Table 5 shows demographic characteristics of our cohort by year of baseline survey. The proportion of respondents disabled at baseline declined over time, from 32% in 1984 to 25.9% in 1994. This 6 percentage point decline in disability at baseline is consistent with other analyses using the NLTCS (Manton and Gu 2001). In addition, there was a slight increase in the mean modified Charlson index score from 0.94 in 1984 to 1.07 in 1994.
Coefficients and standard errors from our estimation results are shown in tables 6-10 for models with all cvd patients, ischemic heart disease patients, stroke patients, heart failure patients and other circulatory disease patients. Each table includes a model with only survey year and demographic characteristics as well as a model with area-level relevant procedures, appropriate pharmaceutical use and interactions between year and time-invariant treatment covariates.
In the all-cvd models, the average respondent from the 1994 survey was significantly less likely to be disabled and dead compared to the average respondent to the 1984 survey, when adjusting for only demographic and health status covariates (Table 6). When we include treatments covariates, the average respondent living in an area with average treatment rates in the 1994 survey was significantly less likely to experience disability and death at follow-up survey compared to the average respondent in an average area in 1984, but inclusion of treatment covariates moderated the associations. In particular, area-level use of treatments explains approximately 3.5% and 29% of the difference in disability and death outcomes, respectively, between the 1994 and 1984 cohorts. Relevant procedures were significantly associated with a decreased likelihood of death (p= 0.044), but not disability (p=0.756), while all AMI treatments together were significantly associated with disability (p= 0.001) and death (p=0.020). Link
Predicted event rates for all cvd patients in areas at the 10th and 90th percentiles are shown in Figures 5 and 6. For relevant procedures, the probability of death is lower at all time points for 90th percentile areas, and due to increasing rates of procedure use, it decreases faster for the 90th compared to 10th percentile over time. When examining time invariant AMI treatment variables together, we find that the probability of disability declined 6 percentage points from 28% in 1984 to 22% in 1994 for the 90th percentile and increased approximately 3 percentage points from 23% to 26% in 10 th percentile areas.