commvita™
Connected care platform
GOVERNANCE & ASSURANCE EDITION

Quality Improvement & Evidence Programme

Problem, baseline, method, tests, evidence, benefits and learning as one governed record — built around a refusal: commvita reports that a measure moved and never claims it improved, because separating signal from variation needs SPC and that isn’t built.

Ahead of the field
A corrective action raised by the incident that caused it. Above a severity threshold the linked record is created in the same step as the report, carrying the source reference, so the action register and the incident register can’t drift apart.This surface is demonstrated on seeded data today; the claim is about how the platform is put together, not about a deployment. From commvita’s own assessment of this module against the systems it competes with. Our assessment, not an independent one.
Live vs demonstrated: Live — real, API-backed platform logic (wired end-to-end today) Demonstrated — representative control surface with seeded data / illustrative UI mock-up

1A control system, not another project tracker

Most improvement programmes are a list of projects and a RAG column. They can tell a board how many projects are running and how many are green. They can’t tell it whether anything got better, because a measure moving and an improvement aren’t the same claim — and almost every tool in this space quietly treats them as one.

commvita holds the whole thing as one governed record: a problem, a baseline with its data source, an improvement method, the measures, the tests (PDSA cycles, Lean artefacts, an agile backlog), the evidence, the benefits and the learning — with the refusal below sitting at the centre of it.

The control the whole module is built around: a measure moving isn’t an improvementbaseline (dashed) and ten recorded pointscommvita reports the direction and refuses the claim`improvement_claimed` is ALWAYS false, and the reason travelswith it.Distinguishing a real signal from ordinary variation needsstatistical process control. It isn’t built.Two points and a green arrow could just as easily be noise,and a board can’t tell the difference from a dashboard.So the chart above is reported as a downward direction over ten points. It’s NOT reported as an improvement, and nothing in the product says the change caused it.
This is the module’s central honesty control. Ten points trending down is a direction. Calling it an improvement would require separating a real signal from ordinary variation, which needs statistical process control — and that isn’t built.

The three-point minimum is commvita’s own, and is attributed to nobody. Below three recorded points no direction is described at all. Published run-chart rules vary by source, none has been read here, so the threshold is stated as a product decision rather than dressed up as a standard. That’s a smaller claim than most tools make and it’s the one that can be defended.

2The lifecycle, and where it refuses

A project moves through eleven phases, and can be run under whichever method the team actually uses — Lean, agile, PDSA, the Model for Improvement, a hybrid, or a local method named as local, not forced into somebody else’s vocabulary.

Eleven phases — and the two points that refuseideaproblemdefinitionaim andmethoddiagnosticinterventiondesigntest anditerateimplementvalidatebenefitsustain andspreadclosedabandonedCannot ENTER DELIVERY withoutA problem statement.A baseline carrying its data source.An outcome measure.And at medium or high risk, a balancing measure — the thing that catches animprovement paid for somewhere else.Cannot CLOSE withoutFinal values against the measures.Every benefit dispositioned.Learning captured — including from an abandoned test, which is retainednot deleted.The refusal names what is missing, instead of warning.
Two gates refuse and not warn, and each names what is missing. The balancing measure at medium and high risk is the one most often left out — it’s what catches an improvement that was paid for somewhere else in the system.

Learning from an abandoned test is kept. A PDSA cycle and its learning survive the decision to abandon. A programme that deletes its failed tests is a programme that will run them again, and the record of what did not work is frequently worth more than the record of what did.

3How credible is this project?

Evidence maturity 0–5 — DERIVED from what has been recorded, never typedSo two projects on one board paper aren’t equally credible just because both are green. A caller-supplied level is ignored.0no evidence1problem evidenced2interventionevidenced3improvement emerging4benefit validated5improvementsustainedLevel 5 means a monitoring period actually elapsedSustained is derived from the dates, not from somebody ticking a box that says sustained. An improvement that held for three weeks and an improvement that held fora year are different claims, and only one of them is level 5.
Maturity is computed from what has actually been recorded, so a project that has evidenced a problem and a project that has sustained a validated benefit don’t read the same on a board paper — which they would if both were simply green.

Alongside it, activity and improvement are reported separately — five counts that are never summed into one. Number of projects, number of tests run, number of benefits validated: each is a real fact, and adding them produces a figure that means nothing while looking like progress.

4Nothing is forked

What it reuses instead of forks — and one claim that was wrong and was correctedRisks and issues/risk-registerOne risk source of truth. A QI project links to a risk; it doesn’t keep itsown register.Financial benefit/frpA financial benefit holds NO amount of its own. It names a recovery scheme andthe validated figure is READ from that scheme.Quality impact of a change/quality-impact-assessmentA pathway change that affects care goes through the QIA, not through a secondassessment here.Process metrics over time/process-qualityCycle time, rework and handoffs are measured there and not re-derived.Approvals are NOT routed through the policy engine — and the product says soAn earlier version of this list claimed QI approvals ran through the governed approval engine. No QI code touches it, so the claim was wrong and was removed.What exists instead is stated plainly: benefit validation is gated at executive, the benefit owner can’t validate their own benefit, and the evidence submittercan’t accept their own evidence.What does NOT exist is listed too — no SLA-timed escalation, no N-of-M quorum, no team authorisers, no hash-chained approval audit. The claim is checked against the code so itcan’t drift from the code again.
A financial benefit carries no amount of its own. It names a recovery scheme, and the validated figure is read from that scheme’s own finance validation — one saving, one number, however many places it’s reported.

5Where it lives, and its honest edges

PartWhereStatus
Programmes, projects, measures, PDSA cycles, Lean artefacts, backlog /quality-improvementLive
API/qi/ — 26 endpoints Live
What is enforced, what is reused, and what can’t be done GET /qi/methodologyLive
Evidence, review and derived maturity POST /qi/projects/{id}/evidence, /evidence/{id}/review Live
Benefits, validated by somebody other than their owner POST /qi/benefits/{id}/validate Live
Organisational learning, retained from abandoned tests too GET /qi/learningLive
Risks — linked, never re-keyedGET /qi/projects/{id}/risks Live
Process metrics over time/process-quality Live
Clinical audit and CAPA/clinical-audit, /capa Live
Statistical process control— Not built

Statistical process control is recorded as an absence, not omitted. Without it, commvita can’t separate special-cause from common-cause variation and therefore can’t say a change worked. It’s in the standards register as a gap not left for a buyer to discover, because the honest version of this product is one that reports direction and stops.

It can’t judge whether a method was applied competently. It records that the artefacts exist — a problem statement, a driver diagram, a PDSA cycle — not that any of them is any good. A complete, well-formed project can still be a bad project.

Improvement work done outside commvita is invisible here, and must not be read as an absence of improvement. And NHS IMPACT is NHS England’s framing: commvita doesn’t assert it applies in any other jurisdiction, which is the same fail-safe direction the platform applies to every other national framework.

Models: QIProgramme · QIProject · QIMeasure · QIMeasurePoint · QIPDSACycle · QILeanArtefact · QIBacklogItem · QIEvidence · QIBenefit · QILearningAPI: /qi/ (26 endpoints) · Route: /quality-improvement · Non-SaMD
© 2026 Commvita Digital Health Solutions Ltd. All rights reserved.