LG · Feature design
Telehealth Smart Rails
- Shipped
- 1-month research, 3-month design + eng (12.2023 – 03.2024)
- Product designer
- 1 PM, 3 engineers
- Web
- Background
- LG's patient observation app lets one remote technician watch up to twelve hospital patients across ICU, step-down and medical–surgical units.
- Problem
- Visual monitoring delays or misses fall risk — through observer fatigue and divided attention, not carelessness. Nearly a million inpatient falls happen in the US each year.
- Approach
- Interviewed seven telesitters and nurses, weighed two detection models with engineering, then designed an AI boundary alert together with the controls that stop it becoming noise.
- Outcome
- Phased launch 04.2024, GA 08.2024. SUS 87, and a measurable drop in falls on top of what human monitoring already caught.
Problem
Limitations of human monitoring
in fall risk detection
Inpatient falls are among the most common and most expensive harms in US hospitals, which is why continuous observation is worth a dedicated pair of eyes. However, one remote technician watches up to twelve patients at once. Attention has hard limits, and falls happen in the gap — not through carelessness, but through fatigue and divided attention.
User research
Seven interviews, three consistent needs
I interviewed seven users: one operations manager, three telesitters, and three nurses with telesitter experience. Three needs came through consistently.
Early detection
Alerts for patient movements that may lead to a fall, early enough to intervene in time.
Tailored alerts
Every patient has a different medical condition, mobility level, and fall risk profile — so detection has to be tuned per person.
Concurrent movement
Catching simultaneous movements across feeds so that none of them is missed.
Design exploration
Two ways to detect a fall risk
Early detection and technical feasibility pulled in opposite directions. I worked through both options with engineering to see what each would cost — in development time, and in false alerts.
Option A — Posture analysis
Reads the patient's posture with a custom-trained dataset to spot movements that lead to a bed exit.
- Pros
- Detects intent early, before the patient reaches the edge
- Cons
- Long development time — high computational complexity, custom dataset, heavy fine-tuning
- More false alerts in cluttered rooms, affected by blankets and poor light
Option B — Boundary crossing
Watches a boundary drawn around the bed and alerts the moment the patient crosses or touches it.
- Pros
- Easier to implement with rule-based alerts
- High accuracy for in-bed and out-of-bed behaviours
- Cons
- Detects later than Option A
- Cannot cover key in-bed movements — raising up, shifting, rolling
MVP design
Alerting when a patient crosses the virtual boundary
Option B shipped first because it detects earlier and more consistently than human monitoring at a fraction of the build cost — lower risk, fast to iterate with existing partners, with the hybrid approach still the long-term vision.
Step 1: Initial Smart Rails setup
Users access Smart Rails setup from the side panel, the central place for configuring monitoring features such as Smart Rails, Canned Messages, and Blur Mode.
The boundary is drawn straight onto the live feed, so the AI knows exactly where the bed edges are and can tell when a patient is trying to get out. Sensitivity and a trigger delay are set per patient, because mobility and fall risk differ.
Step 2: Get alerts and take actions
When a patient crosses the Smart Rails boundary, a colored alert appears directly on their video. Sitters can scan multiple rooms at once and instantly see who is at risk of falling.
Step 3: Documenting what happened
Every event is timestamped into the patient's log — alerts, interventions, messages spoken to the patient, privacy mode, virtual rounding. An incident can be reconstructed afterwards and handed to the care team.
Usability testing
Remote usability testing with 5 targeted users
Remote testing with five users: two Intermountain patient safety monitors, two nurses, and one remote safety tech. The prototype scored an average SUS of 87. The consistent criticism was not about finding things. It was alerts firing when nothing was wrong.
Design update 1: Adding Mute mode & alert delay
A customisable delay filters out momentary movements. Mute mode silences alerts when staff are present or in privacy mode.
Design update 2: Adding Snooze mode
The inline banner lets you snooze false or nuisance alerts; Smart Rails turns back on automatically.
Impact
What changed after launch
Phased launch 04.2024 – 07.2024, GA launch 08.2024. Averages over the last 90 days compared with the previous period.
“The real-time AI alerts help me respond faster and prevent falls more effectively. I'm way less tired by the end of my shift, and it's made a real difference for both patient safety and my job satisfaction.”
Elizabeth N, Patient Safety Monitoring Technician
More projects
