Questions of Life and Death in Old Age

Like the world's population, patients and physicians are living longer, leading to increased interest in how to live well and die later. The article by Fried and colleagues¹ in this issue of THE JOURNAL reports historical and measured attributes as risk factors for death in the Cardiovascular Health Study (CHS), a study conducted in the homes of community-dwelling men and women aged 65 years and older who were followed up for 5 years. The authors make 2 main points: (1) disease markers or quantitative measures are better predictors of death than the medical history; and (2) older people frequently have multiple conditions rather than a single predictor of death.

It is not surprising that objective measures of disease replace the medical history as more accurate predictors of death. As the authors note, good diagnostic tests have fewer false positives and false negatives than does the history (otherwise clinicians would not use them), and afford the detection of subclinical disease and a measure of disease severity. Many common diseases are unrecognized without specific tests; for example, half of elderly patients with diabetes and a majority with osteoporosis do not know they have it.^{2 - 3} In a community-based study from Italy, more than one third of patients with heart failure, arrhythmia, myocardial infarction, or peripheral arterial disease did not report the diagnosis on interview.⁴

However, in the absence of medical records, it is impossible to differentiate the superiority of diagnostic testing from the inferiority of information provided to or recalled by the patient. Older patients cared for during years of well-intentioned but paternalistic medical care may not have been informed of important information or details regarding their health, perhaps to spare them anxiety. In the past, the diagnosis frequently was shared only with the husband of the older woman, and at times only in the vaguest terms. A woman in her 70s recently told us that, when her physician performed her hysterectomy when she was 40 years old, he simply told her husband that she "had lost the nursery but not the playroom." Neither partner knew whether an oophorectomy had accompanied the hysterectomy.

Unfortunately, the design of the CHS cannot tell whether chart review would be a better predictor than other extensive measures or what proportion of the participants were not informed vs those who did not remember their diagnoses. The absence of such information may incorrectly imply the need for more laboratory tests for optimal preventive and treatment strategies, when instead providing each patient with an annually updated copy of his or her medical record would better serve the purpose.

In the CHS, women had a 43% lower mortality risk than men, a difference not explained by atherosclerosis as diagnosed by electrocardiogram, echocardiogram, or a history of heart failure. Unfortunately, whether women were treated with estrogen replacement therapy is not mentioned.

While it is not remarkable that the female survival advantage persists into old age (even without estrogen, women do not "catch up to men," contrary to popular belief),⁵ it is surprising that the sex difference identified by the authors is not associated with atherosclerosis. Could this be a result of using all-cause mortality rather than cardiovascular deaths as the outcome? More than half of all deaths in men and women in the United States are attributed to cardiovascular disease, and cardiovascular deaths are expected to parallel deaths from all causes. However, this may not be true in the elderly. Autopsies reveal no clear-cut cause of death in nearly one third of persons aged 85 years and older.⁶ It is unlikely that death certificate diagnoses (planned for a subsequent CHS publication) will be very enlightening. This is because the law demands that a cause of death be chosen from a standard list to be reported on the death certificate; therefore, death may be assigned without evidence for a common cause, such as cardiovascular disease.

Another finding from the CHS was that participants who weighed more in old age were less likely to die. Similar results were reported by Ostfeld in 1972⁷ and more recently by Dorn et al⁸ and Stevens et al.⁹ Physicians who treat elderly patients will not be surprised that increasing age weakens the association of body size with death. This paradox generally is explained on the grounds that underweight in the elderly is a marker for chronic disease, frailty, depression, or dementia.¹⁰ Conversely, there are multiple plausible reasons that moderate adiposity might improve survival: it may reflect social supports (eg, a partner to cook or cook for), a more adequate diet, the anabolic effects of hyperinsulinemia, or the protection from hip fracture afforded by padding over the greater trochanter.

Although it is well known that elderly people rarely die of a single cause, the observation that 20 variables contributed independently to mortality risk in CHS is surprising. These 20 factors included both nonmodifiable conditions (eg, age, sex, income, serum creatinine level, poor pulmonary function, major electrocardiographic abnormalities, congestive heart failure by history, and impaired cognitive function), as well as a number of factors that would be amenable to behavioral intervention (eg, smoking, lack of physical exercise) or medical intervention (eg, hypertension or carotid artery stenosis).

How should these findings alter clinical practice? Among the 20 variables that predicted mortality, at least 12 are modifiable. But, except for smoking cessation, exercise, and blood pressure control, there is little evidence that modifying these variables will improve outcome in old age. Important considerations involve whether interventions are equally appropriate at age 65 and 85 years, and whether interventions should vary by chronological age or physiological age. Moreover, it is unclear whether the CHS findings can be extrapolated to other populations. The CHS participants were younger and healthier than elderly persons who refused to participate in the study,¹¹ and the CHS death rate seems relatively low—only 12% of an age-stratified sample with a mean age of 73 years died within 5 years, compared with a 5-year death rate of 43% in a study of elderly persons in New Zealand.¹²

Everyone dies too early or too late. For the elderly, the predictors of death are perhaps less important than the quality of life before death. The CHS investigators have the opportunity to use their rich database and considerable skills to study the variables that predict a healthy old age.