CNA Shield

The shift this is built for

The same shift, with the tool running: one tap at the start logs 14 residents assigned, timestamped. Through the night he talks while he works — “Room 306, repositioned to right side, incontinence care complete, resident calm” — thirty seconds, hands never stopped. At the end there's a full draft care log waiting, every room and event timestamped; he corrects two entries and approves it in four minutes.

The handoff is already written: Mr. Davis's BP flagged, the 312 resident's baseline change noted, two items marked for immediate follow-up. The incoming CNA reads it before walking onto the floor. And at month end, 23 documented shifts show a 1:14 ratio — aggregated, anonymized, with no worker's name attached to it.

The workers with the least power carry the most liability

That's the structural gap, and it shows up in three specific ways.

1 · You cannot chart with both hands full

This isn't a discipline problem, it's a physical impossibility — care takes two hands, full attention and constant movement. So documentation happens afterward, from memory, and comes out late and incomplete. When care can't be proven, it becomes the worker's problem. Of 1,324 posts from working CNAs, 92 described charting as a distinct burden stacked on top of an already unsustainable workload:

“In orientation they said 'we have computers in the nursing station for charting but we encourage you NOT to use them. There are kiosks mounted on the wall specifically for the CNAs.' The kiosks make your arm hurt and your feet are already killing you. I'm fed up with the way we're treated.”

Working CNA

The documentation system was designed for nurses. CNAs got wall kiosks.

2 · Nurses have a handoff protocol. CNAs have nothing.

Nurses are taught SBAR in every nursing program. There is no equivalent for the aides, so baselines, behavioral changes and care exceptions simply don't get transmitted between shifts. The consequence, in one CNA's words:

“The charge nurse had to ask me if the patient was forming normal sentences when I last saw them. The nurse prior didn't give a baseline. There was no way of knowing how long the patient had been in that state.”

Working CNA

She had possibly missed a stroke, because the previous shift left no neurological baseline. The outcome is unknown. She carried the guilt for a system-level failure.

3 · Reporting an illegal ratio can cost you the job

Legal ratios — 1:8 in many states — are routinely broken. The analysis found 338 posts describing ratio conditions, in the plainest possible terms:

“Legislation says the ratio is supposed to be 1:8 during the day but most of the time this month it's been 1:21 per hall with a float.”

CNA who tried to report their facility to state regulators

“Ratio went up to 10:1. Third night, 14:1. I was running around non-stop. She's written me up for neglecting and ignoring a patient's needs.”

CNA disciplined for care lapses caused by facility-imposed ratios

The one who filed a state report had to put her name in the submission email — which makes an “anonymous” report traceable. Without objective, timestamped, aggregated evidence, these get filed away as disgruntled-worker complaints.

What 6,333 posts were read for

The evidence base is a computational read of 6,333 organic posts across six caregiver communities — CNAs, eldercare, caregivers, caregiving, aging parents, dementia. Nobody was prompted and no researcher was present; these were written to peers. Pain point extraction over 30 high-signal CNA posts produced 21 structured pain points at an average severity of 4.2 out of 5. Keyword analysis put numbers on the shape of it: 338 ratio posts, 319 on people quitting, 92 on charting, 42 on short staffing. Semantic similarity across 6,236 text chunks ranked low pay, retention and short staffing as the highest-signal themes.

The failure mode isn't unique to CNAs, either. Direct Support Professionals, home health aides and community support professionals do structurally identical hands-on, documentation-heavy shift work, often with less formal training. The voice-first mechanic — talk while working, structured entry, generated handoff — is role-agnostic by design. CNAs are the pilot because that's where the deepest evidence and the readiest technology already are.

How it works

1 · Voice care logging

Activate at shift start, then speak brief observations hands-free as you go: “Just completed incontinence care for Room 12, repositioned to left side, vitals BP 128/82, resident alert and oriented to person and place.” Speech recognition runs on the device; a model structures it into a timestamped entry mapped to the facility's documentation format. At shift end you get an editable log to review.

Nothing is auto-submitted. The CNA reviews and approves every entry. Human judgment governs the record.

2 · Shift handoff assistant

The day's logs get synthesized into a structured handoff: falls and near-falls, behavioral changes from the documented baseline, medication exceptions, and what needs attention immediately. Outgoing CNA reviews and approves, incoming CNA reads it at shift start. That's the continuous baseline record the stroke case above didn't have.

3 · Safe ratio reporting

One tap logs the real assignment at shift start. Ratio data accumulates across shifts and workers into aggregate, anonymized reports fit for state submission, with timestamped documentation behind them. No individual is identified; differential privacy keeps the aggregate from being reverse-engineered back to a reporter. It is worker-owned and worker-initiated — generated only when a CNA logs an assignment, never visible to or controlled by the facility, and it works whether or not the facility adopts anything. The model is the Aviation Safety Reporting System, which transformed airline safety by decriminalizing the act of reporting.

Where it actually stands

The prototype turns raw spoken shift notes into a structured, triaged handoff today, and that's logged rather than merely claimed. Three full shifts are captured. What matters in them is the triage: in the first, one elevated blood-pressure reading goes on watch, while the same resident's second elevated reading is escalated to urgent with the reason stated — exactly the distinction a worker at 1:14 can miss. In the second, a fall is pushed to the top. In the third, a routine shift correctly produces no urgent items at all. Every line traces back to something spoken; no vital sign or event is invented. Alongside that: a 50-case labeled test set scored for precision, recall and F1 per triage class, a 40-case adversarial stress log, and a fake-protocol safety test.

Overall: TRL 3–4. Ambient documentation for doctors is a solved product category sitting at 7–8. The two pieces rated 3–4 — a speech model tuned to care-setting language, and the anonymous reporting pipeline — are honestly at design and early-prototype stage. Those are the build targets, not current capabilities.

The population technology skipped

The architecture here isn't novel; ambient clinical documentation already works, for physicians. The contribution is pointing it at the direct care worker — a workforce systematically left out of healthcare technology investment — with an interaction design for someone whose hands are full, and a data model that protects the worker rather than the facility.