HEARTLAND FHIR candidate 0.3.0 — draft, experimental R4 4.0.1 reference. Prepared for technical review; not evidence of clinical validation or vendor interoperability.

Risk Assessment

Risk Assessment Workflow

This draft represents a proposed, unvalidated heuristic with ten weighted criteria, 0–18 points and three qualitative tiers. It is an educational implementation-support representation, not a patient-care instruction or outcome prediction.

Inputs (historical score definition retained for compatibility)

linkId	Item text	Points
age-75	Age >=75 years	2
hf-hosp-6mo	Prior heart failure hospitalization within 6 months	3
egfr-low	eGFR <45 mL/min/1.73m^2	2
natriuretic-high	BNP >=500 pg/mL or NT-proBNP >=1500 pg/mL	2
sbp-low	Systolic BP <100 mmHg at admission	2
diabetes	Diabetes mellitus	1
lvef-low	LVEF <30%	2
ckm-stage	Cardiovascular-kidney-metabolic Stage 3 or 4	2
distance-far	Distance to cardiology care >50 miles	1
social-support	Lives alone or limited social support	1
Total possible		18

Tier Cutoffs

Total score	Tier	Historical heuristic label, not a prescription
0-4	Low	Standard monitoring
5-8	Moderate	Enhanced monitoring bundle
>=9	High	Intensive monitoring bundle

FHIR Workflow

1. Capture complete inputs using HeartlandRiskInputResponse, referencing HeartlandRiskInputQuestionnaire with version |0.3.0. This specialized profile requires all ten unique link IDs in Questionnaire order, each with exactly one actual Boolean value. Nested, missing, extra and duplicate items are rejected. Status must be completed or amended; this is not clinical approval.
2. Interpret weights by link ID, using each item's HeartlandRiskTruePoints extension. Explicit true contributes the stated weight; explicit false contributes zero. Missing or unknown is not false. A _valueBoolean containing only an absent-data extension is not an answer that can be scored. The extension allows exactly one weight, between 1 and 3, per Boolean item. These are data constraints, not an automated clinical adjudication.
3. Map score to tier using the cutoff table above.
4. Create a HeartlandRiskAssessment with:
   • prediction.qualitativeRisk bound to the resulting HeartlandRiskTier code: low, moderate or high.
   • the heartland-risk-score-total extension on prediction carrying the integer total (0-18).
   • basis referencing the complete response, with separately checked subject and score consistency.
   • method.text exactly "HEARTLAND Protocol v3.2 Risk Score".

Where the Point Total Goes

prediction.probability[x] is prohibited by the profile. FHIR R4 defines that element as the likelihood of a specified outcome, expressed as a percentage; the HEARTLAND total is a count of heuristic points, and writing it there would publish "11 points" as "11% chance of an event". The tier is the canonical result and lives in prediction.qualitativeRisk; the total lives in the heartland-risk-score-total extension.

Legacy Observation and generic QuestionnaireResponse references remain allowed by RiskAssessment.basis for compatibility. An Observation representation uses the HeartlandRiskScore code system, code heartland-risk-score, and a point total rather than a probability. A reference alone does not validate its target, confirm subject identity, recompute the total or prove conformance of an entire Bundle. Only the specialized response profile defines the complete Boolean capture contract; it does not automatically validate a separate RiskAssessment's calculation.

The method text HEARTLAND Protocol v3.2 Risk Score is retained as a compatibility identifier. It is not a statement that the candidate IG or current Toolkit has version 3.2. Historical resources are not rewritten.

Patient Extensions

Two HEARTLAND-specific Patient extensions surface the score's social determinants:

• heartland-distance-to-cardiology — Quantity in miles ([mi_us] UCUM)
• heartland-social-support-score — boolean (true = limited social support)

Use the HeartlandPatient profile to bundle both.

The heartland-synthetic-county-code extension belongs on Patient.address and carries valueCoding with system https://fhir.heartlandprotocol.org/sid/synthetic-county-code and a required actual code. Preserve the opaque code; it is not an ANSI/FIPS county GEOID, postal code or evidence of real residence. The extension also applies to unprofiled synthetic Patient resources; its definition does not certify the rest of an export.

Worked Example

PatientExampleRural is a synthetic example with an illustrative birth date, Montana state label, distance of 87 miles and limited support. The data do not identify a real patient or establish residence in a real county.

QuestionnaireResponseExampleRiskInputs records six of ten items as true:

• age-75 (+2)
• hf-hosp-6mo (+3)
• sbp-low (+2)
• ckm-stage (+2)
• distance-far (+1)
• social-support (+1)

Total = 11 points -> High Risk heuristic tier. No clinical action, notification or outcome is inferred.

RiskAssessmentExampleHigh records this result and references the questionnaire response as basis.

Evidence Note

The score's pragmatic label identifies a proposed heuristic pending validation, not established clinical utility. This guide's tests concern data representation and constraints. They do not establish discrimination, calibration, safety, clinical effectiveness or suitability for patient care.