The gait change before the fall.

A fall is the moment everyone sees. The chart entry, the incident report, the call to the family. By then the question is recovery, not prevention. The interesting signal is upstream — in the days and sometimes weeks before, when the way a resident moves starts to drift from their own normal.

Companion watches that drift. A transfer that used to take four seconds now takes seven. A new pause at the edge of the bed before standing. A subtle asymmetry — the right side hesitating where the left does not. None of these are alarms on their own. Together, over a week, they are a pattern.

Why the baseline is the resident, not the population

A 92-year-old with Parkinson's and an 78-year-old recovering from a hip replacement do not share a normal. Comparing either of them to a population average produces noise — alarms on people who are simply themselves, and silence on the residents whose decline is real but starts from a faster baseline. The right reference is the resident two weeks ago.

So that is what Companion tracks. Transfer time in and out of bed. Hesitation events at thresholds — the bedside, the bathroom door. Left-right asymmetry in how weight shifts during a stand. Each metric is anchored to the resident's own rolling baseline, and the system flags sustained drift, not a single slow morning.

What PT actually gets

The product impact is a heads-up, not a report. When the pattern crosses a threshold, the resident shows up on the PT's Monday list with the specific change attached — transfer time up 40% over ten days, new right-side hesitation on stand — so the assessment is targeted, not exploratory. We are not predicting falls. We are flagging the residents whose own movement is asking for attention. The fall that does not happen is the one the metric is for. Next up: surfacing the same baseline view to families in the resident's weekly summary.