The openEHR specifications can be applied to a different "modules" of an EMR/EHR. Let's say that we have one of those systems, created in the "normal" way: specifying requirements, designing an architecture and an information model based on those requirements, writing code that implements the architecture and information model (user interface, logic, persistence, services), testing and deployed.
Now we have a system that implements our very specific requirements, but we know those will change in the future.
What are the suggested phases or steps to make our custom system more openEHR?
Should we start supporting archetypes or operational templates and for what? Or should we focus on the database / persistence design based on the openEHR information model?
Depends on if you want to keep your original system untouched or not.
The fastest way is to create a wrapper over current systems. There are a handful of tools to achieve this, even some archeytpe-based ones. This keeps original systems untouched and allows you to keep using it as always.
If you want to transform your system to a full openEHR one, the first thing to do is to detect your current clinical models needs and find the equivalent openEHR archetypes in an archetype repository (and specialize if needed). Then you can safely generate an openEHR persistence layer and create your system from there. You probably must keep your current data and transform it to archetype-based data, so the use of a wrapper to transform the data is still needed.