You've read the situation right and you still have to touch them. Why the fight is usually a symptom, and how to get the assessment done anyway.
Notice that being struck changes you faster than thought. Nobody absorbs a hit from a patient and stays exactly as calm as they were a second earlier. One breath, and the next move being slower rather than firmer, is most of the skill on this call.
Start with the finding that should reorganize how you read this call: the behavior you are labeling combative is, in the standard clinical instrument, a pain score.
PAINAD, the validated observational pain scale for advanced dementia, assesses five things: breathing, negative vocalization, facial expression, body language, and consolability. Under body language, the highest pain score is assigned to a rigid body, clenched fists, knees pulled up, pulling or pushing away, and striking out. Those are the same behaviors that get documented as combativeness. In an instrument designed to detect pain in people who can no longer report it, they are the top of the scale.
This isn't a reframe for its own sake. Pain in this population is common and badly missed. Up to 80% of nursing home residents with dementia are reported to experience pain, and it is consistently described as under-recognized and under-treated. As dementia advances, people lose the ability to say where it hurts or to connect the pain to a cause, so it surfaces as aggression, agitation, resistance to care, or wandering instead of as a complaint.
The causal evidence is unusually good for a behavioral claim. A cluster randomized trial across 60 nursing home units in Norway gave 352 patients with moderate to severe dementia and significant behavioral disturbance a structured eight-week pain treatment protocol. Agitated behaviors decreased significantly compared with usual care, across verbal agitation, physically non-aggressive behavior, and aggression. Treating the pain reduced the behavior.
And the setting closest to yours has been studied directly. In a longitudinal cohort of 230 people with dementia admitted to UK general hospitals, pain was strongly associated with behavioral and psychiatric symptoms, with the strongest associations being aggression and anxiety, and the association held for pain assessed on movement. Movement is what you are about to do to them.
Restraint and chemical sedation decisions belong to your protocol and medical control, and this page doesn't touch them. What's worth carrying into that decision is the order of operations. Undiscovered pain does not stop being pain once someone is held down, and an agitation that would have responded to analgesia is a different clinical problem from one that requires sedation. Where your protocol allows the pain question to be asked first, ask it first.
The fight is usually a sentence. Most often it's "that hurts," said by someone who no longer has the word for it.