Memory care technology has spent years getting good at one question: where is the resident. Door sensors, wander bands, GPS, RFID checkpoints. If you always know where someone is, you can go and get them when something goes wrong, and that logic has held up well enough that most facilities still run on it.
Lately the field has started asking a harder question — not where the resident is, but what is changing about them. That shift toward behavioral monitoring is real progress. It also inherits an assumption that does not survive contact with dementia: that somewhere under the noise there is a stable baseline, and that the events worth flagging are the ones that depart from it.
A Baseline That Will Not Hold Still
For a lot of conditions that assumption is fine. You establish what normal looks like, you watch for deviations, you surface the ones that matter. Dementia does not hold still long enough for that to work. What counts as normal for a resident in the spring may be gone by winter — the routine narrower, the sleep broken, a walking pattern that used to read as agitation now just part of how they move through the afternoon. The baseline itself is drifting, because the disease keeps moving it.
A resident might have walked the same loop after lunch for years — down to the lobby, a pause at the front doors, back past the nurses’ station a little before two. A behavioral system learns that as hers and does well with it. Then the loop shortens; she starts turning back at the dayroom. Months later she is making smaller circuits closer to her own room, more of them, at hours that don’t line up with anything. None of it shows up as an event, and each week looks about like the one before. The resident the model knows is steadily falling behind the one in the building.
When the reference point moves and the system’s picture of normal does not, two things go wrong. It raises alarms against expectations that expired months ago. Or it does the quieter and worse thing, adjusting so gradually that real decline settles in as the new ordinary and nobody is told. Neither is a sensor failure. The person has changed and the model has not kept up. Families already know this part — someone with dementia stays unmistakably themselves while almost everything about how that self comes through the day keeps shifting. The continuity is real. What changes is its expression.
The Shape of a Day
A day with its structure intact has a recognizable shape. People wake around the same time, move along familiar routes, string tasks together in an order that mostly holds. The shape is never rigid — a bad night, a visitor, a change in the weather all bend it — but it stays coherent enough that you could describe someone’s ordinary Tuesday and be mostly right.
Dementia works on that coherence slowly. Routes that used to run start to stop and restart, transitions between rooms stretch out or go missing, and the timing loosens until things that once happened before lunch now happen whenever. Researchers who study movement and daily activity reach for a rough measure of this and call it behavioral entropy — not a score to chase, just a way of asking how much predictable structure is left in a day. A low reading looks like steady routes and stable timing; a rising one looks like repeated loops, journeys that don’t complete, an afternoon that no longer resembles the one before it. What makes the measure worth having is that the loss of coherence tends to show up before anything an incident report would recognize.
Which Way, and How Fast
The more useful question is not how far today sits from baseline. It is which direction the resident is heading, and how quickly. Picture two residents with the same level of disorganized behavior right now. One has looked like that for six months. The other looked settled three weeks ago. Same snapshot, very different situations, and they do not call for the same response. What matters is usually not the reading on any given day but the slope: whether the structure is eroding, holding, or slowly coming back.
Why This Is Harder Than an Alert
An alert has a simple job. Something crosses a line, and it fires. Following a trajectory is different work: you carry a model of one particular person, keep updating it as they change, and stay useful the whole time the model is a little out of date — which is most of the time, because the thing being modeled never settles.
Several other difficulties come with the territory, and none of them really resolve. A new resident arrives with no history, so the system has to earn its keep while it is still working out what their normal was, and that normal keeps moving while it learns. It has to hold everyday variation lightly — people have bad weeks, routines loosen over a holiday and tighten again — while still catching the occasional week that turns out to be the first of a longer decline. No two people go the same way, so population patterns only take you so far; what matters for a given resident is how far their own trajectory has bent from their own earlier one. And the record that would teach a system how dementia behavior unfolds over years is thin, not the sort of thing you can buy or simulate convincingly. It accumulates one deployment at a time, through cases and outcomes and the slow work of watching people change. Given enough of it, that accumulated understanding may end up mattering more than any single model, because it is the part that moves at the same speed as the disease.
Before the Incident
The industry counts incidents because incidents are visible. You can document them, report them, put a number on them. But an incident is usually the far end of a drift that began much earlier and much more quietly. Seeing that drift sooner does not promise prediction, and it does not remove the uncertainty that comes with this disease. What it can do is return a little time while the situation is still recoverable, which is most of what protecting an ordinary day comes down to. A baseline was never really a definition of normal — more the beginning of understanding what one particular life looks like when it is steady, and what is worth noticing when that starts to slip.
A Note on Uncertainty
None of this amounts to certainty. No model represents a person’s behavior faithfully, and dementia care will always carry more unknowns than a system can close. The aim is narrower and more honest than prediction: a better chance of noticing, while there is still room to act, that an ordinary life has started to drift.
Sources
¹ Enshaeifar, S., et al. "Health management and pattern analysis of daily living activities of people with dementia." npj Digital Medicine, 2018.
² Martino-Saltzman, D., et al. "Travel behavior of nursing home residents perceived as wanderers and nonwanderers." The Gerontologist, 1992.
³ Alzheimer's Association. Alzheimer's Disease Facts and Figures.
⁴ Koester, R.J. Lost Person Behavior.
Originally Appeared
This article originally published by McKnight's Long-Term Care News as part of its Marketplace Experts series.
We are grateful for the opportunity to contribute to the industry's ongoing conversation about dementia care, wandering prevention, caregiver burden, and resident safety.
The article is archived here as part of Rientro OS and may be updated periodically as our understanding of these challenges continues to develop.
Original publication: The baseline moves: Why static intelligence struggles with a progressive disease — McKnight's Long-Term Care News