Table 1.
Model Variables, Coding Algorithms, and Mean Baseline Characteristics
| Mean for Observations
|
|||
|---|---|---|---|
| Baseline Characteristic and Model Variable | Coding Algorithm | Included in Analytic Sample (n = 3,129) | Excluded From Analytic Sample (n = 4,318) |
| SRH | 95 = excellent, 90 = very good, 80 = good, 30 = fair, 15 = poor | 72.8 | 57.6 |
| SRH trajectory | Calculated (see Methods section) | 2.88 | — |
| SRH trjectory SSE | Calculated (see Methods section) | 637.5 | — |
| Self-rated change in health statusa | 1 = much better, 2 = somewhat better, 3 = about the same, 4 = somewhat worse, 5 = much worse | 3.30 | — |
| Age | Age in years at baseline | 75.5 | 79.0 |
| Men (%) | 1 = man | 35.3 | 39.8 |
| White (%) | 1 = White | 88.0 | 82.4 |
| Live alone (%)b | 1 = live alone | 35.9 | 36.4 |
| High school graduates (%) | 1 = high school graduate | 67.7* | 49.3 |
| High income (%)c | 1 = income in top fourth or fifth quintile of all 7,447 AHEAD respondents | 64.6 | 46.8 |
| Religion very important (%)d,e | 1 = religion is very important | 67.7 | 65.0 |
| Ever smoked (%)f | 1 = former or current smoker | 51.0* | 52.4 |
| Ever drunk alcohol (%) | 1 = has drunk alcohol or currently drinks alcohol | 53.0 | 39.5 |
| Health conditiong | |||
| Hypertension (%)c | 1 = high blood pressure/hypertension | 43.7 | 47.6 |
| Diabetes (%)c | 1 = diabetes | 9.9 | 14.1 |
| Cancer (%) | 1 = cancer or malignant tumor (excluding minor skin cancers) | 11.3 | 14.1 |
| Hip fracture (%) | 1 = fractured hip | 3.0 | 6.6 |
| Lung disease (%)c | 1 = chronic lung disease | 6.2 | 11.9 |
| Heart condition (%)c | 1 = heart attack, coronary heart disease, angina, congestive heart failure, or other heart problem | 23.7 | 33.3 |
| Stroke (%) | 1 = stroke | 5.4 | 14.3 |
| Obese (%)c,h | 1 = BMI ≥ 30 kg/m2 | 15.1 | 12.6 |
| Number of ADL limitationsc | Sum of ADLs with which respondent needs help or has difficulty (0–6) | 0.161 | 0.531 |
| Number of IADL limitationsc | Sum of IADLs with which respondent needs help or has difficulty (0–5) | 0.159 | 0.851 |
| CES-D8 scorei | Sum of “yes” responses for 8 items from the Center for Epidemiologic Studies-Depression scale | 1.25 | 1.88 |
| Prior year health service use | |||
| Hospital admissionsc | Number of different times in a hospital overnight in the past 12 months | 0.214 | 0.468 |
| Nursing home admissionsc | Number of different times as patient in a nursing home in the past 12 months | 0.006 | 0.029 |
| Doctor visitsj | Number of times talked to a medical doctor about own health in past 12 months | 4.22 | 5.45 |
| Probability of living at least 10 more years (%) | Chances of living at least 10 years, from 0 to 100; 0 = absolutely no chance; 100 = absolutely certain | 45.9 | 36.7 |
| Refused, missing (%)k | 1 = refused, don’t know, missing, born before 1903 | 9.8 | 17.8 |
| Percentage proxy (%) | 1 = proxy respondent | 0.0 | 18.3 |
Notes: SRH = self-rated health; SSE = sum of squared errors; AHEAD = Asset and Health Dynamics Among the Oldest Old; BMI = body mass index; ADL = activity of daily living; IADL = instrumental ADL; CES-D = Center for Epidemiologic Studies–Depression scale. Comparison of means based on 1993 (baseline) interview responses. All p < .001 unless otherwise indicated.
Missing observations (n = 2) imputed based on changes in SRH between 1998 and 2000, if available; otherwise, set to “no change.”
p = .6703.
Missing observations set to sample mode.
p = .0134; compared to those indicating religion is either somewhat or not important.
Missing observations (n = 4) set to religion not important or religion somewhat important.
p = .2097.
Dichotomous marker based on respondents reporting ever having or having been told by doctor that they have the listed condition; 0 = not having condition.
p = .0019.
Higher score represents more depressive symptoms. Proxy respondents (none in analytic sample) did not answer this question.
Missing observations set equal to: Medianbaseline × (Visitswave# × Medianwave#); where wave# is first wave in which respondent’s doctor visits were available. If no data were available in any of the four waves (n = 2), set to sample mode.
This question was not asked of individuals born before 1903. Value = 0 if individual answered the question or if questioned answered by a proxy respondent (there were no proxy respondents on the analytic sample).
Not significant at the .05 level.