top of page
IHP_Patient.png
IHP_Logo_Green.png

I.H.P.

Imminent Hospitalization Probability

Predict Hospitalization - Before it Happens

What is IHP?

A transparent algorithm that estimates a patient’s likelihood of hospitalization on a daily basis. IHP produces a 0–1 probability, classifies into Critical / Very High / High, and shows Top Contributors so clinicians understand why a patient surfaced.

​

How it works (3 steps)

  1. Aggregate daily signals: vitals, device adherence, activity, and condition-specific metrics.

  2. Score via a rule-driven composite that reflects current status and short-term trends.

  3. Calibrate to probability with a logistic transform (α/β externalized per customer), then band into Critical / Very High / High.

​

Bands & Time-to-Event

  • Critical: ~1–7 days (highest near-term risk; governed SMS option during pilot)

  • Very High: ~1–14 days (proactive outreach)

  • High: ~3–30 days (routine outreach & coaching)​

 

Explainability (Top Contributors)

Every IHP result includes a ranked list of Top Contributors—e.g., O₂ saturation events, HR trend ↑ (7d), recent weight change—so staff can act with confidence and context.

​​

Built-in Governance

  • Weekly PPV & volume monitoring versus targets

  • Monthly false-page audit (chart review sample)

  • Versioned parameters, daily prediction logs, and audit trails​

 

Alerts & Workflows

  • In-app alerts by band with clear reasons and daily email report

  • Critical-band SMS during pilot under governance criteria

  • Daily worklist sorted by Band → IHP → Days-to-event

IHP FAQ

GET A DEMO

​​Is IHP a black box?

IHP (Imminent Hospitalization Probability) is a transparent, SQL-based algorithm that estimates a patient’s likelihood of hospitalization each day (0–1) and places them into three action bands: Critical (~1–7 days), Very High (~1–14 days), High (~3–30 days). It also surfaces Top Contributors so clinicians see why a patient appeared.

​

Is IHP a black box?

No. It uses a rule-driven composite score 𝑆 mapped to probability via a logistic transform. The calibration parameters (α/β) and band cutoffs are externalized (stored in parameter tables) and governed, with a full, auditable prediction ledger.

​

Who is IHP for?

Remote Care/RPM programs at provider groups, ACOs, plans, and care-management teams that want earlier, explainable signals to focus outreach before an admission.

​​

Clinical & Operational Use

How should teams act on each band?

  • Critical (~1–7d): same-day review; governed SMS + in-app alerts during pilot; rapid outreach/escalation.

  • Very High (~1–14d): proactive outreach, clinical review of contributors, short follow-up loop.

  • High (~3–30d): routine outreach and adherence coaching; watch trend.

​

How are patients prioritized each day?

  1. The worklist sorts by Band (Critical → Very High → High), then IHP (desc), then estimated days-to-event (ascending). Each row shows Top Contributors and a data-sufficiency indicator.

​

What if alerts become too frequent (alert fatigue)?

  • Governance includes weekly PPV & volume monitoring against targets and a monthly false-page audit (chart-review sample). If PPV drops or volumes exceed capacity, cutoffs/gates are adjusted.​

 

Can we suppress re-alerts?

Yes. Cool-downs/hysteresis can require a minimum ΔIHP or a time window before the same patient re-alerts—configurable per-band.

​​

Does IHP replace clinical judgment?

No. It’s a decision support signal. Final decisions remain with licensed clinicians following local SOPs.

IHP_Bands.png

Implementation & Pilot

Typical onboarding steps?

  1. Data mapping & quality checks

  2. Parameter seeding (α/β, band cutoffs)

  3. UAT with back-tests and shadow mode

  4. Pilot launch with governance guardrails

  5. Review → calibrate → scale

​

How long to first results?

In data-ready orgs, a shadow run can begin soon after mappings; pilot timing depends on governance approvals and training.

​

What training do staff need?

A 60–90 minute session covers the worklist, Top Contributors, escalation hints, and documentation expectations. Playbooks and quick-reference guides are provided.

 

Getting Started

Book a short session. We’ll review data readiness, define targets (e.g., PPV by band and alert volumes), and propose a pilot plan with timelines and responsibilities.

bottom of page