We conclude that it is not currently feasible to set up a multicentre registry in the way that we had envisaged. We did not identify all patients who were eligible for the study: patients with unobstructed arteries (i.e. those that had an emergency angiogram but did not receive PPCI) were difficult to identify from catheter laboratory databases. We successfully consented patients but obtaining individual, opt-in consent would not be feasible for a national registry. We explored several consent models (e.g. opt-out on procedural consent form and standard consent at discharge) but none was implemented because of logistic difficulties. Linkage of data from HISs with hospital episode data was feasible, but data from HISs are not uniformly available/exportable. Information about whether or not participants had had CMR in CMR hospitals was obtained successfully from HISs, although some referrals for CMR were for research rather than clinical purposes.27
In the cost-effectiveness models, the differences in QALYs between strategies were very small, and the results were driven largely by the differences in costs. Sensitivity analyses around the two models identified the diagnostic accuracy of the ischaemia tests as the key driver of cost-effectiveness.
We defined five subgroups of patients who activate the PPCI pathway for whom there was consensus that CMR changes patient management in a clinically important way.4 All subgroups could be identified in either the local HIS data set, hospital episode data or both. It is feasible to identify some, but not all, important changes in management in follow-up hospital episode data. The main constraints on identifying important changes in management were as follows: (1) outpatient hospital episode data were poorly coded with respect to diagnosis, (2) medication data were not available at baseline or follow-up (preventing inspection of prescription of/changes in medication after the index event) and (3) data about diagnostic investigations were poorly coded in both acute care and outpatient data sets. More research would be needed to develop a specific composite outcome for a patient subgroup.68
Any multicentre registry set up in this patient population would need to be able to capture new and emerging diagnostic/interventional technologies. During the study, priority for CMR decreased because of the introduction of pressure-wire testing for diagnosing ischaemia in patients with multivessel disease. However, it is not clear whether or not pressure-wire testing is the optimal management strategy from the perspective of health services. This question could potentially be answered in a national registry, given that the use of pressure wire is (currently) driven largely by the preferences of interventional cardiologists rather than patient characteristics. However, we are uncertain whether or not CMR is used sufficiently to allow this question to be addressed.