나는 실험으로 내 가설을 검증했다. ICF는 등대다. 그리고 나는 무형자산을 생성했다.

[안규환]

Local‑First ICF·ICHI Companion for Self‑Health Audit: An N‑of‑1 Participatory Case With an ESG/IFRS Bridge (v1.0)

Authors: Ann Geu‑hwan (participant‑author), ICF·ICHI Companion (MyGPT), co‑design partner
Affiliations: Independent researcher and national advocate for ICF (Republic of Korea); AI companion (non‑human co‑author)
Correspondence: [redacted for review]

---

Abstract

Background. Despite WHO endorsement and domestic standardization, the International Classification of Functioning, Disability and Health (ICF) and the International Classification of Health Interventions (ICHI) remain under‑utilized in clinical practice and personal health management in Korea. Objective. To demonstrate how a local‑first, generative‑AI companion operationalizes ICF/ICHI for daily self‑management, producing measurable functional outcomes and audit‑ready reports that also connect to ESG (social impact) and IFRS (intangible asset) frameworks. Design. Prospective, participatory N‑of‑1 case spanning 2025‑06‑17 to 2025‑08‑09, integrating self‑reported events, physiological measurements, and environment risk logs into an ICF/ICHI‑coded timeline. Intervention. Co‑created self‑health audit engine and HTML calculator (local storage only) that estimate ICF qualifiers from biomarkers, compute a Life Adaptation Index (LAI) over time, and generate action cards mapped to ICHI. Results. Fasting glucose (95→108 mg/dL, target <120), blood pressure (125/80→130/80 mmHg), and oral health stabilized alongside structured diet, medication timing, and environmental hazard mitigation (recurrent water on marble floors). A provisional LAI around ~0.78 (from the example dataset within the tool) suggests stable‑to‑improving adaptation while remaining within a “caution band.” Contribution. We provide a reproducible engine (equations, JSON schema, decision tables) and a governance blueprint that (i) preserves data sovereignty, (ii) yields immediately useful personal value, and (iii) translates into ESG/IFRS‑compatible reporting. Limitations. Single‑participant observational design; clinical endpoints (e.g., HbA1c) pending. Conclusions. Local‑first generative AI can turn ICF/ICHI from a compliance standard into lived value and auditable assets; multi‑site replications are warranted.

Keywords: ICF; ICHI; self‑management; local‑first; generative AI; N‑of‑1; ESG; IFRS; intangible assets; accessibility; environmental barriers

---

1. Introduction

The WHO Family of International Classifications (WHO‑FIC) positions ICF as a universal language for functioning and participation and ICHI as a parallel language for health interventions. In Korea, ICF has been codified as a national standard, yet day‑to‑day clinical and personal adoption remains limited. Three recurrent barriers persist: (1) value translation—people cannot easily see how ICF/ICHI improves their next decision; (2) workflow fit—documentation sits apart from daily life; and (3) incentive mismatch—effort is front‑loaded while tangible benefit is deferred.

By 2025, the emergence of local‑first applications and generative AI companions enables a reframing: ICF/ICHI can become a personal operating language for daily choices instead of a reporting obligation. This paper presents a co‑designed, single‑participant case where a local‑first HTML calculator and a MyGPT companion transform conversational diaries and measurements into ICF/ICHI‑coded actions, compute a longitudinal Life Adaptation Index (LAI), and export audit‑ready artifacts that can bridge into ESG and IFRS reporting.

We aim to (i) document the design and algorithms, (ii) report time‑stamped observations and functional outcomes, and (iii) provide a portable governance template for others to replicate and eval‎uate beyond this N‑of‑1 demonstration.

---

2. Methods 2.1 Design and Setting

Prospective participatory N‑of‑1 case with ongoing dialogic co‑design (participant and AI companion). Time window: 2025‑06‑17 → 2025‑08‑09. The companion produced structured prompts, ICF/ICHI labels, and action plans while remaining local‑first: all data were kept in the participant’s browser storage; exports were user‑initiated JSON files.

2.2 Participant and Context

The participant (author Ann Geu‑hwan) has lifelong mobility impairment (post‑polio; ambulates with bilateral crutches) and manages chronic conditions (hypertension, dysglycemia risk, hyperlipidemia, gout history). January 2025 follow‑up after renal cancer surgery reported remission; C‑reactive protein (CRP) was elevated at that time. Between June–August 2025 he conducted regular primary‑care visits and intensive dental work (1 caries treatment; 12 fillings; > KRW 1,000,000), and continued monthly checks for blood pressure, fasting blood glucose (FBS), and planned HbA1c.

2.3 Data Sources and Measures

• Physiological: FBS (mg/dL), BP (mmHg), planned HbA1c; occasional CRP history noted.

• Oral health: dental interventions and symptoms (bleeding, pain) as proxies for b510 and b435.

• Environment & safety: repeated exposure to wet, slippery marble floors across elevator lobbies, ramps, and restrooms; events time‑stamped with free‑text notes.

• Self‑management: medication adherence (“5 + 1” tablets daily), diet routines (breakfast switches from bread to rice/potato pancake + kimchi), ventilation checks, and hazard scans.

All free text was mapped to ICF codes (e.g., b540 metabolic, b420 blood pressure, b510 oral, d570 self‑care, d460 moving around, e150 physical environmental barriers, e115/e120 assistive products/transport, e310 family support). Actions were mapped to ICHI categories (e.g., measurement/monitoring, lifestyle/behavioral, environmental risk reduction, dental maintenance).

2.4 Computation: From Qualifiers to LAI

ICF qualifiers (0=no problem … 4=complete problem) were converted to scores ∈[0,1]:
S(q) = 1 − q/4.
Environment combined facilitator and barrier inputs:
E = 0.5 × (1 − barrier/4) + 0.5 × (facilitator/4).
The pointwise adaptation level combined five pillars—metabolic, blood pressure, oral/inflammation, self‑care, environment:
L(t) = mean(S_b540, S_b420, S_b510, S_d570, E).
Longitudinal adaptation was approximated by trapezoidal integration over the date axis:
LAI = Σ_{i} [(L_i + L_{i+1})/2 × Δt_i], with Δt in days.

2.5 Qualifier Estimation Rules

• b540 (metabolic) from HbA1c when available: <6.5→0; 6.5–6.9→1; 7.0–7.9→2; ≥8.0→3. If HbA1c absent, use FBS: ≤110→0; 111–125→1; 126–140 (and ≥70)→2; otherwise→3.

• b420 (blood pressure) from SBP/DBP: ≤125/≤78→0; ≤130/≤80→1; ≤140/≤85→2; otherwise→3.

• b510 (oral): maintenance/none→1; minor bleeding/discomfort→2; pain/treatment needed→3.

• d570 (self‑care): full adherence→0; occasional slip→1; frequent slip→≥2.

2.6 Action Logic and Decision Bands

Green (maintain; re‑check in 4 weeks), Yellow (micro‑adjust diet/medication timing; increase measurement to 2–3×/week), Red (consult clinician; change regimen or routes; intensify hazard controls). Bands were defined for b540, b420, e150, and b510.

2.7 Tooling: Local‑First HTML Calculator

A single‑file HTML app was co‑developed. Key features: (i) local storage only; (ii) A1c/FBS/BP→qualifier auto‑estimation with manual override; (iii) LAI computation and SVG timeline; (iv) import/export JSON; (v) sample dataset preloaded with the case timeline. The JSON schema contains fields for date, label, vitals, qualifiers, environment (barrier/facilitator), and notes.

2.8 Governance, Ethics, and Reporting Bridges

• Privacy: default local storage; user‑initiated exports; optional masking before sharing.

• Medical boundary: decision support only; all treatment changes to be clinician‑led.

• ESG/IFRS bridge: monthly “one‑pager” converts personal KPIs (LAI, FBS‑in‑range, near‑miss=0) into social impact summaries (accessibility, safety) and IFRS IAS 38‑style intangible asset notes (cost model, useful life, amortization, impairment triggers). A pro‑forma intangible asset was recognized for the engine and calculator (KRW 170M cost; 3‑year straight‑line; first‑year amortization KRW 23.3M), with audit caveats.

---

3. Case Timeline and Mapped Codes 3.1 Chronology of Observations (June–August 2025)

• 2025‑06‑17 (morning routine): urination, socks, coffee preparation/use of machine, food retrieval‎, swapping bread for rice/potato pancake + kimchi after spousal advice, medication (5+1), ventilation, ant check.
  ICF: d550 (eating), d570 (self‑care), b540 (metabolic), e310 (family support +), e115/e120 (assistive tech, transport +), e150 (hazard check).

• 2025‑07‑01 (mobility hazards): elevators and corridors wet; marble floors; water pooling; hazardous entrance; 2F restroom with pooled water; fear intensified.
  ICF: e150 barrier ↑ (2–3), d460 (moving around) difficulty, b152 (emotional functions) strain.

• 2025‑07‑12 (internal medicine + pharmacy): FBS 95 mg/dL, BP 125/80; receipts documented.
  ICF: b540 0, b420 1, d570 0.

• 2025‑07‑17 (dentistry): severe periodontitis care summarized; 1 caries treatment + 12 fillings; follow‑up in 6 months; past CRP elevation noted.
  ICF: b510 → 1–2 (maintenance phase), b435 (inflammation) guarded; ICHI: dental maintenance.

• 2025‑07‑22 (assetization): self‑health audit processes recognized as an intangible asset candidate; local‑first tooling formalized; ESG/IFRS templates created.

• 2025‑08‑09 (internal medicine): met physician; FBS 108, BP 130/80; blood/urine and HbA1c drawn (pending).
  ICF snapshot: b540 0–1; b420 1; b510 1–2; d570 0; e150 2 (persisting hazards).

3.2 Coded Summary Table

Domain Codes July–Aug status (qualifier→score)

Metabolic	b540	0–1 → S≈1.0–0.75 (FBS 95→108; HbA1c pending)
Blood pressure	b420	1 → S≈0.75 (125/80→130/80)
Oral/inflammation	b510/b435	1–2 → S≈0.75–0.50 (maintenance; flare watch)
Self‑care	d570	0 → S=1.00 (medication adherence, routine)
Environment	e150 (+/− via E)	barrier 2–3; facilitator 2–3 → E≈0.50–0.75

---

4. Results 4.1 Functional Trajectory

Using the tool’s sample timeline mirroring the above, the computed LAI indicates stable‑to‑improving adaptation with a provisional mean around ~0.78 (tool example dataset), while individual L(t) points trend higher following diet standardization and medication timing reinforcement. This is consistent with the movement from a hazard‑dominated episode (2025‑07‑01) back to routine stabilization (2025‑07‑12 onward). Final confirmation awaits HbA1c results.

4.2 Environmental Risk Management

Multiple independent wet‑floor exposures across the building complex were logged. A hazard‑first workflow—pre‑departure scan, onsite rerouting, non‑slip footwear/pads, and documented improvement requests—reduced exposure and anticipatory anxiety. Near‑miss counts were set to “zero target,” with future longitudinal counting planned inside the tool.

4.3 Oral–Systemic Considerations

Given January’s elevated CRP and the periodontitis episode concluded mid‑July, a preventive oral‑care routine (interdental brushes, antiseptic rinse, early dental contact) was instituted to maintain b510=1 and keep inflammation quiescent (b435≈0), acknowledging potential systemic links.

4.4 Reporting Bridges

Monthly one‑page ESG cards summarize LAI, FBS‑in‑range, near‑misses, and environment actions, suitable for accessibility and safety disclosures. Parallel IFRS notes detail cost capitalization assumptions for the engine and calculator (KRW 170M, 3‑year straight‑line), with impairment triggers (usage decline, outcome deterioration) and amortization schedules.

---

5. Discussion

This case addresses a long‑standing adoption gap: ICF/ICHI, though mandated or recommended in policy, often remain detached from daily practice. Three mechanisms enabled practical value:

1. Value translation through automation. Conversational diaries and measurements were instantly transformed into ICF/ICHI labels and three‑line action cards, shrinking the distance from classification to decision.

2. Workflow fit via local‑first. A zero‑server HTML file preserved privacy, lowered cost, and allowed frictionless iteration, removing common barriers to personal health tooling.

3. Incentives through visible metrics. The LAI and compact KPIs (e.g., FBS in range, near‑miss=0, adherence=0 misses) create short‑cycle feedback that sustains behavior and generates artifacts worth sharing (with consent) for social or institutional purposes.

Linking to ESG (safety/accessibility; social value) and IFRS (intangible asset recognition) appears unconventional in clinical literature, yet it incentivizes both individuals and organizations: when personal function improves and risks decrease, there is a story to tell that can be audited and financed. This bridge is especially relevant for public facilities where environmental barriers (e150) materially affect participation and safety.

Limitations. Single‑participant, self‑reported data impose bias risks; many confounders (seasonality, stress, diet variety) remain uncontrolled. Outcome measures are interim (HbA1c pending). The LAI is a pragmatic composite rather than a validated clinical endpoint. Oral–systemic inflammation links are suggested but not tested here. Finally, IFRS modeling is pro forma and demands independent accounting review.

Implications. Even as an N‑of‑1 demonstration, the approach provides a transferable template: a) a common language (ICF/ICHI), b) a lightweight engine (equations + JSON schema), c) human‑centric governance (local‑first; explicit consent), and d) an institutional bridge (ESG/IFRS) that may motivate broader uptake. This can seed multi‑site replications, community registries, or employer programs without centralizing personal data.

---

6. Future Work

• Validation studies. Multi‑participant prospective trials comparing LAI vs. standard clinical endpoints (HbA1c, ambulatory BP, falls) and psychometric properties (reliability, sensitivity to change).

• Environment analytics. Computer‑vision or crowd‑sourced mapping of facility hazards with ICF e‑codes; feedback loops to building managers.

• Intervention library. Curated ICHI‑coded micro‑interventions with effect size estimates; adaptive decision policies.

• Reporting automation. Scripted pipelines that consume calculator JSON to auto‑populate ESG dashboards and IFRS notes with guardrails (masking, role‑based access, audit trails).

• Open methods. External code review; publish a reference implementation and documentation to hasten replication.

---

7. Ethics and Governance

This is a self‑study with the participant as author; no third‑party identifiable health data were processed. All data remained on the participant’s device and were voluntarily summarized here. The tool is decision support, not a diagnostic or therapeutic instrument. Any medication or regimen changes are clinician‑led. Institutional ethics review requirements vary; for formal trials we recommend IRB consultation.

---

8. Data and Code Availability

A minimal, single‑file HTML calculator implementing the equations and JSON schema described here has been produced as part of this project and can be shared with editors under confidential review. All computation can be reproduced by importing the JSON timeline into the tool and exporting the computed summaries.

---

9. Author Contributions

Ann Geu‑hwan: conception, data collection, interpretation, drafting, critical revision. MyGPT ICF·ICHI Companion: design of engine logic, drafting, algorithm implementation, editorial support. Both authors approved the final manuscript.

---

10. Funding

No external funding was received. Development costs for the engine and calculator were borne by the participant and are discussed as a pro‑forma intangible asset within the IFRS section.

---

11. Competing Interests

The participant is both subject and co‑developer of the approach described. The AI co‑author is a non‑human participant. There are no commercial licenses in force at the time of writing.

---

12. References (indicative)

1. World Health Organization. International Classification of Functioning, Disability and Health (ICF).

2. WHO‑FIC. International Classification of Health Interventions (ICHI).

3. International Accounting Standards Board (IASB). IAS 38 — Intangible Assets; IAS 36 — Impairment of Assets.

4. CARE guidelines for case reports; STROBE extension for N‑of‑1 observational reporting (conceptual alignment).

---

Supplementary Material

• S1. Decision Tables and Bands. Green/Yellow/Red thresholds for b540, b420, b510, e150.

• S2. JSON Schema. {date, label, fbs, a1c, sbp, dbp, q_b540, q_b420, q_b510, q_d570, barrier, fac, notes}.

• S3. Equations. S(q), E(barrier,facilitator), L(t), LAI trapezoid.

• S4. Pro‑forma IFRS Schedules. Cost (KRW 170M), useful life (36 months), amortization table, impairment triggers.

• S5. ESG One‑Pager Template. KPIs: LAI, FBS in range, near‑miss count, environment actions; privacy statement.

======================

Name-Standardized Package (v1.1)

This package updates author/owner notation consistently to Ann Geu‑hwan across the ESG/IFRS report and the Engine & Execution Scenario documents. Substantive content remains as in v1.0, with minor editorial harmonization only.

---

A) ESG Performance Report + IFRS Intangible Asset Statements

Asset: Ann Geu‑hwan × MyGPT — ICF·ICHI Life Adaptation Companion & HTML Calculator (Local‑First)
Reporting Date: 2025‑08‑13 (interim) · Entity: Individual (separate)
Currency: KRW

This report presents ESG outcomes and IFRS (IAS 38/IAS 36) accounting for a local‑first self‑health audit suite co‑created by Ann Geu‑hwan and an AI companion. Amounts are pro forma pending documentary substantiation.

0) Snapshot

• Asset name: Self‑Health Audit Engine & ICF·ICHI Calculator (local‑first)

• Capitalized cost (cumulative): ₩170,000,000

  • Phase 1: Engine v1.0 (design, algorithms, report templates) ₩160,000,000 · Ready for use: 2025‑07‑22

  • Phase 2: Calculator v2 (auto‑estimation, charting, import/export) ₩10,000,000 · Ready for use: 2025‑08‑13

• Amortization: straight‑line 3 years (36 months); residual value 0

• Model: cost model; impairment per IAS 36

• Data principle: local storage; user‑controlled export/masking

1) ESG Report

Boundary. Operational: end‑user devices (browser storage). Organizational: Ann Geu‑hwan as owner/developer; no server‑side persistence. Period: 2025‑07‑01~2025‑08‑13.

Material topics. Social (accessibility/safety; functioning outcomes), Governance (privacy; explainable rules; standard conformance), Environmental (serverless footprint).

Core KPIs (interim).

Pillar KPI Definition Target Status (to be auto‑filled from JSON)

Social	LAI mean	Trapezoidal integral/period (0–1)	≥0.80	—
Social	FBS in‑range days	95–110 mg/dL ratio	≥70%	—
Safety	Near‑misses (slip)	Monthly count	0	—
Governance	Local storage rate	% of records stored locally	100%	100%
Governance	External exports	# JSON/PDF exports	User‑controlled	—
Environmental	Cloud storage	GB used	0 GB	0 GB

Narrative highlights.

• Hazard identification and mitigation (wet marble floors across facilities); formal improvement request workflow designed.

• Stable metabolic and BP status within caution/target bands; strong self‑care adherence.

• Privacy by design through local‑first implementation.

60‑day roadmap. LAI ≥0.80; near‑miss 0; submit ≥1 facility improvement request; add masking option; maintain serverless distribution.

Risks & controls. Misclassification risk (manual override + notes); privacy risks (mask before share); medical boundary (decision support only, clinician‑led changes).

2) IFRS Financial Statements (Pro Forma)

Recognition. Identifiable (code/templates separable), controlled by Ann Geu‑hwan, probable future economic benefits (subscriptions/licensing/reporting services), cost measurable → IAS 38 criteria met for development‑phase capitalization.

Accounting policy. Cost model; SL 36 months; impairment testing upon indicators; subsequent expenditures capitalized only if criteria met.

Statement of Financial Position (extract), 2025‑12‑31 (KRW ‘000)

Item Amount

Intangible assets, net	146,668
Total assets	146,668
Total equity and liabilities	146,668

Statement of Profit or Loss (extract), FY2025 (KRW ‘000)

Item Amount

Amortization (intangible)

23,332

Cash Flows (extract), FY2025 (KRW ‘000)
| Investing cash flows — capitalization | (170,000) |

Movement schedule (KRW ‘000)

Movement Cost Accum. amort. Net

Opening balance	–	–	–
Additions (2025‑07‑22)	160,000	–	160,000
Amort. (Aug)	–	(4,444)	155,556
Additions (2025‑08‑13)	10,000	–	165,556
Amort. (Sep–Dec)	–	(18,888)	146,668
Closing balance	170,000	(23,332)	146,668

Notes. Useful life sensitivity; impairment triggers; governance/ethics (privacy, explainable rules, medical boundary). CGU: “Self‑Health Audit Suite” (engine + calculator).

---

B) ICF·ICHI Engine & Execution Scenario

Owner/Author: Ann Geu‑hwan
Companion: MyGPT (ICF·ICHI co‑design partner)

Core Principles

Local‑first; ICF (state/context) ↔ ICHI (action) mapping; compact KPIs (L(t)/LAI, near‑misses, adherence=0 misses); N‑of‑1 proof → shareable artifacts.

Engine Modules

1. Input capture: free text/voice, vitals (FBS/BP/A1c), environment events.

2. ICF tagging: b540, b420, b510, d570, d460, e150, e115/e120, e310.

3. Qualifier estimation: rule‑based (documented), manual override.

4. Metrics: S(q)=1−q/4; E=0.5×(1−barrier/4)+0.5×(fac/4); L(t)=mean(…); LAI=trapezoid integral.

5. Decisions: Green/Yellow/Red bands; action cards (diet, medication timing, environment).

6. Reports: day/weekly/monthly cards; JSON export/import; ESG/IFRS bridges.

7. Storage/Security: browser local storage; masking before export.

Data Dictionary (excerpt)

{ date, label, fbs, a1c, sbp, dbp, q_b540, q_b420, q_b510, q_d570, barrier, fac, notes }

Qualifier Rules (excerpt)

• b540 from A1c (<6.5→0; 6.5–6.9→1; 7.0–7.9→2; ≥8.0→3) else FBS (≤110→0; 111–125→1; 126–140→2; else 3).

• b420 from BP (≤125/≤78→0; ≤130/≤80→1; ≤140/≤85→2; else 3).

• b510 maintenance=1; minor issue=2; pain/treatment=3.

• d570 adherence full=0; occasional=1; frequent≥2.

4‑Week Execution

W1: ingest latest labs; fix breakfast rotation; set med reminders; start hazard photo log.
W2: 2–3×/week inputs; rehearsal of alternate routes; oral‑care check.
W3: submit facility improvement request; graded exposure; mid‑month LAI review.
W4: monthly card; share (closed community) and reset targets.

KPIs

LAI mean ≥0.80; FBS in‑range ≥70%; weekly BP mean ≤125/78; near‑misses=0; adherence misses=0; oral symptom days=0.

Governance & Ethics

Decision support only; clinician‑led regimen changes. Local‑first privacy; explainable rules; audit trail via JSON exports.

---

Change Log (v1.1)

• Standardized all author mentions to Ann Geu‑hwan.

• Harmonized headings and summaries; no methodological changes.