Why these three: early synthetic testing of the platform's simulator pointed at three gaps worth building tools around — incomplete referrals, undocumented phone advice, and inconsistent TC decisions. That early testing is not evidence of anything about real referral patterns (it reflects how the simulator was configured, not a clinical finding) — it only motivated what to build. Tab 1's required fields and Tab 3's TC weights are grounded instead in KDIGO 2024, NICE NG203, and the KID-LINK protocol's own referral-completeness and "grey zone" definitions (see KID-LINK_Referral_TC_Criteria.md).

Structured referral builder

Every field marked required must be filled before the letter can be generated — this is the point. No more single-eGFR, no-med-list referrals.

Patient

Patient pseudonym ID *

Age / sex *

eGFR trajectory — #1 complaint: single values with no trend

eGFR now *

eGFR ~6 months ago *

Date of most recent value

Labs

uACR (mg/mmol) *

K+ (mmol/L) *

HCO3 (mmol/L)

Medications & context

Current medication list, with doses *

Home BP readings, if available

Comorbidities *

The actual question

Specific clinical question for the nephrologist *

Patient's own priorities / goals-of-care note (flagged as commonly missing for late-stage referrals)

5 required fields still empty

Referral Completeness Score: 0/5

Scored per the project's own primary outcome definition (Study I): presence of eGFR with a trend, ACR, medication list, comorbidities, and a stated clinical question — 1 point each, 0–5. This is not a KID-LINK invention; it operationalises what the project already intends to measure.

Curbside consult log

A place for the phone call or corridor conversation that currently disappears. Not a referral, not a TC — just a record that advice happened. Structured on the Champlain BASE eConsult model (Ontario, Canada): a closed 3-option response type plus a visible target turnaround, rather than free text with no taxonomy — BASE's own real-world figure is 74% of ~19,000 annual cases closed without a face-to-face specialist visit at all, which is what this structure is meant to make comparable against. See KID-LINK_Refinement_Research_Plan.md §1.5 / §3.6.

Target turnaround (business days)

BASE's own target is ~1 week with a ~1.2-day actual median; 2 days is a starting placeholder here, not a validated local figure — adjust once the supervisory team/nephrologists agree on a realistic one.

Patient ID (optional — can be anonymous topic)

Who spoke to whom

Date asked

Date resolved

Response type

Question asked

Advice given / outcome

Entries (0)

No entries yet.

TC eligibility criteria

Grounded in the project's own "grey zone" population definition (CKD stage 3b/4 + ≥2 qualifying comorbidities, per the methodology's §3.2.2) and the protocol's WP3 intervention menu — TC is one option among several (outreach visit, e-consult, contact-physician), not the default. Weights below trace to KDIGO 2024, NICE NG203, and the protocol itself — see KID-LINK_Referral_TC_Criteria.md.

Protocol-grounded population match

CKD stage 3b or 4 (eGFR 15–44) with ≥2 qualifying comorbidities — T2DM, hypertension on ≥2 agents, heart failure, anaemia of CKD, hyperuricaemia, or uACR≥30 ("grey zone", methodology §3.2.2)+3

Active, unresolved therapy decision pending (SGLT2i continuation, RRT-timing, RAASi vs hyperkalaemia trade-off)+2

Supporting clinical/relational triggers

Diagnostic uncertainty (disproportionate proteinuria, atypical trajectory)+1

Medication complexity — competing indications, electrolyte constraints+1

Goals-of-care conversation needed (frailty, dialysis-avoidance wishes) and not yet had — a named TC debrief function in the methodology+2

GP has a specific question they don't feel a letter exchange can resolve+1

Against a TC (counts negative — favours the protocol's lighter-weight options instead)

Question is narrow/factual — answerable by an online case-based e-consult or letter alone (WP3's lighter alternative to TC)−2

No slot feasible within a clinically appropriate timeframe−1

Logistics / preference — recorded, not scored

Patient/family anxiety, repeated questions, or significant travel burden — real considerations, but not part of the clinical/protocol grounding, so kept out of the score. Note them in the referral's free text instead.+0

Select criteria above to see a suggestion.

This produces a suggestion, not a rule — the GP/nephrologist judgement call stays central. Thresholds (≥5 strong case, 2–4 consider TC or e-consult, <2 standard letter) are a first grounded pass, still pending review by the supervisory team via the protocol's own WP3 co-design step.

Referral criteria check

The published, guideline-defined test for "does this patient need specialist assessment" — NICE NG203 (August 2021 update), unchanged on these thresholds as of the most recent guideline review (Nov 2023). Referral is indicated if any one of the 8 criteria below is met. This is decision support, not a diagnosis — it does not replace clinical judgement, and criteria involving suspicion (genetic cause, renal artery stenosis) or clinical assessment (BP control) require the clinician's own determination, not a lab value.

1 · Kidney Failure Risk Equation

Age (years)

Sex

Computed automatically from age, sex, eGFR and ACR (entered below) using the published 4-variable Kidney Failure Risk Equation (Tangri et al. 2011), non-North-American calibration — the version NICE NG203 and UK Kidney Association use, appropriate for a Danish/European population. Coefficients and baseline survival constants verified against two independent published sources (see kidneyfailurerisk.com for a second opinion / cross-check on any individual result). Refer if 5-year risk >5%.

Known limitation (2026-09-08): a 2025 recalibration study on a Mediterranean/European primary-care cohort found this generic non-North-American calibration tends to overestimate risk, especially in older patients and women (see criteria doc §2a for sourcing). A population-specific recalibration exists but its full formula wasn't confidently extractable from available sources, so it is not implemented here — treat a borderline result (close to the 5% threshold) in an elderly or female patient with extra caution rather than as definitive.

Override with a value calculated elsewhere (optional)

5-year risk of kidney failure (%) — manual override

Only fill this in if you have a reason to distrust the automatic calculation above (e.g. a different validated equation, or a documented discrepancy) — when set, this value is used instead of the computed one.

2–3 · Proteinuria

ACR (mg/mmol)

Known to be caused by diabetes and already appropriately treated?

Haematuria present

Refer if ACR ≥70 (unless diabetic and already appropriately treated), or ACR >30 (category A3) together with haematuria.

4–5 · eGFR trajectory over 12 months

eGFR now

eGFR ~12 months ago

Refer if a sustained fall of ≥25% accompanies a change in eGFR category (G1/G2/G3a/G3b/G4/G5), or if the fall is ≥15 mL/min/1.73m² over the 12 months regardless of category change.

6 · Blood pressure control

Antihypertensive medications

BP remains above this patient's individual target despite the above

Refer if BP is poorly controlled (above individual target) despite ≥4 antihypertensive medicines at therapeutic doses, across distinct drug classes. This tool counts distinct classes from the medication text; confirming doses are therapeutic and the target is genuinely unmet is a clinical judgement.

7–8 · Suspected cause

Known or suspected rare or genetic cause of CKD

Suspected renal artery stenosis

Criteria per NICE NG203, cross-checked against two independent clinical-reference summaries of the guideline (GPnotebook, Primary Care Notebook) on 2026-09-08. All 8 criteria are now checked automatically, including the Kidney Failure Risk Equation (§1) — see the KID-LINK_Referral_TC_Criteria.md changelog for the formula sources and verification method.

Framework comparison

Enter a patient's data once, see what three independently-sourced referral algorithms each conclude — NICE NG203 (UK), the Ontario KidneyWise Toolkit (Canada, a real deployed primary-care algorithm), and KDIGO 2024's own categorical + risk-based criteria. This is exploratory, not a fourth rule to follow — the point is seeing where three real, differently-designed systems agree and where they don't, on the same patient. Where a guideline's own criterion isn't numeric (e.g. KDIGO's "sustained decrease in eGFR" gives no specific %), this tool says so explicitly next to that result rather than inventing false precision. Sourcing: KID-LINK_Referral_TC_Criteria.md §2 (NICE, KDIGO) and KID-LINK_Refinement_Research_Plan.md §1.1 (KidneyWise).

Known limitation shared by NICE and KidneyWise here (both use KFRE): the non-North-American calibration used across this tool is a generic international default. A 2025 recalibration on a Mediterranean/European cohort found it tends to overestimate risk, especially in elderly and female patients — see the criteria doc §2a. Not yet implemented (the population-specific formula wasn't confidently extractable from available sources); a borderline KFRE-driven verdict in that group deserves extra clinical judgement.

Patient

Age (years)

Sex

Known diabetic (and, if so, already appropriately treated)

eGFR trajectory

eGFR now

eGFR ~6 months ago

eGFR ~12 months ago

6-month value feeds KidneyWise's rate-of-decline trigger; 12-month feeds NICE's two eGFR criteria. Leave either blank if not available — that framework's trajectory-based criterion(a) will show as not assessable rather than defaulting to a guess.

Proteinuria / haematuria

ACR (mg/mmol)

Haematuria present

Labs

K+ (mmol/L)

HCO3 (mmol/L)

Feeds KDIGO's "metabolic abnormality requiring specialist management" criterion — this tool treats K+ >5.5 or HCO3 <22 as meeting it, a threshold choice made here (not itself dictated by a specific KDIGO number), since KDIGO names the category without giving one.

Blood pressure

Antihypertensive medications

BP remains above this patient's individual target despite the above

NICE's resistant-hypertension criterion requires ≥4 antihypertensive classes; KidneyWise's requires ≥3 — same medication list, different thresholds, which is exactly the kind of disagreement this tab is for.

Other red flags

Known or suspected rare/genetic cause of CKD

Suspected renal artery stenosis

Suspected glomerulonephritis/vasculitis (KidneyWise-specific red flag, not in NICE's 8 or KDIGO's list)

Recurrent kidney stones needing metabolic workup (KidneyWise-specific)

Complex/uncertain case needing an advanced therapy decision (SGLT2i, RAS-inhibitor titration, MRA) — KDIGO-specific

Diagnostic uncertainty about the cause/trajectory of this patient's CKD — KDIGO-specific

None of these three verdicts overrides another — a disagreement between them is itself the useful output here (e.g. NICE's ACR≥70 threshold vs KidneyWise's diabetic-split 30/60 will genuinely disagree on some real patients). Not clinically validated for this population; see KID-LINK_Referral_TC_Criteria.md §5.

referral-toolkit.html · build 2026-09-07