In the present large, prospective randomized controlled study in which all patients were followed for 19 months, a comprehensive, intensive program designed to optimize secondary prevention in cardiac patients failed to reduce the primary end points of rehospitalizations and total days in hospital per patient, as well as the secondary end point of mortality. We anticipated a positive result because of the known effectiveness of secondary prevention, as proven in other randomized placebo-controlled trials (1
). Given the near-optimal physician compliance with guidelines for secondary prevention in our hospital region, there was limited potential for improving appropriate prescription rates for secondary prevention medications; thus, the study was effectively an attempt to improve outcomes by optimizing patient adherence to therapy. Patient education and monitoring for adherence are widely perceived to be suboptimal in our region. Therefore, we had anticipated that our efforts to improve adherence would have a substantial effect on outcomes. Our results indicate that even our intensive efforts to improve adherence were not sufficient to incrementally improve outcomes when prescription rates were already near optimal.
We considered many factors that might have affected our results. The power of the trial should have been adequate to demonstrate a biologically important effect, given the accumulation of more than 3500 patient-years of observation. It is possible that the total duration of the study (19 months) was too short for benefits to be more apparent; however, given the cumulative proven benefits of secondary prevention, it is not clear that this explains our results. There was little potential for improved prescription rates, as indicated by the near-optimal rates in the usual care arm, except for those for lipid-lowering agents. To illustrate the effectiveness of the program with this well-tolerated class of drugs with clear indications, despite the high prescription rates in the usual care group (approximately twice that in other parts of the country and higher than most other reported rates during the same period [11
]), prescription rates were still somewhat higher in the intervention group. However, the differences between the high rate in the usual care group and the slightly higher rate in the intervention group were not expected to result in a substantial benefit within 19 months compared with what was expected versus placebo (53
). This is well illustrated by the results from a recent study (57
) that focused on lipid-lowering drugs; a very large number of patients needed to be studied for a longer time to demonstrate a significant but relatively small effect. Our patient-reported adherence to therapy data also suggest that there was only a small difference in adherence to beta-blocker and angiotensin-converting enzyme inhibitor prescriptions, which were not expected to result in detectable differences in outcomes. We also considered that study nurses might have initiated processes, leading to more readmissions than might have occurred if patients had not sought medical attention on their own. However, a detailed analysis of the reasons for rehospitalization did not support that hypothesis.
Post hoc analysis
Without prior consideration of the data, we reasoned that there might have been an important learning curve related to rapid growth of the study team, as well as a possible effect from the restructuring of cardiac services during the first half of the study. The post hoc analysis provided some intriguing data. The outcomes during the first half of the study were similar to those during the whole study period. However, the analysis also suggested that the program was more effective during the second time period –patients in the intervention arm spent substantially fewer days in hospital during the second half of the study. Statistical analysis of time-dependence of differences in the number of days in hospital also supports this interpretation of our results; there was a trend (P=0.08) suggesting increased effectiveness in the intervention group over time. Thus, the post hoc analysis suggested that after the study team had acquired sufficient skills to be most effective and the team was stable, with minimal staff turnover, an important difference in the number of days in hospital might have been achieved by the program with considerable cost savings.
If this interpretation of the results is valid, there are important implications regarding implementation of such programs. Our experience leads us to believe that such a program is complex and requires a stable team with considerable experience to be effective; the learning curve is perhaps as long as six to 12 months. Because this was not our primary, a priori, hypothesis, we cannot conclude a cause and effect relationship. Nevertheless, this finding is consistent with our observation that such programs are more difficult to implement than we had originally considered.
In contrast to patients involved in multicentre drug trials, our patients more likely reflected the mix normally encountered in many tertiary care institutions. Only patients with terminal diseases, those in nursing or long-term care facilities and those unable to communicate were excluded from the study; approximately 80% of those approached agreed to be in the study. Thus, the observed readmission rates and hospital days should have more closely reflected outcomes than in more patient-selective treatment trials (1
). Our data highlight the substantial cost burden related to rehospitalization of cardiac patients. Overall, there were more than 130 readmissions per 100 patients over 19 months, with approximately one-half occurring within six months. The vast majority of these were not elective. It is also noteworthy that data from the provincial database identified admissions to other hospitals that would otherwise have been missed, despite the detailed data collection by the study team from the patients themselves.
The random assignment was designed to study the effects of the intervention program on prescription rates, and to study the effects of the patient-directed interventions. There was a slight trend toward reduced mortality in the intervention arm (P=0.15), so a larger study might have shown a significant difference. There were almost no opportunities to cue the cardiologists, so simpler random assignment would have sufficed. Nevertheless, there were no differences among any of the characteristics between groups (in those listed and in many others considered), which supports the validity of the random assignment. Transfer of care of patients between cardiologists from both groups was frequent because of short clinical service blocks; thus, patients were frequently cared for by both usual care and intervention cardiologists. Furthermore, because there was no demonstrable effect of the attending cardiologists on the results, we suggest that although the reassignment of two cardiologists (after fewer than one-fifth of patients were enrolled) from the usual care to the intervention group might be considered to have been inappropriate, this should not have impacted the results. Although the study can be considered a cluster trial, the lack of interventions related to the prescription rates suggests that the trial can be considered to be a nonclustered study for analysis purposes.
Physician behaviour was not modified, and analysis of potential cardiologist-related effects also showed no effect, which supports the absence of a cluster effect. Secondary prevention in cardiac diseases has evolved since the study was performed and standards have changed. However, there is little reason to expect these kinds of programs to be more or less effective as guidelines evolve. The patient-directed component would still have the same goals when patient education and adherence to therapy are the main issues.
There are significant limitations to using patient-reported adherence to therapy data – some patients are reluctant to admit to failure to take medications as prescribed. However, the data remain consistent with a small, positive effect in adherence with some prescriptions in the intervention group.
Our results suggest that an intensive program designed to optimize secondary prevention in heart disease was ineffective in improving outcomes in our region in which the standard of care, measured by prescription rates and mortality from acute myocardial infarction, is very high (58
). However, the post hoc analysis of our results suggests that there may be an important learning curve in implementing such a program and that improved outcomes may be achievable with a well-trained team. Our results suggest that the environment in which a disease management program is being considered should be carefully evaluated and that it may be best to focus on areas that can be improved to maximize cost-effectiveness. Comprehensive programs in health regions where secondary prevention is not practised according to guidelines may be effective, whereas programs that focus only on specific goals, such as patient adherence to therapy, may suffice where the standard of practice is already high.