Agitation in Dementia: When Non-Drug Measures Are Enough, and When They Aren't
Two patients present with agitated behavior in dementia. Guidelines agree non-pharmacologic measures should come first in both — the two cases diverge only once that step has actually been done, and the question becomes what to do when it succeeds versus when it doesn't.
Ruth A., 84, has moderate Alzheimer's disease and has lived in the same memory-care unit for two years without significant behavioral concerns. Over the past ten days, staff report she has become newly agitated each afternoon — pacing, calling out, twice pushing away a caregiver who approached to help her to the bathroom. Nothing about her medical status has obviously changed; there is no fever, no new medication, and her family reports nothing different at home visits.
The team applies the DICE framework — Describe the behavior specifically, Investigate contributing causes, Create a targeted plan, Evaluate the result — before reaching for any medication. Investigation finds two concrete, modifiable contributors: her afternoon agitation clusters tightly around a newly reassigned aide whose approach is brisker and less narrated than the aide Ruth is used to, and a room reassignment two weeks ago moved her further from the unit's sunniest common area, coinciding with less daytime light exposure and a flatter sleep-wake pattern per nursing logs.
She pushed a caregiver away yesterday. I don't want to wait a full week on an investigation that might not find anything actionable before we put some safety margin in place — even a low, well-tolerated dose of something now seems reasonable while the behavioral work continues in parallel.
Before we reach for that, let me walk through what investigation actually found, because I think it changes the calculus. Her agitation clusters tightly around one specific reassigned aide's brisker approach, and around a room move that measurably reduced her daytime light exposure. The Kales, Gitlin, and Lyketsos framework for behavioral symptoms in dementia, published in the BMJ in 2015, argues specifically against reflexive medication use exactly when a cause this concrete and fixable is already in view.
One pushed-away hand, without injury, in a patient with two years of no prior behavioral concerns, isn't yet the kind of ongoing safety risk that should override a plan this specific and this likely to work.
That's a fair read of the investigation, and I'll concede the triggers are concrete enough to be worth trying first. I'd still want the plan documented with a short, explicit re-evaluation window rather than an open-ended one, given what already happened.
Agreed: no medication started. The environmental plan is implemented today with a one-week re-evaluation scheduled and documented as a decision point, not an open-ended wait.
At re-evaluation one week later, the afternoon agitation had resolved with the aide reassignment and increased light exposure alone — no medication was ultimately needed for Ruth.
Frank D., 79, has moderate-severe Alzheimer's disease and, like Ruth, is evaluated with the DICE framework when new agitation appears — in his case, escalating physical aggression toward staff during morning care over three weeks, including a scratch that broke a caregiver's skin and required a tetanus check. Investigation finds no clear environmental trigger: no staffing changes, no schedule disruption, no identifiable pain source after a full exam and review of his medication list for anything newly constipating or otherwise uncomfortable.
A two-week trial of a consistent, familiar morning caregiver, slower unhurried pacing during care tasks, and a structured pre-care routine — the team's best environmental intervention, built the same way Ruth's was — produces no meaningful change. The aggression continues at the same frequency and severity, and the unit's most experienced aide, the one now specifically assigned to his mornings, reports the same outcome regardless of how the approach is modified. Two staff injuries and no identifiable modifiable cause after a genuine, structured attempt change the calculus in a way Ruth's case never reached.
Unlike Ruth, Frank has actually been through the full non-pharmacologic process and it didn't work. I'd start citalopram now. The relevant trial is CitAD — Porsteinsson and colleagues, JAMA, 2014, 186 patients with clinically significant agitation in Alzheimer's disease, randomized to citalopram at a target of 30 mg daily or placebo. Forty percent of the citalopram group showed moderate-or-marked improvement on caregiver-rated global impression, versus twenty-six percent on placebo.
I'd support starting it, but I want the QTc data named plainly before we do. In that same trial, citalopram produced an 18.1-millisecond greater increase in QTc interval than placebo at three weeks — a real, statistically significant difference, not incidental. The trial itself was amended mid-course to add mandatory ECG monitoring after an FDA safety communication about citalopram's dose-dependent QTc effect in older adults.
I'm agreeing this is the right call given where we are, but I don't want the CitAD benefit cited without also naming that its own investigators couldn't recommend the 30 mg dose without qualification, specifically because of this cardiac finding.
Fair, and I'll build that into the plan directly: baseline ECG before starting, 10 mg to begin rather than jumping to the trial's 30 mg target, and a repeat ECG at the first dose increase.
One thing I want to say out loud, though, because CitAD's 30 mg target invites exactly this mistake: we cannot titrate him to the trial dose. He's 79. The FDA capped citalopram at 20 mg a day for anyone over 60 — that warning came out while CitAD was still running, which is why the trial's own investigators ended up qualifying the dose they'd studied. So 20 mg is our ceiling, not a waypoint on the road to 30. Given the ongoing risk to staff and the exhausted non-pharmacologic options, treating within that ceiling with ECG monitoring is the right balance — better than withholding treatment over a QTc signal we can actually watch.
Agreed: citalopram started at 10 mg with baseline ECG, titrated no higher than the 20 mg/day labeled maximum for patients over 60, continued alongside the existing environmental measures, with a two-week re-evaluation of both aggression frequency and repeat ECG at the first dose increase.
Not agreed: whether 20 mg — the most he can be given at his age — meaningfully preserves the benefit CitAD demonstrated at 30 mg, or whether the age-based cap simply means treating him at a dose the trial never tested. The pharmacologist accepts the ceiling as non-negotiable and would judge response against it; the psychiatrist thinks that if 20 mg fails, the honest next conversation is about a different drug rather than a higher dose. Both positions were left on record rather than resolved.