Abstract and Introduction
Official statistics in Australia and the United States show large recent increases in dementia mortality rates. In this study, we assessed whether these trends are biased by an increasing tendency of medical certifiers (predominantly physicians) to report on the death certificate that dementia was a direct cause of death. Regression models of multiple-cause-of-death data in Australia (2006–2016) and the United States (2006–2017) were constructed to adjust dementia mortality rates for changes in death certification practices. Compared with official statistics, the recent increase in adjusted age-standardized dementia death rates was less than half as large in Australia and about two-thirds as large in the United States. Further adjustment for changes in reporting of dementia anywhere on the death certificate implied even lower increases in dementia mortality. Declines in reporting of cardiovascular diseases as comorbid conditions also contributed to rises in dementia mortality rates. The increasing likelihood of dementia's being reported as directly leading to death largely explains recent increases in dementia mortality rates in both countries. However, studies have found that reported dementia on death certificates remains low compared with clinical evaluations of its prevalence. Improved guidance and training for certifiers in reporting of dementia on death certificates will help standardize mortality statistics within and between countries.
In Australia and the United States, dementia will become an even more important public health issue in coming decades because of continued population aging. Accurate dementia mortality data enable estimation of how long a person with the condition can be expected to live and for how many of those years they will live with severe disability, and so are an important source of health intelligence to better inform clinical understanding of dementia and planning by various authorities for health and care services. Official dementia mortality statistics, collected by national authorities for all registered deaths in a population, provide valuable evidence with which to estimate the burden of disease from dementia and associated economic costs, to obtain a nuanced understanding of differentials by demographic and socioeconomic characteristics, and to identify conditions that are commonly comorbid with dementia mortality. As a consequence, there has been increasing interest by the public health community in obtaining more reliable and comparable data on the burden of dementia.
Official mortality statistics in high-income countries show that mortality rates from Alzheimer disease and other forms of dementia have been increasing in recent years.[4–7] In Australia between 2010 and 2019, the reported number of dementia deaths increased by 67% and the dementia age-standardized death rate (ASDR) rose by 27%, with dementia now reported to be the leading cause of death for females and the second-leading cause for males. In the United States between 2009 and 2018, the number of dementia deaths rose by 51% and the dementia ASDR increased by 23%, with dementia now being the leading cause of death for US females and the third-leading cause for males. Based on these trends, the Australian Bureau of Statistics has predicted that dementia will be the leading cause of death for both sexes in Australia in coming years.
These reported trends in dementia mortality rates contrast with studies suggesting that dementia incidence and prevalence are either stable or decreasing.[10–14] In an analysis of 7 population-based cohort studies conducted in Europe and North America, Wolters et al. reported a 13% decennial decline in dementia incidence rates over the past 25 years, faster for males than for females. However, in another review of studies, Prince et al. concluded that trends in prevalence are less clear because declines in incidence may be offset by improvements in survival. In Australia, there is less evidence on trends in dementia incidence and prevalence, with one study finding marginal declines in prevalence in recent years, standardized for the aging of the population, both for people accessing long-term care and for people using home care.[15–18] Projections of dementia prevalence and associated economic costs generally assume that prevalence will remain constant in the future.[19–22]
One possible reason for the divergent trends between official dementia mortality statistics and dementia incidence and prevalence is changes in medical certification practices. Dementia mortality reported in official statistics in high-income countries is based on what the medical certifier (predominantly physicians, but the certifier can be a coroner or, in the United States, a nurse practitioner or another designated agent, depending on the state) reports about causes of death on the International Form of Medical Certificate of Cause of Death (i.e., the international death certificate—see Web Figure 1, available at https://doi.org/10.1093/aje/kwac047). Dementia can be reported either as leading directly to death (part 1 of the certificate) or as another significant condition contributing to death (part 2 of the certificate). This information is used, with International Classification of Diseases, Tenth Revision (ICD-10) coding rules, to select an underlying cause of death (UCOD), which is the information on mortality reported in official statistics.
For dementia to be reported on the death certificate, the decedent must 1) have had dementia, 2) have been diagnosed with dementia, and 3) have had dementia assessed by the medical certifier to have either led directly to death or been another significant condition that contributed to death. Studies suggest that dementia prevalence trends are essentially stable, or at least not increasing anywhere near the extent of dementia mortality according to official statistics. In this context, trends in dementia mortality may be affected by changes in the tendency of medical certifiers to report dementia anywhere on the death certificate, as well as whether they report dementia in part 1 or part 2. These changes may be due to improved diagnostic practices for dementia, shifting diagnostic preferences or "diagnostic fashions," improved recognition and knowledge, and better-informed attitudes about dementia's role in leading to death.[20,24] For example, 2 studies from the United Kingdom showed that reporting of dementia anywhere on a death certificate has increased substantially in recent decades, with one study finding that the prevalence of any mention of dementia on the death certificate increased from 40% of decedents with dementia in 2006 to 63% in 2013.[25,26] Underreporting of dementia on the death certificate has been found to be significant as compared with clinical evaluations in the United States (2 studies specifically focused on Alzheimer disease)[27,28] and estimates from administrative data linkage studies from the Netherlands and Australia.[16,27–29] This reflects the underdiagnosis of dementia more generally, with one systematic review and meta-analysis finding that most dementia is undetected. Various studies have also found that dementia is more likely to be reported on a death certificate if it is more severe or if the deceased lived in an institution, was older, or had Alzheimer disease, and less likely if the deceased died in a hospital.[25,26,31]
Trends in dementia mortality statistics are also influenced by changes in the number and type of comorbid conditions present at or around the time of death, which are especially common among the very old, and whether medical certifiers report these in part 1 or part 2 of the death certificate. For example, according to ICD-10 coding rules, when ischemic heart disease and dementia are both reported in part 1 of the death certificate, ischemic heart disease is selected as the UCOD. However, if ischemic heart disease is reported in part 2 and dementia in part 1, then dementia is selected as the UCOD. Differing certification practices can lead to substantial variations in dementia mortality rates; for example, dementia death rates at age 85 years or more in Australia are 6 times higher than those in Japan, despite very little difference in prevalence between the 2 countries.[32,33] Certification of deaths where there are multiple comorbid conditions, with often complex interactions, is challenged by their reliable diagnosis and also accurate attribution of the death to a chain of causes.
These measurement issues make comparisons of official dementia mortality statistics over time challenging. To develop improved estimates of trends in dementia mortality in Australia and the United States, we analyzed multiple-cause-of-death (MCOD) data, which comprise all conditions reported on death certificates for all registered deaths (i.e., the data used in official statistics). MCOD data are an underutilized data source which are particularly useful for better understanding of the sequential role of morbid conditions in leading to death, and so are particularly relevant for dementia. The specific objectives of our study were to:
develop and apply statistical models to estimate age-standardized dementia death rates and their trends in Australia and the United States, accounting for the likely bias arising from changing death certification practices; and
analyze the presence of comorbid conditions reported on the death certificate and how they influence trends in dementia mortality rates.
This additional intelligence should help to inform interpretation of data on dementia mortality, including evaluation of reported statistics on the comparative importance of dementia as a cause of death in Australia and the United States, and potentially elsewhere.
Am J Epidemiol. 2022;191(7):1270-1279. © 2022 Oxford University Press