Cost comparison is usually done badly

On-premise costs are visible and lumpy: hardware, licences, a refresh every few years. Hosted costs are smooth and recurring. Comparing them fairly means including the on-premise costs that are usually hidden in other budgets — data centre space, power, the proportion of infrastructure staff time spent on that system, and the cost of the capacity you bought but do not use. Most naive comparisons flatter on-premise because those costs sit somewhere else.

Recovery is the question that actually matters

Ask how long your service can operate without the system, and how much data you can afford to lose. Those two numbers — recovery time and recovery point — determine more about the right answer than cost does. Then ask, honestly, whether your current arrangement has ever been tested by performing a real recovery. A backup job that completes is not evidence of anything. Many organisations discover their recovery position is worse than they believed, and that discovery should happen in a test, not an incident.

Data residency is a constraint, not a preference

Where patient data may physically reside is determined by regulation and by your own information governance position, not by architectural preference. Establish it early and in writing. It rules out some options entirely, and finding that out late has derailed otherwise sound migrations.

Connectivity becomes a clinical dependency

A hosted clinical system makes your network connection part of the care pathway. That is manageable, but it must be designed for: resilient links, a defined degraded mode, and clarity about what clinical staff do when connectivity fails. Sites with weak connectivity are a genuine argument for keeping systems local, and it is worth checking the real position at every site rather than the head office.

Hybrid is a legitimate answer

The choice is often presented as binary and rarely is. Keeping a system local because it has a hard latency or connectivity requirement, while moving others, is a reasonable outcome rather than a failure to commit. What matters is that each decision has a stated reason you could defend later.