Intended use & regulatory status

Intended purpose

Echo Pathways is a result-surveillance platform. It is intended to be used by qualified health professionals, within a healthcare service, to reduce the delay between a clinically significant pathology result and the clinical action that result should trigger.

In use, the system:

1. retrieves pathology results and patient context from the record system that already holds them;
2. structures those results — reading FHIR, HL7 v2, PDF, image and free-text reports, and normalising units, scales and censoring conventions;
3. matches a patient to an active, clinically signed-off protocol;
4. evaluates that protocol with a deterministic rule engine to detect trends, threshold crossings, and results that are expected but have not arrived;
5. surfaces a prioritised recommendation, from a bounded vocabulary of actions, to the clinician responsible for that patient — with the source, the rule and the reason attached; and
6. keeps that recommendation visible until a human reviews it.

Intended users

Qualified health professionals and the clinical governance teams that author and approve protocols. Echo Pathways is not intended for use by patients, and has no patient-facing interface.

Intended clinical setting

Hospital and specialist services with an existing electronic record system and an existing clinical governance process capable of authoring, reviewing and signing off protocols. The initial validated pathway is serial molecular monitoring in acute myeloid leukaemia (AML).

What Echo Pathways does not do

These are design constraints enforced in the product, not aspirations.

• It does not diagnose. It detects that a rule has been met and says which rule and why.
• It does not prescribe or recommend a treatment. The action vocabulary is bounded and operational — review, repeat, escalate, contact — not therapeutic.
• It does not act autonomously. A high-severity action always requires a human to authorise it. The system never auto-executes a consequential step.
• It does not let a generative model make the decision. Harmonisation, event detection, schedule logic and action selection are deterministic rule execution. A model is used only for constrained perception — reading a value off a document, schema-constrained and source-linked, with no inference and no arithmetic. Anything it cannot read confidently is routed to a human rather than guessed.
• It does not write back to the record system. FHIR context is read-only. Write-back is a separate, later, explicitly-scoped decision.
• It does not replace the clinician's inbox, the laboratory, or clinical judgement. It is a safety net over them.

Residual risk the user must manage

No surveillance system is complete, and we will not imply otherwise.

• Echo Pathways can only evaluate results it receives. A result that never reaches the connected source system cannot be surveilled — though where a protocol expects a result, its absence is itself detected and raised as high-severity.
• A protocol only detects what it was authored to detect. It is not a general-purpose safety monitor, and a clinically significant finding outside the protocol's scope will not be surfaced.
• Document reading can fail. Where confidence is low the item is routed to a human — but a human must then act on it.
• The clinician remains responsible for the patient. Echo Pathways records who was alerted and when; it does not discharge the duty.

Why we are describing it this way

Our sister product, Echo-Health, is positioned as an information-retrieval tool and is engineered to stay within the clinical-decision-support exemption — it surfaces information and never grades, ranks or recommends.

Echo Pathways does not make that claim, because it would not be true. Applying an approved rule to a result, assigning it a severity, and escalating it to a named clinician is clinical decision support. We are therefore building Echo Pathways under a medical-device quality system from the outset, and we will say it is certified when, and only when, it is.

Reporting a concern

If you believe Echo Pathways has behaved unsafely — a missed event, a wrong recommendation, an alert that reached the wrong person — tell us. Every decision the engine makes is replayable from its recorded inputs, so a specific report can be investigated exactly.

Clinical safety: safety@echo-pathways.com
Security vulnerabilities: security@echo-pathways.com