Anatomy of 2,000 Falls

How the Teton system sees a room

The Teton sensor detects every person in the room, including residents, visitors, and staff, and classifies what they are doing 10 times per second. The system recognises 14 distinct states a person can be in. A transition is any movement from one state to another, such as getting out of bed, sitting down in a chair, leaving the room.

Every time a person moves from one state to another, the system logs a transition. Across 2,000 falls, we analysed the 3 million transitions surrounding them to understand which ones are dangerous and which ones are not.

See more of Teton's system →

Not all transitions are equal

A care home monitoring system tracks dozens of state changes per resident per day. A resident goes from lying in bed to sitting up. From sitting to standing. From standing to walking. From in the room to out of view. Most of these transitions are completely routine. A few of them are dangerous.

We measured the relative fall risk of every transition type across 3 million events:

The first two make intuitive sense. The transfer from a supported position (lying, sitting) to an unsupported one (standing) is the moment of maximum instability. But the out-of-view finding is just as important: transitions to and from the bathroom, the corridor, or out of the room are significantly less likely to lead to falls. This means many monitoring systems are alerting on exactly the wrong transitions, generating noise from out-of-view events while under-alerting on bed and chair exits.

When falls happen

Time of day has a dramatic effect on fall risk:

A transition at 3 AM is 63% more likely to result in a fall than the same transition at 3 PM. This isn't surprising: reduced staffing, darkness, disorientation from waking, and medication effects all play a role. But the magnitude is what stands out. Night-time monitoring isn't just "nice to have." It's where the majority of the preventable risk sits.

Inactivity is a risk multiplier

The longer a resident has been inactive before a transition, the more dangerous that transition becomes:

A resident who has been lying in bed for over an hour and then transitions to standing is 2.4 times more likely to fall than one who has been resting for less than 15 minutes. The clinical explanation is straightforward (orthostatic hypotension, stiffness, disorientation) but the monitoring implication is significant. An alert for "getting out of bed" at night after an hour of sleep is a fundamentally different event from the same transition after 5 minutes, and should be treated accordingly.

The best predictor of falls is falls

Residents who had a fall in the last 3 months are 5 times more likely to have another fall in the next 30 days. This is the single strongest predictor in our model, stronger than any health metric, mobility aid usage, or sleep pattern.

This has a direct operational implication: after a fall, a resident's monitoring sensitivity should increase immediately and stay elevated for at least 90 days. Most facilities don't do this systematically.

Walking aid users: 32% of residents, 52% of falls

This was one of the most striking findings. Walking aid users are 3.2 times more likely to fall than residents who don't use mobility aids. Despite being only about a third of the population, they account for more than half of all falls.

The data gets more specific:

Two-thirds of walking aid user falls happen when the aid isn't being used. And in 43% of cases, the walking aid was physically in the room, just not within reach. This is a significant finding because it points to a category of falls that is preventable through environmental adjustment rather than clinical intervention. Repositioning a walking aid closer to the bed is a simple change that care teams can act on immediately.

Wheelchair users and bed transfers

Wheelchair users face a compounded risk. Getting out of bed is already 2.5x more dangerous than baseline. For wheelchair users, it's 5.3x: the 2.5x base risk multiplied by an additional 2.1x from the wheelchair transfer.

In care settings where bed-exit monitoring was active for wheelchair users, fall rates in this group dropped by 66% (from 16.8 to 5.7 per 1,000 patient-days), the largest reduction of any population segment in the dataset.

Sleep and respiration: the quiet signals

Two health metrics showed meaningful correlation with fall risk:

Sleep regularity: We measure how consistently a resident falls asleep and wakes up at the same times each day. This is scored on a 0 to 100 scale, where 100 means perfectly consistent sleep patterns and lower scores mean increasingly irregular timing. Residents scoring below 70 fall 1.5x more often. Below 10, they fall 2xmore often. This relationship holds for both short-term (3-day) and long-term (30-day) predictions. When a resident's sleep pattern starts shifting, it's often the first visible sign that something has changed: new medication, increasing pain, an emerging infection, or cognitive decline. It shows up in fall risk before it shows up anywhere else.

Respiration rate: The sensor measures breathing rate overnight. When a resident's respiration rate starts increasing from their normal baseline, it often means they are getting ill. And when someone is getting ill, they are also more likely to fall. Any increase from their baseline raises fall risk by 23% within the next 3 days. A significant shift increases it by 36%. The connection is straightforward: infection, fluid retention, and medication changes all affect balance and cognition, and they show up in breathing patterns before a fall happens.

Neither of these are traditional fall risk factors. They're not on any standard assessment checklist. But they're measurable, they're predictive, and the Teton sensor captures both of them passively, every night.

From data to prevention: how Teton uses these findings

Understanding when and why falls happen is only useful if it changes what care teams do. Teton's system translates these findings into three prevention mechanisms, each addressing a different part of the problem:

The results: 42% reduction, and the path to 70%

Together, these three mechanisms have produced measurable results. Communities using Teton's monitoring system saw an overall fall rate reduction of 42%, from 11.3 to 6.6 falls per 1,000 patient-days. The reduction was most dramatic for the transition types that the alarm system is designed to catch:

These are the transitions where the Teton system helps the most: moments where a resident moves from a supported position to an unsupported one. The system detects the transition, alerts staff, and gives them time to intervene.

But the analysis shows that of the 60% of falls that still occurred, 37% could have been prevented. The reasons they weren't:

The first two categories, representing over 37% of remaining falls, are addressable with better alarm configuration and fewer, more targeted notifications.

The path to 70%+ fall reduction isn't theoretical. It's the sum of closing these specific, measured gaps.

All statistics in this analysis are based on production data from Teton-monitored care settings. No estimates or projections.