Skip to content

06-module-indicators-hapi.md

Latest commit

 

History

History
110 lines (82 loc) · 7.91 KB

06-module-indicators-hapi.md

File metadata and controls

110 lines (82 loc) · 7.91 KB
Raw

06 — Module ③ Indicators: the Healthy Aging Policy Index (HAPI)

中文概览

这是作者最期待的模块。核心主张:不直接照搬政府指标,而是自建一个透明、可复现的独立指数——Healthy Aging Policy Index(HAPI)。

  • 六个一级域:Health(健康)、Independence(独立性)、Social Participation(社会参与)、Financial Security(经济保障)、Care Access(照护可及性)、Digital Inclusion(数字包容)。
  • 每个子指标必须标注:定义、计算公式、数据源、归一化方法、方向(越高越好/越低越好)、覆盖的管辖区与时间范围。
  • 评分方法学:原始值 → 归一化(0–100)→ 按方向对齐 → 域内加权汇总 → 域得分 → 综合 HAPI。方法版本化(method_version),权重与输入可审计。
  • 政策自动评分:政府出台新政 → 映射到相关指标 → 系统按指标变化自动给出该政策领域的 HAPI 影响评估。这本身就是论文(指数设计 + 评估框架)。
  • 严谨声明:HAPI 衡量的是结果状态与趋势,把某项政策的"功劳"归因到指标变化需要 07 的因果设计,不能仅凭评分下因果结论。

1. The core idea

Governments publish metrics that suit governments. To evaluate policy independently, the observatory maintains its own index: the Healthy Aging Policy Index (HAPI) — a transparent, documented, reproducible composite that scores how well a jurisdiction supports healthy aging, and how that changes over time.

HAPI is not a single number pulled from a report. It is a methodology: defined indicators, sourced data, explicit normalization, and versioned scoring. That methodology is a publishable research contribution (Paper 1; see 09-research-roadmap.md).

2. The six domains

flowchart TB
    H["HAPI<br/>composite"]
    H --> D1["Health"]
    H --> D2["Independence"]
    H --> D3["Social Participation"]
    H --> D4["Financial Security"]
    H --> D5["Care Access"]
    H --> D6["Digital Inclusion"]
    D1 --> i1["sub-indicators…"]
    D5 --> i5["home-care hours · LTC beds ·<br/>avoidable ED visits · wait times…"]
Loading
Domain What it captures Illustrative sub-indicators
Health Health status of older adults Healthy life expectancy at 65; chronic-disease prevalence; avoidable ED visits (65+); self-rated health
Independence Ability to live independently Activities-of-daily-living limitation rate; aging-in-place rate; functional disability
Social Participation Engagement & connection Volunteer/community participation; social isolation/loneliness; transport access
Financial Security Economic stability in old age Low-income rate (65+); pension/GIS coverage; out-of-pocket health spend
Care Access Access to needed care Home-care hours per capita; LTC beds & wait times; home-care wait times; unmet care needs
Digital Inclusion Access to the digital world Internet access (65+); digital-service use; digital-literacy support

Care Access is prioritized in v1 because it maps most directly to the author's long-term-care work and to high-signal CIHI data.

Each sub-indicator is an Indicator row (see 03-data-model.md §2.4) and must declare: definition, formula, data source(s), normalization method, direction, and coverage. No indicator enters HAPI without all six.

3. Scoring methodology

flowchart LR
    O["Observations<br/>(raw values)"] --> N["Normalize to 0–100<br/>per indicator method"]
    N --> DIR["Align by direction<br/>(invert 'lower_is_better')"]
    DIR --> W["Weighted aggregate<br/>within each domain"]
    W --> DS["Domain scores (0–100)"]
    DS --> C["Composite HAPI<br/>(weighted across domains)"]
    C --> STORE[("HapiScore<br/>method_version + inputs")]
Loading
  1. Normalize. Each raw value → 0–100 via the indicator's declared method (e.g. min-max against a reference range, or z-score rescaled). Parameters stored in Indicator.normalization.
  2. Align direction. "Lower is better" indicators (e.g. avoidable ED visits) are inverted so higher always means healthier-aging.
  3. Aggregate within domain. Weighted mean of an domain's normalized indicators → domain score.
  4. Composite. Weighted mean across domains → overall HAPI, using theory-anchored "expert" tiers (Tier 1: Health, Care Access; Tier 2: Financial Security, Independence; Tier 3: Social Participation, Digital Inclusion — grounded in the WHO healthy-ageing and HelpAge AgeWatch frameworks), renormalized over the domains present for a jurisdiction × year. The choice is sensitivity-tested: hapi weights reports the composite under equal, expert, and empirical (coefficient-of-variation) schemes side by side, per the OECD/JRC Handbook on Constructing Composite Indicators. Documented in weighting.py; adjustable in a future method version.
  5. Persist with provenance. Each HapiScore records method_version and the inputs (indicator codes + weights), so any score is fully auditable and reproducible (see 03-data-model.md §2.8).

Versioned methodology. Weights and normalization can evolve; method_version ensures past scores remain interpretable and past papers remain reproducible.

4. Automatic policy scoring

The payoff: when a government announces a new policy, the system can position it against the index automatically.

flowchart LR
    P["New policy<br/>(Policy Library)"] --> MAP["Map to indicators<br/>(policy_indicator)"]
    MAP --> BASE["Baseline indicator trend<br/>(pre-policy)"]
    BASE --> WATCH["Track post-policy trend<br/>in mapped indicators"]
    WATCH --> SCORE["HAPI-domain movement<br/>attributable region"]
    SCORE --> FLAG["Report: which domains the<br/>policy touches + how they moved"]
Loading

This produces, for any policy, a structured view of which HAPI domains it targets and how those indicators have moved since it took effect.

5. Rigor: what HAPI does and does not claim

  • HAPI measures outcome states and trends for a jurisdiction over time. That is a descriptive, reproducible measurement.
  • HAPI does not, by itself, prove a policy caused a change. Attributing indicator movement to a specific policy requires the quasi-experimental designs in 07-module-policy-analytics.md (interrupted time series, difference-in-differences, synthetic control), each with stated assumptions.
  • The automatic-scoring view above is explicitly framed as "policy targets these domains; here is how they moved" — an evidence-gathering step, not a causal verdict.

Holding this line is what makes HAPI academically credible rather than a vanity score.

6. v1 scope

  • Define the six domains and a first, defensible set of sub-indicators, weighted toward Care Access and Health.
  • Full indicator definitions (all six required attributes) for the v1 set.
  • Working normalization + scoring producing NS + Federal HAPI domain scores over time.
  • Documented method_version v1.

Out of v1: an exhaustive indicator set for all six domains. v1 establishes the method on a focused indicator set; breadth is added incrementally without changing the model.

7. Visualization (web)

/hapi (and the homepage) render a domain-profile radar (DomainRadarOverTime): one polygon per jurisdiction across the scored domains, 0–100. A year slider + ▶ play scrubs the profile over time using last-observation-carried-forward, so it fills in smoothly as each domain's indicators come online rather than blinking on irregular cadences. Per-domain TrendCharts sit below, optionally overlaid with policy-event markers for the policies targeting that domain. Every score stays auditable to its raw inputs in an expandable table. See RUNBOOK §F.