A change to a service is a decision about care until somebody says otherwise. The QIA is the record that they did — nine domains, scored before and after mitigation, versioned rather than edited, and refused at four points rather than warned about.
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
1What a Quality Impact Assessment is for
A service redesign, a savings scheme or a change to who is eligible is a
decision about care whether or not anybody writes that down. A Quality Impact
Assessment is the record that somebody did — what was expected to get
worse, on what evidence, what was done about it, and who accepted the residual
risk. It is made before the change, and reassessed after it.
commvita implements the National Quality Board quality impact assessment
framework (NHS England, publication reference PRN01920_i, published 24 June
2025) together with its assessment tool and RASCI template. The framework
explicitly replaced 2012 guidance that framed the whole exercise around cost
improvement — and the list of triggers shows why that mattered.
Cost improvement is one trigger of twelve. Reading a QIA as a savings document is the specific error the 2025 framework was written to correct, and it is the reason a decommissioning or a leadership change reaches the same gate as a CIP.
2Nine domains, scored twice
Every domain carries an initial consequence and likelihood and a
revised pair after mitigation. Both are kept, because a single score
cannot show whether the mitigation moved anything — and the gap between
the two is precisely what a reviewer is being asked to accept.
Six are the National Quality Board quality domains; workforce, performance and strategic objectives sit alongside them in the assessment tool. The risk score is DERIVED from consequence × likelihood at read time, so a stored rating can never drift away from the numbers beside it.
Insufficient evidence is not a neutral impact. The
tempting simplification is to score an unevidenced domain as neutral and move
on — which converts we do not know into no harm expected
on the face of a board paper. An evidence gap is recorded as a gap, reported
separately from impact, and requires an action plan before the assessment can
be submitted.
3Where it refuses
Four conditions block submission outright, each naming the failing domains
rather than warning about them. A gate that warns is a gate that gets clicked
through.
The gate sits at SUBMIT, not at review, so the author finds out what is missing before a panel is convened rather than in front of one.
4Review, and the fail-closed default
Review has two stages and four possible outcomes: approved,
requires further information and re-submission, amendments required to
reduce impact on care quality, and not approved to proceed. Every
outcome requires a rationale — an outcome with no reasoning cannot be
challenged or learned from.
An unconfigured escalation threshold routes everything to the fuller review rather than expediting on a number nobody chose. That is the same fail-safe direction the platform applies to jurisdiction resolution: unknown sees the more rigorous path, never the convenient one.
Reassessed, never edited. A material change creates a
new version that supersedes the previous one, which is retained. What a
panel actually saw when it approved something stays readable afterwards. An
assessment that can be quietly edited after approval is not evidence of
anything, and the question an inquiry asks is what was assessed when this
decision was taken.
5The seam to financial recovery
The framework's own trigger list includes cost improvement programmes
“including the effect of cumulative CIPs within a service over
time”. That sentence is the direct join to Financial Recovery &
SIP: four individually acceptable schemes landing on one population is the
failure a per-scheme assessment cannot see, and GET /qia/cumulative
adds them up and names the schemes behind the total.
A scheme's impact level either has an assessment behind it or it does not, and the API says which. Where a QIA is linked the level is computed and the typed word is discarded.
6Where it lives, and its honest edges
Part
Where
Status
Assessment, nine domains, mitigations, reviews, RASCI
/quality-impact-assessment
Live
API
/qia/ — 15 endpoints
Live
Framework, triggers, panel roles, RASCI key
GET /qia/framework
Live
What is enforced, and what cannot be
GET /qia/methodology
Live
Cumulative impact across schemes
GET /qia/cumulative
Live
Scheme impact with provenance
GET /qia/scheme-impact
Live
Seeded assessments for demonstration
—
Demonstrated
Source provenance, stated because the API states
it. The framework, the assessment tool and the RASCI template were
supplied as files and parsed. They were not retrieved from
england.nhs.uk, which is refused by this build environment's egress policy
(measured: CONNECT tunnel failed, response 403), so any of them may
have been superseded. Read-from-a-supplied-document is a weaker fact than
retrieved-from-the-publisher and the product does not collapse the two.
One reference framework, because one was read.
Offering a second “international” or “generic” domain
set would attribute a construct to an authority nobody consulted. An
organisation in any jurisdiction may adopt this framework as good
practice — a recorded decision with a named adopter — or
define its own domain set. Both are first class; neither is a default,
and an unconfigured jurisdiction gets neither.
What commvita cannot do. It cannot judge whether an
assessment is correct — only that the steps were taken, the
evidence was cited and a named person accepted the residual risk. A well-formed
QIA can still be wrong. It supplies no clinical judgement and neither convenes
nor trains the panel. Publication is an organisational decision: commvita
records whether a QIA was published and does not publish it. And an assessment
done by email or in a document outside commvita is invisible here —
which must never be read as an absence of assessment.