Evaluation of Underlying Causes of Death in Patients with Dementia to Support Targeted Advance Care Planning

Abstract

Background: Insight in causes of death in demented patients may help physicians in end-of-life care.

Objectives: To investigate underlying causes of death (UCD) in demented patients stratified by age, sex, dementia subtype [Alzheimer’s disease (AD), vascular dementia (VaD)] and to compare them with UCD in the general population (GP).

Methods: A nationwide cohort of 59,201 patients with dementia (admitted to a hospital or visiting a day clinic) was constructed [38.7% men, 81.4 years (SD 7.0)] from 2000 through 2010. UCDs were reported and compared to the GP by calculating relative risks (RRs).

Results: During follow up [median follow up time 1.3 years (IQR 0.3– 3.0)], 64.2% of women and 69.3% of men died. Leading UCDs were dementia (17.5% in men and 23.7% in women) and cardiovascular disease (CVD) (18.7% and 19.2%, respectively). When compared to the GP, dementia was a more common UCD (RR in men 4.65, 95% CI 4.43–4.88), while CVD (RR in men 0.67, 95% CI 0.65–0.68) and cancer (RR 0.40, 95% CI 0.39–0.41) were less common. These differences were more pronounced in patients aged between 60–69 as compared to those aged≥90 years. Patients with AD died less often of cerebrovascular diseases as compared to VaD (RR in men 0.53, 95% CI 0.47–0.59).

Conclusion: UCDs in patients with dementia differs from that of the GP, as dementia is more often and cancer less often an UCD. Although less frequent compared to the GP, CVD also is one of the leading UCDs in patients with dementia. This information is valuable for targeted advance care planning.

Keywords

Cardiovascular disease cause of death cohort dementia mortality

INTRODUCTION

Dementia is increasingly becoming a major health concern. The incidence of the disease is increasing with the aging of the population and numbers of death attributed to dementia are considerably growing in recent decades; the proportion of deaths resulting from dementia increased 68% in the US between 2000 and 2010 [1 –3]. It is expected that dementia will become the leading cause of death in the near future in industrialized countries [4, 5].

Cardiovascular disease (CVD) and pneumonia are the main attributable causes of death in patients with dementia, as demonstrated in several registry- and autopsy-based studies [6 –11]. The majority of these studies investigated the causes of death in the entire group of demented patients without making a distinction between sex, age, and type of dementia [6 , 9–12] or they were of a small sample size [6 , 10]. As a result, little is known about the potential differences in causes of death by age, sex, and dementia subtype.

More insight in causes of death stratified by different patient characteristics can be valuable for patients, clinicians, and carers concerning decision-making in daily practice with respect to targeted advance care planning (ACP). ACP is a process of communication between individuals and their health care providers to discuss and plan for future health care, based on personal preferences and goals. One of the components of ACP is the discussion on whether to initiate or refrain from diagnostic or therapeutic interventions. Yet, uptake of ACP is increasing in the general elderly population and in patients with CVD and cancer, but is known to be low among people with dementia [13, 14]. However, life expectancy in dementia is poor and decision-making in an acute setting concerning diagnostic and therapeutic interventions (including living wills, instructions regarding antibiotic therapy, resuscitation, or admission to an intensive care unit) is less complicated within the context of a previous discussion on preferences and personal goals of terminal care and management. Therefore, ACP is extremely important particularly in this group of patients. Causes of death can support this discussion, e.g., if CVD appears to be among the leading causes of death, instructions regarding prevention (whether or not to initiate preventive measures) or treatment (including instructions with respect to resuscitation or admission at an intensive care unit), can be more firmly established.

Therefore we aimed to examine the underlying cause of death in a large nationwide cohort of patients with dementia stratified by age, sex, and dementia subtypes [Alzheimer’s disease (AD) and vascular dementia (VaD)]. To put the underlying causes of death into perspective, we compared them to those observed in the general population.

METHODS

Databases

Patients with dementia were identified by linkage of three nationwide databases, the Dutch Hospital Discharge Register (HDR), the Dutch Population Register (PR), and the National Cause of Death Register. Since the 1960 s, medical and administrative data for all admitted and memory/day clinic patients visiting a Dutch hospital are recorded in the HDR; no information from outpatient visits and nursing home residents is available. Patients in The Netherlands are referred to the day/memory clinic either in case of memory-related disorders (memory clinic) or with multi-morbidity, which also might include memory-related disorders (day clinic). In The Netherlands, a day clinic visit is a one-day hospital admission and therefore considered to be inpatient care. Around 100 hospitals participate in the register. The HDR contains information on patients’ demographics (date of birth, sex), type of hospital, admission data, and principle diagnosis at admission. The principal and secondary diagnoses are determined at discharge and coded using the ninth revision of the International Classification of Diseases (ICD-9-CM) [15]. The PR contains information on all legally residing citizens in The Netherlands, including date of birth, sex, current address, postal code, nationality, and native country. In the National Cause of Death register, all principal and any underlying causes of death are reported. In The Netherlands, it is obliged to fill out a death declaration form after the death of any person which has to be sent to the national cause of death statistics. Death reports are coded according to the International statistical Classification of Diseases and Related Health Problems, 10th version [16]. The overall validity of these registries is proved to be high[17 –19].

Cohort identification

All patients first ever hospitalized or first ever referred to the day/memory clinic for dementia (either a principal or secondary diagnosis: ICD-codes 290.0; 290.1; 290.3; 290.4; 294.1; 331.0; 331.1; 331.82) aged between 60 and 100 years were selected from the HDR between 1 January 2000 and 31 December 2010. In the Dutch population, there are about 2.9 million people aged 60 years and older. A recent validation study showed high validity of the use of ICD-9 codes to identify patients with dementia (positive predictive value was 93.2%) [20]. Collected cases were linked with the PR by using the record identification number assigned to each resident in The Netherlands with a unique combination of date of birth, sex, and the numeric part of the postal code. The use of the unique record identification number enables to identify different admissions, even in different hospitals, from the same person. Through linkage of these selected cases with the National Cause of Death registry, follow-up information on date of death, and principal and underlying causes of death could be obtained. No information on severity of disease or medication use was available in the registry. The approach resulted in a cohort consisting of 59,201 patients.

Privacy issues

Linkage of data form the different registries was performed in agreement with the privacy legislation in The Netherlands [21]. All linkages and analysis were performed in a secure environment of Statistics Netherlands.

Cause of death

Cause of deaths are obtained from the National Cause of Death Register in which underlying causes are reported. The underlying cause of death is defined as the disease that started the chain of events leading to death. We divided causes of death into eight different subgroups (1. Cardiovascular disease: ICD-10 codes I00-I99, R00-R01, Q20-Q28; 2. Cerebrovascular disease: ICD-10 codes I60-I69; 3. Cancer: ICD-10 codes C00-C97; 4. Pneumonia: ICD-10 codes J09-J22; 5. Chronic respiratory diseases: ICD-10 codes J40-J47; 6. Genitourinary diseases: ICD-10 codes N00-N98, R30-R39; 7. Gastrointestinal diseases: ICD-10 codes K00-K93, R10-R19; 8. Dementia: ICD-10 codes G30, G318, F00-F03).

Data analysis

Continuous data were summarized as mean and standard deviation or as median and interquartile range where appropriate. Categorical data were summarized as percentages. Patients were followed up from their earliest date of dementia hospitalization or referral to the day/memory clinic for dementia. Patients were censored in case of death or the end of the study period at 31 December 2010.

Underlying cause of death frequencies among patients with dementia were calculated and presented as percentages, stratified by sex and age of death (classified in 5-year groups). Since numbers of death in age groups 60–64 and 65–69 years were rather low, we combined these in one category of 60–69 years. We also checked whether there were differences in underlying causes of death with respect to setting of care (day clinic visits versus inpatient care). Additionally, we compared underlying causes of death among patients with AD and VaD, stratified by sex. Relative risks were calculated (AD versus VaD) and presented with corresponding 95% CI.

To put the percentages of underlying causes of death of overall dementia into perspective, we also compared them to the causes of death in the general population. Age and sex specific causes of death for men and women aged 60–99 are available online from Statistics Netherlands [22]. A direct method for age-standardization was used on the basis of the age distribution of the 2005 Dutch population with 5-year age groups. Relative risks (RRs) of underlying cause of death in dementia versus the general population were calculated and presented with corresponding 95% confidence intervals (95% CI).

Data were analyzed with SPSS software, version 20.0 (SPSS Inc, Chicago, Illinois, USA). A two sided p-value <0.05 was considered statistically significant.

RESULTS

In total, 59,201 patients diagnosed with dementia were identified through cross-linkage of the three registries. Mean age was 81.4 years (±7.0). The majority was women (61.3%). Median follow up time was 465 days (Inter Quartile Range 117–1081). Baseline characteristics are shown in Table 1.

Underlying causes of death among patients with dementia

During follow up, 23,269 women and 15,895 men died (64.2% and 69.3%, respectively). Table 2 shows the percentages of causes of death in patients with a first hospitalization or first day clinic visit in The Netherlands between 2000–2010, stratified by sex and setting of care. The three leading underlying causes of death in women with dementia were dementia (23.7%), CVD (19.2%) [in which 28.1% consisted of ischemic heart disease (IHD)], and cerebrovascular diseases (10.2%). In men, leading causes of death were CVD (18.7%) including 35.1% IHD, dementia (17.5%) and cancer (11.6%). Age-adjustment did not yield a statistically significant difference between causes of death in both care settings (data not shown).

Underlying causes of death among patients with AD as compared to VaD

Among men and women with AD, the risk to die from CVD slightly increased with increasing age. A similar pattern was observed for the risk to die from dementia in both sexes. The latter was also found among patients with VaD. The risk to die from cancer, however, decreased with increasing age in both sexes and dementia subtypes. Since numbers of death stratified by subtype, age, and sex were rather low, we were not allowed to report them in line with the Dutch data protection guidelines. Therefore, Table 3 shows underlying causes of death and relative risks for patients with AD versus VaD merely stratified by sex. Men and women with AD died less often from cerebrovascular diseases as compared to VaD patients (RR for women 0.66, 95% CI 0.62–0.71, RR for men 0.54, 0.50–0.60). There were no significant differences with respect to CVD as underlying cause of death. Patients with AD died more often of cancer compared to patients with VaD (RR for women 1.32; 95% CI 1.21–1.43 and for men 1.37; 95% CI 1.24–1.50).

Underlying causes of death in patients with dementia as compared to the general population

Table 4 shows the distribution of causes of death in patients with dementia as compared to the general population stratified by sex. Although CVD is among the leading causes of death in patients with dementia, it was less frequently observed as compared to the general population (age-standardized RR in men 0.67, 95% CI 0.65–0.68; age-standardized RR in women 0.73, 95% CI 0.71–0.75). Cancer and chronic respiratory diseases were also a less common cause of death as compared to the general population (age-standardized RR for cancer in men and women 0.40, 95% CI 0.39–0.41 and 0.41, 95% CI 0.40–0.42, respectively, and for chronic respiratory diseases RR 0.89, 95% CI 0.85–0.93 and RR 0.75, 95% CI 0.71–0.80).

Figure 1 shows the differences per five-year age-group in underlying cause of death between patients with dementia and the general population (since results for men and women are comparable, Fig. 1 merely shows results for men; results for women can be seen in Supplementary Figure 1). The age-specific differences were more pronounced in younger age groups. A comprehensive table with all RRs stratified per sex and age-group can be found in Supplementary Table 1. We observed no differences regarding the distribution of subtypes of cancer between patients with dementia and the general population (data not shown).

More common causes of death were dementia and infectious diseases (pneumonia and genitourinary diseases). While 17.5% of all men with dementia died of dementia, 3.5% of men in the general population died of dementia (age-standardized RR 4.65, 95% CI 4.43–4.88). In women the difference was 23.7% versus 8.2% (age-standardized RR 2.82, 95% CI 2.74–2.91). The largest differences were found in the youngest age groups. Similar age-specific patterns were also found for pneumonia and mortality from genitourinary diseases in which 56.5% consisted of urinary tract infections.

With respect to the remaining underlying causes of death (cerebrovascular diseases, chronic respiratory diseases, and gastrointestinal diseases), differences between men and women with dementia were less pronounced as compared to the general population.

DISCUSSION

In the present study using a large nationwide cohort of patients with dementia, the most important underlying cause of death were dementia itself and CVD. The latter was less frequently observed as compared to the general population. When compared to the general population dementia as cause of death was more often and cancer remarkably less common. These differences were strikingly more pronounced at younger age. Regarding dementia subtypes, patients with VaD died more often of cerebrovascular diseases and less often of cancer as compared to patients with AD. We observed no statistically significant differences with respect to setting of care.

Although dementia is not an acute life-threatening disorder, the risk of death is high and dementia was among the leading causes of death. As dementia progresses patients become increasingly frail, which gives rise to complications, such as swallowing impairment, weight loss, and incontinence. Consequently, dementia results in a chain of negative events leading to death (the underlying cause of death). Previous studies reported percentages of dementia as underlying cause of death ranging from 7.2 to 23.5%, which is in line with the results found in this study. These studies even found higher RRs when comparing demented patients (community-dwellings with mean age 85.5 years [SD 7.2] [8] and patients recruited from a memory clinic with mean age 78.6 years, [SD 7.5] [10]) with controls (RRs 11.1 and 11.6) [8, 10]. These studies did not provide age specific causes of death. Especially in the relatively young olds the risk to die from dementia is seriously increased compared to non-demented peers.

Although CVD was a less common underlying cause of death than in the general population, it was the leading cause of death in men with dementia and the second leading cause of death in women in all age categories. Two studies have found CVD to be among the leading underlying cause of death as well [8, 11], whereas another study have found CVD to be the fifth leading cause of death [10]. RRs in comparison with the general population ranged from 0.37–0.83. Two of these studies also demonstrated a remarkable difference with respect to cancer as underlying cause of death among patients with dementia compared to controls (RRs ranging from 0.35 to 0.86) [8, 10]. This difference was also underlined by the results of the present study. However, we additionally showed that the RR was remarkably more pronounced at younger age. There was no difference regarding the distribution of cancer subtypes; all subtypes were less likely to be the underlying cause of death in patients with dementia as compared to the general population. There are several proposed explanations for the relatively low numbers of death due to cancer in demented patients. First, cancer may not be detected as symptoms of a malignancy are not adequately recognized in patients with cognitive disorders, or physicians might be less willing to thoroughly search for cancer in demented individuals. Furthermore, varying not yet completely unraveled, pathophysiological mechanisms have been proposed, but a discussion thereof is beyond the scope of this article [23 –26].

Mortality from pneumonia was more common among patients with dementia than in the general population, especially at younger ages. Comparable with our study, Chamandy et al. [8] found an overall RR of 2.2 (95% CI 1.54–3.27), whereas Todd et al. [10] found a RR of 0.71 (95% CI 0.37–1.36). The latter RR might be explained by the selection of the controls. The control group in this particular study was recruited from an ortho-geriatric rehabilitation unit and geriatric day care hospital, which is probably not a representative group of the general population. Studies focusing on immediate causes of death (defined as the disease that leads directly to death) have found higher frequencies of pneumonia as the cause of death ranging from 26.6% to 47.3% [6 , 12].

Our results concerning underlying causes of death among patients with AD versus VaD are in line with a previous study [8]. This study reported an even more pronounced decreased risk of cerebrovascular mortality in patients with AD versus VaD (OR 0.32, 95% CI 0.19–0.55). The difference was not unexpected since cerebrovascular pathology contributes substantively to the occurrence of VaD. They also found a decreased risk of cancer mortality in patients with VaD relative to AD.

In addition to previous studies we provided age and sex specific RRs to further guide ACP. The risk to die from dementia was increased at any age compared to the general population, but particularly at younger age. This means that ACP in all patients with dementia should have a focus on preferences and goals with respect to progression of disease. Pneumonia was a remarkably more common cause at younger age. It is therefore important to discuss initiation of antibiotic therapy in case of pneumonia, also in the relatively young olds. CVD as underlying cause of death showed a similar age-distribution as the general population. Furthermore, CVD is a relevant cause of death at any age and should be therefore seriously considered in ACP (e.g., discussion on whether or not to resuscitate). Cancer was remarkably less common and is therefore of less importance in ACP as compared to CVD for patients with dementia.

Strengths and limitations

The strengths of this study are the relative large sample size, the complete follow up of the entire cohort, and the stratification by age, sex, and dementia subtype. The validity of the linkage of registries in The Netherlands has been proved to be high [18, 19]. Another strength is the high validity of the ICD-9 code to identify patients with dementia in the Dutch HDR, which has been shown in a previously performed study [20]. During total follow-up, more than 60% of our patients died. This high percentage is mainly driven by the high mortality risk among patients hospitalized with dementia. In a previous study, we found a remarkable difference in prognosis between patients visiting a day clinic and those hospitalized with dementia. However, this will not influence our results since causes of death between both groups did not differ significantly. The study might be limited in that the accuracy of death certificates in the geriatric population is low which has been shown in literature [6 , 27]. However, these studies focused on immediate causes of death showing that pneumonia and dementia are underestimated. Less is known about the accuracy of underlying causes of death. In The Netherlands, autopsies are not frequently performed, especially not in patients who died outside the hospital since costs attached will not be reimbursed by insurance companies. As a consequence, information to perform a validation study on the accuracy of underlying causes of death registered on the death certificates is scarce. An urgent need to investigate the accuracy of death certificates with autopsy information in the near future still remains. In our paper, we compared causes of death between patients with dementia and non-demented peers with information derived from the same registry. It is not likely that misclassification or accuracy differs between both groups.

Furthermore, generalizability of results is restricted to patients with dementia visiting a hospital. This means that results are applicable to approximately 22% –30% of the patients with dementia in The Netherlands based on referral rate and incidence of the disease.

Clinical implications

Insight in age and sex specific underlying causes of death is valuable for patients, carers, and physicians in personal end-of-life care. And although causes of death differ not that much between patients with or without dementia, the presented results can certainly help the direction of end-of-life care. Given that CVD is among the leading causes of death in patients with dementia at any age, and the relatively high frequency of pneumonia at younger ages as underlying cause of death, we urge for timely and targeted ACP for patients with dementia. More attention is required not only for the oldest olds, it is even crucial for the relatively ‘young olds’ with dementia since differences with respect to the general population were more pronounced at younger age.

To illustrate the importance of causes of death in ACP we outline two scenarios. First, a 64-year-old woman visits the day clinic to discuss her preferences with respect to terminal care management, because she has recently been diagnosed with dementia. She has a 20% risk to die from dementia, 16% to die from CVD, and 17% to die from cancer, whereas the risk for a non-demented peer is 1%, 15%, and 54%, respectively. In contrast to her non-demented peer, the discussion on ACP should have a primary focus on preferences and goals when dementia progresses. Since CVD is an important cause of death in patients with dementia and in non-demented peers, it should be considered as an important topic of ACP in both groups of patients. And although cancer is obviously less common in dementia as compared to non-demented patients, it still is an important cause of death and should therefore also be discussed in ACP. Another example is an 85-year-old man with dementia. He has a high risk of dying from dementia itself (24%) or cardiac disease (20%). In agreement with his relatives, he has decided to refuse any further therapy if his situation deteriorates including preventive cardiovascular measures and cardiopulmonary resuscitation. These examples show that insight in causes of death offers specific guidance for ACP and that ACP is inevitably based on patients’ preferences and goals in combination with the expected prognosis.

Yet, uptake of ACP is known to be low among people with dementia [13, 14] while prognosis is poor. Decision-making in an acute setting concerning diagnostic and therapeutic interventions (including living wills, instructions regarding antibiotic therapy, resuscitation, or admission to an intensive care unit) is less complicated within the context of a previous discussion on preferences and personal goals of terminal care and management.

CONCLUSION

In conclusion, the distribution of underlying causes of death in patients with dementia differs from that of the general population, as dementia is more often and cancer less often an underlying cause of death and they are more pronounced at younger age. Although less frequent compared to the general population, CVD is also one of the leading underlying causes of death in patients with dementia. This information stresses the importance of targeted advance care planning in patients with dementia even at a younger age.

Footnotes

ACKNOWLEDGMENTS

This study was financially supported by Alzheimer Nederland (project no WE.03-2012-38).Dr. Vaartjes was supported by a grant from the Dutch Heart Foundation (grant DHF project ‘Facts and Figures’).

Authors’ disclosures available online ().

Appendix

References

Weuve

, Hebert

, Scherr

, Evans

(2014) Deaths in the United States among persons with Alzheimer’s disease (2010-2050). Alzheimers Dement 10, e40–e46.

James

, Leurgans

, Hebert

, Scherr

, Yaffe

, Bennett

(2014) Contribution of Alzheimer disease to mortality in the United States. Neurology 82, 1045–1050.

Fargo

, Bleiler

(2014) Alzheimer’s Association report. Alzheimers Dement 10, e47–e92.

Poos

(2013) Sterfte aan de ‘VTV-ziekten’ naar geslacht in 2011, RIVM, Bilthoven.

Minino

(2013) Death in the United States, 2011. NCHS Data Brief (115) 1–8.

Attems

, Konig

, Huber

, Lintner

, Jellinger

(2005) Cause of death in demented and non-demented elderly inpatients; an autopsy study of 308 cases. J Alzheimers Dis 8, 57–62.

Brunnstrom

, Englund

(2009) Cause of death in patients with dementia disorders. Eur J Neurol 16, 488–492.

Chamandy

, Wolfson

(2005) Underlying cause of death in demented and non-demented elderly Canadians. Neuroepidemiology 25, 75–84.

Kammoun

, Gold

, Bouras

, Giannakopoulos

, McGee

, Herrmann

, Michel

(2000) Immediate causes of death of demented and non-demented elderly. Acta Neurol Scand Suppl 176, 96–99.

10.

Todd

, Barr

, Passmore

(2013) Cause of death in Alzheimer’s disease: A cohort study. QJM 106, 747–753.

11.

Keene

, Hope

, Fairburn

, Jacoby

(2001) Death and dementia. Int J Geriatr Psychiatry 16, 969–974.

12.

Beard

, Kokmen

, Sigler

, Smith

, Petterson

, O’Brien

(1996) Cause of death in Alzheimer’s disease. Ann Epidemiol 6, 195–200.

13.

Taylor

, Cameron

(2002) Advance care planning in Australia: Overdue for improvement. Intern Med J 32, 475–480.

14.

Meeussen

, Van den Block

, Echteld

, Boffin

, Bilsen

, Van Casteren

, Deliens

(2012) Older people dying with dementia: A nationwide study. Int Psychogeriatr 24, 1581–1591.

15.

(1979) The International Statistical Classification of Diseases, Injuries and Causes of Death. Ninth Revision. Clinical Modification Department of Health and Human Services, Washington, DC.

16.

(1992) The International Statistical Classification of Diseases, Injuries and Related Health Problems. Tenth revision World Health Organization, Geneva.

17.

Harteloh

, de Bruin

, Kardaun

(2010) The reliability of cause-of-death coding in The Netherlands. Eur J Epidemiol 25, 531–538.

18.

Paas

, Veenhuizen

(2002) Research on the Validity of the LMR [in Dutch], Prismant, Utrecht.

19.

De Bruin

, Kardaun

, Gast

, de Bruin

, van Sijl

, Verweij

(2004) Record linkage of hospital discharge register with population register: Experiences at Statistics Netherlands. Stat J UN Econ Comm Eur 21, 23–32.

20.

van de Vorst

, Vaartjes

, Sinnecker

, Beks

, Bots

, Koek

(2015) The validity of national hospital discharge register data on dementia: A comparative analysis using clinical data from a university medical centre. Neth J Med 73, 69–75.

21.

Reitsma

, Kardaun

, Gevers

, de Bruin

, van der Wal

, Bonsel

(2003) Possibilities for anonymous follow-up studies of patients in Dutch national medical registrations using the Municipal Population Register: A pilot study. Ned Tijdschr Geneeskd 147, 2286–2290.

22.

Statline. Den Haag: Centraal Bureau voor de Statistiek. [Internet].; 2013. Available from: https://statline.cbs.nl.

23.

Behrens

, Lendon

, Roe

(2009) A common biological mechanism in cancer and Alzheimer’s disease?. Curr Alzheimer Res 6, 196–204.

24.

Ganguli

(2015) Cancer and dementia: It’s complicated. Alzheimer Dis Assoc Disord 29, 177–182.

25.

Boonen

, van Tijn

, Zivkovic

(2009) Wnt signaling in Alzheimer’s disease: Up or down, that is the question. Ageing Res Rev 8, 71–82.

26.

Driver

, Lu

(2010) Pin1: A new genetic link between Alzheimer’s disease, cancer and aging. Curr Aging Sci 3, 158–165.

27.

Bordin

, Da Col

, Peruzzo

, Stanta

, Guralnik

, Cattin

(1999) Causes of death and clinical diagnostic errors in extreme aged hospitalized people: A retrospective clinical-necropsy survey. J Gerontol A Biol Sci Med Sci 54, M554–M559.