condition [
3]. Estimates of survival after disease onset have varied between 3.3 and 11.7 years [
4] – and the uncertainty over prognosis creates difficulties for patients and their family carers about planning their life and how to prepare for end of life care [
5]. The variability in survival time in dementia is most likely a consequence of other illnesses affecting survival. Some studies report a trend in recent years towards compression of morbidity, where the onset of the first chronic illness occurs at a later age, thus compressing the time living with chronic illnesses [
6].
There is increasing emphasis on providing good end of life care in dementia, which tends to focus on the challenges of delivering it to people with severe dementia [
7]. The proportion of people who have mild, moderate, and severe dementia at end of life has important implications for clinician- and public-awareness of prognosis, how healthcare services plan and implement treatment, and how clinicians provide evidence-based end of life care. Severe dementia is associated with lack of insight
and capacity [
8] to make decisions about healthcare, requiring others to make decisions on behalf of patients [
9]. Someone with mild dementia is more likely to be able to make these decisions, but may require support to do so and to implement them. If many people who have dementia die with a mild form of the condition, this would have implications for palliative settings and may suggest a greater need for support for decision making and management strategies. One study reported that 31/68 (46%) people with dementia being managed at end of life by a general practitioner were described by the clinician as having mild dementia, [
10] however there has been, to our knowledge, no study examining the distribution of dementia severity in a large cohort of people with dementia at time of death.
In this study, we therefore aim to investigate the severity of dementia at death, in a secondary mental healthcare service cohort of older people with a diagnosis of dementia and explore the association between demographic and clinical factors, including medication use (antidepressant, sedative and antipsychotic drugs, and those for cognition in dementia), and dementia severity at death.
Discussion
This is the first study, to our knowledge to report the severity of dementia at death of a cohort of people with dementia. Importantly we find that only one quarter had severe dementia, half had moderate dementia and the rest had mild dementia at death. The mean MMSE at death was 15.3 and we found a normal distribution of estimated MMSE scores at death except for those with estimated MMSE score of 0, a consequence of the floor effect of the MMSE assessment.
This result was unexpected as we had anticipated that most people in the cohort would die with severe dementia. It suggests that most of our cohort died from other illnesses rather than solely from the consequences of advanced dementia, although the presence of dementia may have led to a worse prognosis than others with the same illness [
21]. The age-specific incidence of dementia has decreased in the UK over the past 20 years [
22], meaning that the condition occurs later now than in previous generations, and is therefore more likely to coexist with other illnesses. Around 70% of people with dementia have at least two comorbid chronic diseases and these increase the risk of hospitalisation and mortality [
23]. This elevated mortality may have contributed to a large proportion of our cohort dying with milder forms of the disease. Furthermore, a major drive in dementia policy in the UK and other countries has been early diagnosis of dementia [
24] and the services whose data we used have been successful at increasing dementia diagnosis [
25], so our naturalistic cohort includes many who had dementia diagnosed at very mild stages, who may have died from other conditions, with less severe dementia.
The mean MMSE was slightly lower in the whole sample than in our sensitivity sub-group of people assessed within one year of their death, as the whole cohort included a higher number of people (46 v 24 people) whose estimated MMSE was 0, due to a long interval between their last MMSE and death. In the sensitivity analysis, a higher proportion of people had severe dementia than in the main analysis, which may be because our method underestimated the decline, which may be more precipitous before death, or because people with more advanced dementia were more likely to have been seen and cognitively assessed by the clinical service close to death.
More severe dementia at death was associated with being female, from ethnic minority background, more socioeconomically deprived, showing symptoms of agitation and taking antipsychotic medication. Divorced people died with less severe dementia. Our observational analyses of the association between sociodemographic and clinical factors and dementia severity cannot establish a causal relationship. The identified factors may cause dementia to progress more rapidly; they may indicate people with lower cognitive reserve [
26] and therefore more severe cognitive impairment at the same pathological stage of the disease; or they may indicate those with better physical health who therefore live longer with dementia and have the condition more severely when they die.
We found that women died with more severe dementia than men, which may result from this cohort of women having received less education [
27] and therefore having less cognitive reserve so showing more severe cognitive symptoms for the same level of pathology, or be a consequence of their greater life-expectancy, allowing more dementia progression before death. Our finding that people from minority ethnic background died with more severe dementia than their white counterparts may also reflect lower cognitive reserve related to less education [
28]. In addition, people from ethnic minority backgrounds develop dementia at a younger age [
29] and may therefore live longer with the disease. A final explanation is that people from minority groups present later [
30], so those who die with less severe illness may never have presented to memory services. Alternatively, people from ethnic minority origins may engage less with management of their condition [
31], thus experience a more precipitous cognitive decline. Higher neighbourhood level socioeconomic deprivation being associated with more severe dementia at death is also likely to result from lower cognitive reserve and younger development of dementia [
32]. Our finding that divorced individuals died with less severe dementia compared to married people may be due to the mortality risks associated with being divorced [
33] meaning divorced individuals died earlier and with milder dementia than married people.
Our results showed agitation to be associated with more severe dementia at death and this is likely to be because agitation at diagnosis is a consequence of more severe dementia, which increases in severity as dementia progresses [
34]. An alternative is that agitation caused more rapid dementia progression, supported by studies suggesting that neuropsychiatric symptoms increase neuropathological burden [
35]. Antipsychotics are used for psychosis or sometimes agitation, which are markers of more severe dementia, although the potential contribution of antipsychotic use to cognitive decline [
36] needs further exploration.
Strengths and limitations
This is the first study, to our knowledge, to evaluate dementia severity at death and examine demographic and clinical predictors. Our database allowed analysis of all routinely collected clinical data, with no requirement for explicit consent to involvement in the database, meaning that we had a large naturalistic cohort of people who had been clinically-diagnosed with dementia before dying. Our results are likely to be representative of people with clinically-diagnosed dementia living in similar areas, as memory services are the mainstay of UK dementia diagnosis and assessment [
24] and CIFT has a high estimated diagnosis rate – 84% of people in the catchment area estimated to have dementia from epidemiological studies, have a formal diagnosis [
25] – but our findings may not generalise to people with dementia who have not been diagnosed, who may have been more likely to die with milder dementia.
Our study has limitations; the use of electronic health records which were not collected for research purposes meant that we could only adjust for routinely recorded factors. We would have liked to adjust for education, which would be a more sensitive marker of cognitive reserve than socioeconomic deprivation, and would have ideally had more sensitive measures of physical-ill health, enabling us to examine in detail the potential confounding effect of physical illness on dementia severity at death. Our analysis assumed a linear decline in MMSE at end of life, and applied the average MMSE decline to the whole sample, whereas this may not be the case. Rate of decline likely depends on a number of factors including initial severity, type of dementia, age, neuropsychiatric symptoms, medication, sex and education and we did not have data on all these domains to allow us to predict decline for each patient individually, so we used the mean rate of decline in the sample. The mean annual rate of decline in our sample was 1.5 MMSE points. One study reported 6 month MMSE decline of 0.9 points [
37], whilst others have reported an annual decline of 2.2–2.3 points [
38,
39], although these studies only examined people with Alzheimer’s disease, whereas we included people with all dementia types and the rate of decline differs according to dementia type [
40]. Finally, for some patients, it is possible that MMSE testing was abandoned when they had more severe dementia as the patient may struggle to complete testing and/or the clinician may deem that it does not add significant information, although clinicians would often record the MMSE score at the point of abandoning the test; however, reduced recording of MMSE in severe dementia may mean that dementia severity is underestimated in this study.
Conclusions
We found the majority of people in the sample died with moderate dementia and only one quarter had severe dementia at death. This study provides important information for clinicians and the public about the prognosis of people with dementia and should inform the development of end-of-life services tailored to the condition. Dying with severe dementia is a major concern for people with early disease [
41], but we found that this occurs less frequently than expected. People with dementia, their family carers, and healthcare professionals involved in their care may delay decisions related to end-of-life, assuming this may not be relevant, and research suggests that take-up of advance care plans has been low; [
42] however, our research supports the need for professionals to encourage early discussions about who may help with decisions and end-of-life preferences. All palliative care in older people should consider if the individual has dementia and how this affects their decision-making ability, the possible treatments and their ability to carry out a plan without help. People with mild to moderate dementia often appear unimpaired in social contexts to people who do not know them and it is easy to miss the disability that their cognitive impairment may cause. Clinicians involved in end of life care should consider the presence of dementia in people with less obvious signs of the condition and take this into account when explaining clinical information and making treatment decisions.