There were 9447 individuals 70 years and older interviewed at baseline. Of these persons, follow-up information could be obtained on 7998 (84.7%), of which 2338 (29.2%) were proxy interviews. For another 47 individuals, contact was achieved but information about NH use was missing. Covariates included in the final statistical model were missing for another 115 persons, leaving 7836 people in the primary analysis. There were 938 (11.7%) persons with decedent interviews. Of these individuals, 678 were excluded, who died and did not use NHs, leaving 7158 for the secondary sensitivity analysis. The only statistically significant differences between persons with missing NH use information and those for whom information could be obtained was higher mortality in the former group (20% compared with 15% [P<.001] and higher proportions of black and other races [9% vs 7%, 7% vs 4%, respectively, P<.001], table presenting these data is not shown).
From the cohort of 7836 individuals, there were 707 persons (9%) who used NHs within 2 years of their baseline interview. The weighted proportions of men and women in the sample were 39.9 and 60.1, respectively. Details about this population and the proportions of persons who used NHs according to each variable are presented in . The weighted proportion of people who used NHs increased progressively from 5.8% for persons at ADL stage 0 (no limitation) to 30.9% at ADL stage III (severe), and declined to 17.9% at ADL stage IV (complete limitation).
NH Use in a Cohort of 7836 Community-Dwelling Persons 70 Years and Older
presents ORs for all NH use models beginning with unadjustment and showing increasing adjustment through models I, II, and III. All hypothesized associations were statistically significant. ORs associated with each variable were adjusted for all other variables, shown in the fully adjusted model III. The OR associated with the perception of home environmental barriers versus no perception of home environmental barriers was 2.80 (95% CI, 2.15–3.66) before and 1.43 (95% CI, 1.05–1.96) after full adjustment. The OR associated with living alone versus not living alone was 1.83 (95% CI, 1.53–2.19) before and 1.42 (95% CI, 1.15–1.75) after adjustment. In the fully adjusted model, the OR (95% CI) of NH use was 1.56 (1.24 –1.98), 2.17 (1.62–2.91), and 3.12 (2.20 – 4.41) for stages I, II, and III, respectively, compared with stage 0, but then declined to .96 (.33–2.81) for stage IV.
Relationship Between Sociodemographics, Diagnostic Category, ADL Stages, and NH Use
The C statistic for model III, the final model, was .779 (95% CI, .760 –.798) compared with .747 (95% CI, .727–.768) for model I and .754 (95% CI, .734 –.774) for model II.
After entering all available predictors in our dataset in a model, then using backward selection to remove insignificant ones (ie, high school graduation), only the diagnoses that were previously hypothesized to be strongly associated were positively associated with risk of NH use. Associations with ADL stage and NH use were stronger after excluding those who died before wave II and did not use NHs. As in the primary analysis, ADL stage IV was not statistically significantly different from stage 0, suggesting robustness of this finding. Directions of association did not change for the key variables of interest including living alone, perceived home environmental barriers, or stage in the sensitivity analyses setting missing to NH use or no NH use. The effect sizes were reduced particularly when we assumed all those with missing NH use used NHs (rather than no use) (). Finally, the effect sizes associated with perceptions of home environmental barriers did not change after forcing high school graduation into the final model.