Background Results

3. Data and Methods

3.1 The 1984 and 1994 Surveys of the Aging Population (SOA)

Data used come from the 1994-95 and 1984 Supplements on Aging to the National Health Interview Surveys (SOAI and SOAII). These surveys were designed to provide information on the health of the community dwelling population of the United States 70 years of age and over at two dates - 1984 and 1994 [Note 1]. Both surveys were designed and completed by researchers at the National Centers for Health Statistics with support from the National Institute on Aging. In 1984, 7541 persons 70 years of age and over were included in SOAI and the number for SOAII was 8767. Because of the complex sample design used to select respondents, the data are weighted to reflect the U.S. noninstitutionalized population 70 years of age and over.

In both surveys people were asked about the presence of diseases in an almost identical manner. Questions used to elicit this information are shown in Table 1. We examine trends in three sets of diseases or conditions: mortal diseases, morbid conditions and impairments. Mortal conditions, which are major causes of death among people this age, include heart disease, stroke, cancer, and diabetes. Hypertension is also included in this category as it can be considered both a disease and a risk factor for other diseases. Morbid conditions include arthritis, osteoporosis, broken hip, and glaucoma. Most of the impairments represent sensory impairments: cataracts, blindness, deafness, and other hearing trouble. We also include the symptom dizziness, which could be the result of a large number of underlying conditions. While the surveys were designed to be compatible, for some questions the wording was changed at the second survey. The most important change is in the question about heart disease, for which the four subquestions asked in 1984 were combined into a single question in 1994.

In order to examine severity of conditions, we look at the functioning loss and disability level among persons with each condition. Disability and functioning in the survey is measured by a series of questions on ability to perform activities of daily living (ADLs), instrumental activities of daily living (IADLs), and a set of functions called Nagi functions. Inability to perform personal care is indicated by inability to perform the seven ADL tasks including bathing, dressing, toiletting, eating, getting in and out of bed or chair, walking, and getting outside. Inability in performing the five IADL tasks (including preparing meals, managing money, using the telephone, heavy housework, and light housework) indicates additional inability in the area of self-maintenance. Finally we look at inability to perform a series of nine Nagi functions which represent physical functioning deficits that should precede ADL and IADL disability. These functions include walking a quarter of a mile, walking up ten stairs, standing for 2 hours, sitting for two hours, stooping, crouching or kneeling, reaching up over head, reaching out as if to shake someone's hand, using your fingers to grasp, lifting or carrying something as heavy as ten pounds.

3.2 Methods

In order to examine the significance of change over time in each disease, we pool data from the two surveys and look at the effect on disease presence of being in the later year in a series of sex-specific logistic regression equations including only two covariates: a continuous measure indicating year of age and year of survey indicated by a dummy variable. This use of the regression technique allows us to summarize the effect of time change across all ages. The models presented do not include age/year of survey interactions which would indicate age differences in time change; however, these models have been tested and results indicate that most results are consistent across ages from 70 through 90, therefore models without interactions have been presented. When the coefficient indicating change over time is statistically significant, we use the results of these equations to estimate the prevalence of diseases at age 70 for 1984 and 1994 to indicate the change over time. We use OLS regression to estimate disability and functioning problems among those with specified diseases. Again we estimate sex-specific regressions including age and survey year as the covariates. We also use OLS regression to determine change between 1984 and 1994 in comorbidity or the average number of diseases after controlling for age. Finally we use results of logistic regressions including age and year of survey to estimate the probability of having no disease or three or more diseases by type for each of the two dates.


Background Results

Change in the Prevalence of Diseases among Older Americans: 1984-1994
Eileen M. Crimmins, Yasuhiko Saito
© 2000 Max-Planck-Gesellschaft ISSN 1435-9871