In the model with demographic and health characteristics, circulatory disease patients in the 1994 survey were significantly less likely to be disabled and dead at follow-up compared to the 1984 survey (Table 9). When we include treatment covariates, we explain approximately 18% of the association between 1994 survey year and death, and the 1994 survey year coefficient on death is no longer statistically significant. The 1994 survey year coefficient on disability becomes more negative with the inclusion of treatment variables, i.e. -0.38 vs. -0.37, and remains statistically significant. Relevant procedures are not associated with either disability or death in this model. AMI treatments are associated with disability (p=0.044), but not with death (p=0.146). Predicted outcomes rates by AMI treatments show that the probability of disability declines by 13 percentage points in 90th percentile areas from 32% to 19% and increases by 8 percentage points from 25% to 33% in 10th percentile areas (Figure 12).
Interpreting the Results
Use of effective treatments contributed to the decline in disability and death among cardiovascular disease patients. The association between treatment use and reduced disability and death was strongest in the ischemic heart disease model, explaining over half of the decline in disability for the 1994 cohort and 65-70% of the decline in mortality for the 1989 and 1994 cohorts. With the exception of the stroke model, use of AMI treatments was significantly associated with declining disability. In the all-CVD and stroke models, AMI treatments were significantly associated with reductions in mortality. Increased use of relevant procedures resulted in lowered disability among stroke patients as well as lowered death in the all-cvd and ischemic heart disease models.
Improved medical treatment after an acute cardiovascular event resulted in improved survival and reductions in disability. It may also affect medical spending. While we cannot do a complete evaluation of the impact of these changes, we can provide some information. We begin with the change in quality-adjusted life expectancy. To consider how reductions in disability in one year translate into long-term changes in quality-adjusted life expectancy, we estimate regression models for future survival and disability status as a function of disability in a base year. For a cohort in year t, we estimate linear probability models of the form:
The coefficient у indicates how changes in disability in year t affect long-term health outcomes. Xit is a set of demographic and health status variables. The identifying assumption in equation is that people who are not disabled because of medical treatment are subsequently equivalent in their health to those who never had an incident. This may or may not be the case. If this is not true, and survivors of events are less healthy, conditional on disability status, we will overstate the benefits of reductions in disability. Thus, one should properly view these estimates as an upper bound on the impact of medical interventions to reduce disability. buy wellbutrin online
Figure 13 shows the survival rate by year, conditional on disability status in the base survey year. We report the results for the 10 years after the survey for the 1989 cohort, and the five years after the survey for the 1994 cohort due to data limitations on long-term follow-up. Not surprisingly, survival for the disabled is below that for the non-disabled, by a large margin. The difference is about 20 percentage points, and remains at that level throughout the decade.