The forecast layer for home care.
Pulse watches the operational and clinical signal flowing through Carefold and tells you what's about to need attention. Not at the end of the month. Tonight.
Four predictions. Live, all the time.
Pulse runs continuously in the background of every Carefold tenant. There's no model to configure, no separate dashboard to learn — its outputs surface inside Helm, Move, and Home as you'd expect.
Missed-visit risk
Wellbeing decline
Continuity matching
Capacity strain
No new data needed. The signal is already there.
Every input Pulse uses is data your team already records during routine care delivery. There is no new chart for a carer to fill in, no new device to install. Pulse is the difference between collecting data and using it.
Rota & ECM
Visit start/end timestamps, geofence breaches, mileage between visits, carer availability, leave patterns.
- Tamper-proof timestamps
- Carer absence runs
- Cluster density
Daily charts
Body map deltas, NEWS2 vitals, fluid & food against target, repositioning compliance, bowel/bladder, sleep.
- EPUAP staging
- Hydration trend
- Sleep variance
Notes & mood
Sentiment in care notes, observed-mood scores, activity engagement, social interaction frequency.
- Note tone delta
- Mood 7-day window
- Withdrawal signals
External
Met Office weather feeds, traffic, bank holidays, postcode density — used as covariates, never as the only signal.
- Met Office DataPoint
- Bank holiday calendar
- Postcode covariates
Three forecasts. All with reasoning.
Every Pulse forecast cites the signals it watched and the weights it gave them. No black-box outputs. A coordinator can interrogate any prediction in two clicks.
Wellbeing decline · Albert Reeves
Mood ↓ in 4 of last 5 visits. Fluid intake 41% of daily target (1.2L vs 2.0L) for three days running. Sleep quality logged "poor" by carers on Mon/Tue. Note sentiment dropped sharply on 19 Mar.
Camden evening run
Six visits clustered in NW1/NW3, two carers off (sickness + leave), Friday pattern shows historical 31% miss in this cohort, Met Office issuing yellow rain warning from 18:00.
Continuity match · Joan Sykes
Preferred carer F. Patel is the closest geographic match available Tue/Thu evenings for the next 14 days. Auto-pre-assigning would lift continuity score from 78% → 92% and remove the 19 Mar gap.
Predictive, but never opaque.
Healthcare-grade prediction lives or dies on whether the operator trusts the output. Pulse is engineered around transparency: every forecast carries its reasoning, and every model is auditable.
Reasoning, every time
Every forecast lists the signals it watched and the weights it gave them. No "the model says". A coordinator can interrogate any prediction in two clicks and the reasoning is logged.
Decision, not prescription
Pulse never auto-acts on clinical decisions. It surfaces, recommends, and asks. The Registered Manager remains the decision-maker — and the audit trail captures both Pulse's input and their judgement.
Tenant-isolated models
Pulse models are scoped per tenant. Your data trains your forecasts. No cross-tenant training. No third-party LLM gets your client records, ever.
DCB0129 / 0160 documented
Pulse is treated as a clinical safety case. Hazards, severity, mitigations and post-deployment monitoring are written up against DCB0129 (manufacturer) and DCB0160 (deployer).
Bias monitoring built in
Forecast accuracy is monitored across demographic and geographic cohorts. Drift triggers a hold on dependent automations until the model is reviewed.
Opt-out per tenant
Pulse is included in every plan. Any agency can disable any of the four forecast types per branch, per role, or wholesale — and turn them back on whenever they want to.
Bring two months of your CSV. We'll show you the forecasts you would have had.
In a 30-minute demo we can load a redacted CSV from your current system and run Pulse's missed-visit and wellbeing models against it. You see what you would have caught, and when.