1 · Pilot Rationale

Finnish health registries are among the most comprehensive in the world. The integration failure documented in WP-016 is not a data absence problem — it is an allocation problem. HDCI is a diagnostic instrument for identifying where integration failure is highest, enabling resource allocation decisions under conditions of fiscal constraint.

Pohjois-Savo is selected as the pilot welfare area for three structural reasons:

2 · Data Access Pathway

All required data exists within Finnish health registries. Patient-level access requires formal secondary use licensing through Findata (the Health and Social Data Permit Authority).

Required datasets

Licensing timeline

Current status: Step 1 not yet initiated. This document is preparatory material for the pre-application consultation.

3 · Cohort Definition

Inclusion criteria

Exclusion criteria

Expected cohort size: 30,000–40,000 patients (based on ~250,000 population, ~15–20% adult multi-morbidity prevalence).

4 · Calculation Protocol

DCI — Data Capture Index

IAI — Integration Achievement Index

RVI — Response Velocity Index

RKI — Risk Accumulation Index

PAI — Prevention Activity Index

HDCI v1 — Core Index (v0.7)

DCI — Diagnostic side-indicator (not in composite)

PAI — Dashboard metadata (not in composite)

5 · Validation Hypotheses

H4 is the primary research contribution of this pilot. If supported, it demonstrates that HDCI reveals a high-risk patient state that conventional metrics systematically miss. If unsupported, it constrains the interpretation of HDCI to a measurement instrument without topological properties — which remains valid and useful.

Statistical methods: H1–H3: multilevel regression, patient-level predictors, outcome measures as dependent variables, controlling for age, sex, morbidity index. H4: k-means clustering (k=3) and latent class analysis in RKI–IAI–RVI space; cluster outcome comparison via ANOVA and logistic regression.

6 · Analogy: WEM → HIM

DT-006 pilot results will populate a conceptual dashboard structurally analogous to the Winter Endurance Monitor (WEM). The Health Integration Monitor (HIM) is a design target for post-pilot development.

6 · Measurement Validity and Identifiability — v0.6

The following validity threats are identified in advance of data access. Each is addressed in the pilot design. This section is a prerequisite for a credible Findata application.

DCI — Endogeneity and exogenous instrument

DCI has a structural endogeneity problem: morbidity drives care contacts, which drive Kanta entries — the same causal structure produces both predictor and outcome. Without correction, DCI risks measuring documentation intensity rather than data capture quality. The pilot adds Avohilmo contact rate as an exogenous instrument:

This separates documentation completeness (Kanta entries per contact) from care frequency (contacts per patient). Negative binomial regression; sensitivity vs Poisson and stratum-mean.

IAI — Structural proxy, not clinical truth-metric

IAI is a structural proxy for integration, not a clinically invariant measure. The 30-day episode window is a heuristic that treats all care episodes identically — false in clinical reality across diabetes, orthopaedics, and psychiatry. IAI measures whether the system produces multi-specialty contact within a defined window: a system-level observation, not a patient-level truth claim. Sensitivity analysis: 15, 30, 90-day windows. Unstable rankings require disease-stratified analysis.

RVI — Institutional throughput, not clinical appropriateness

RVI measures institutional response velocity — how quickly the system converts a recorded finding into a recorded action. It does not measure clinical appropriateness. "First finding" is not neutral: it depends on testing frequency and coding practices. RVI is a system-level throughput indicator; this scope must be stated explicitly in all publications.

RKI — Adjusted for Kanta adoption rate (v0.6 revision)

RKI is the strongest HDCI component. v0.6 introduces an explicit adjusted measure:

RKI_adjusted isolates coordination failure from documentation infrastructure variation. Composite HDCI uses RKI_adjusted. Both values reported. Cross-validated against H2 (adverse medication events).

Weight revision — v0.6

RKI: 0.20 → 0.25 (strongest component, direct outcome validation). PAI: 0.10 → 0.05 (input-effort indicator only; does not measure prevention effectiveness). All weights remain Bayesian priors subject to outcome-constrained re-estimation.

HDCI as composite proxy index

HDCI does not model "integration" as a unitary latent variable. Integration is multidimensional — information flow, responsibility coordination, and clinical decision consistency are distinct constructs. HDCI is a composite proxy index: a diagnostic signal about where integration-relevant observables are weakest, not a measurement of an underlying integration quantity.

Causal structure — DAG

EXOGENOUS
  Morbidity · Age/Sex · Avohilmo contact rate · Kanta adoption rate

  Morbidity + Age/Sex → Care Contacts ← Avohilmo rate (instrument)
  Care Contacts → Kanta entries (DCI) · Multi-spec episodes (IAI)
                → First finding → First action (RVI)
  Prescriber multiplicity + Kanta adoption → RKI_raw → RKI_adjusted
  Preventive contacts / total → PAI

  HDCI composite → H1 Preventable hospitalisations
  RKI_adjusted  → H2 Adverse medication events
  HDCI components → H3 Inter-municipality variation

8 · From Topology to System Behaviour

HDCI does not prescribe interventions. It reveals where the system's own coordination logic produces elevated risk without corresponding visibility in current metrics. This section describes what the RKI–IAI–RVI space shows about system behaviour — not what should be done about it.

8.1 Core principle: HDCI is a friction map, not a control system

The three-dimensional RKI–IAI–RVI space identifies:

• Where information does not become action — RVI: response latency
• Where action does not coalesce into responsibility — IAI: fragmentation
• Where responsibility diffusion produces risk before it is visible — RKI: silent ownership collapse

HDCI does not tell the system what to do. It tells the system where its own coordination logic is producing the highest unobserved friction.

8.2 Intervention framing: transitions, not patient classification

The diagnostic value of HDCI lies not in labelling patients but in identifying the transition zones where patients move from coordinated care into high-risk invisibility. Any interventions, if undertaken, address the drift mechanism — not individual patients.

8.3 Three observable system behaviours

1. RKI → IAI coupling — responsibility diffusion before fragmentation. When RKI rises without IAI decrease, medication responsibility diffuses across prescribers before the care pathway fragments into multiple uncoordinated specialties. The system accumulates polypharmacy risk silently. Minimal feedback: a Kela + Kanta medication consolidation view — not a new workflow, only increased visibility of prescriber multiplicity.

2. IAI → RVI coupling — coordination without decision. When IAI is high but RVI remains low, the system makes contact across multiple specialties, records findings, but does not convert findings into timely action. The system sees everything but does not react. Minimal feedback: organisational-level latency reporting — an escalation log, not automated decision support.

3. RKI × RVI risk zone — silent high-risk drift. When RKI is high and RVI is low simultaneously, responsibility is maximally diffused and response velocity is minimal. No current metric (cost, queue, morbidity) flags this state. Minimal feedback: a review flag — not a clinical decision, only a signal that the coordination profile warrants attention.

8.4 Why this framing is necessary

If HDCI is presented as "Group 3 requires intervention," it becomes a clinical decision model — which it is not — and loses standing as a measurement framework. Presented as "HDCI reveals structural friction zones that current metrics cannot see," it remains a diagnostic instrument. This distinction determines whether DT-006 remains a research design or becomes an unimplementable policy proposal.

8.5 Institutional hyväksyttävyys — viisi periaatetta

7 · Milestones and Status

Document versioning