As described in more detail in (Stewart, Landrum et al. 2006), treatment covariates explain between 50 and 70% of the association between disability and death over time in the IHD cohort. Relevant procedures were significantly associated with death, but not disability, and pharmaceutical care was significantly associated with disability but not death Stroke patients in 1994 were significantly less likely to experience disability and death compared to patients in the 1984 survey in the model with only demographic and health status covariates (Table 7). Including treatment covariates explains approximately 42% and 23% of the association between 1994 survey year and disability and death, respectively. The coefficient on 1994 survey year for disability is not statistically significant in the model with treatments, suggesting that in areas with average levels of treatment there were no significant declines in disability. The coefficient on 1994 survey year for death remains statistically significant, suggesting that additional factors are important in explaining the decreased probability of death in 1994 relative to 1984.
Use of relevant procedures for stroke patients is significantly associated with lower likelihood of disability (p = 0.010), but not with death (p=0.157). Use of AMI treatments measured at the HRR-level from the CCP data and increased use of warfarin and anti-thrombolytics measured at the state level are together significantly associated with death (p = 0.001) but not disability (p=0.175). These results are apparent in predicted outcome rates displayed in Figures 7 and 8. The predicted probability of disability for stroke patients in 90th percentile areas of relevant procedures falls from 29% in 1984 to 25% in 1994 (?) and the probability of disability in 10th percentile areas rises from 33% in 1984 to 36% in 1989 and back to 33% in 1994 (Figure 7). In areas at the 90th percentile for AMI and stroke treatments, the probability of death declines from 48% in 1984 to 35% in 1989 and then increases to 40% by 1994 (Figure 8). The trend is opposite in 10th percentile areas, with the probability of death increasing from 43% in 1984 to 51% by 1989 and then decreasing to 42% by 1994.
Heart Failure and Arrhythmia
Table 8 shows that the heart failure and arrhythmia patients in 1994 were significantly less likely to die by follow-up compared to patients in 1984 in the model with only demographic and health status covariates. Including treatment covariates strengthens the association between survey year and death, i.e. the average respondent in the average area is even less likely to die in 1994 compared to 1984. Relevant procedures for heart failure and arrhythmia are not associated with either disability (p = 0.903) or death (p = 0.574). Use of AMI treatments and evaluation of ejection fraction are significantly associated with disability (p=0.043) but not death (p=0.087). buy starlix online
As expected based on estimation results, there is little difference in the probability of death and disability between 10th and 90th percentile areas of relevant procedures (Figure 9). In examination of the AMI and heart failure treatments (Figure 11), the probability of disability is higher in 90th percentile areas in 1984 and 1989 compared to 10th percentile areas. By 1994, the probability of disability is much lower in the 90th percentile areas (19%) compared to the 10th percentile (26%).