Evidence level: B
导言
P3-EXP-048 把中医辨证跨文化版本外部漂移分解为 α(设备/信号漂移)、β(诊断标准/标签漂移)、γ(病机概念漂移)三轴,本研究在 P3-EXP-040(设备校正链)与 P3-EXP-036(互转公式)之间架设一行级连接层:在 OMOP CONCEPT_RELATIONSHIP 表的「Maps to」「Is a」「Mapped from」三类关系上,把 P3-EXP-036 的”标准版本→互转公式”以 custom CONCEPT_RELATIONSHIP 的形式落库,从而对 β 轴漂移做行级可追溯归因。
正文
1. 三轴漂移的 OMOP 落点设计
α 轴漂移(设备信号)已在 P3-EXP-040 通过 MEASUREMENT 表 + ComBat 仿射校正处理,β 轴漂移发生在 OBSERVATION.value_as_concept_id 与 CONCEPT_RELATIONSHIP 的语义边迁移中。当 TCM 四版辨证(1993/2002/2017/2023)映射到 OMOP 概念时,同一证候条目在不同 OMOP 词表版本(v5.0 / 当前季度版)下会指向不同 concept_id,但 CONCEPT_RELATIONSHIP 表本身已记录「Maps to」与「Is a」两类边,可以反向构造”四版→统一 concept_id”的归一化链路。这是 β 轴归因的天然锚点。
2. 「Maps to」与「Mapped from」边作为互转公式载体
OMOP 标准约定 relationship_id = 44818821(「Maps to」)是”source-to-standard concept”的方向,relationship_id = 44818823(「Mapped from」)是反向记录。在四版辨证 OMOP 概念原子化映射(P3-EXP-050 阶段 1-2)完成后,可在 CONCEPT_RELATIONSHIP 表里追加 custom 关系(如 relationship_id = 99990001「TCM version 1993/2002/2017/2023 alignment」),并把 P3-EXP-036 互转公式的每个权重参数以 concept_id → concept_id 的边形式落库。这样 β 轴漂移归因转化为「在行级 OMOP 数据上,跨 OMOP 词表版本切换时 CONCEPT_RELATIONSHIP 边集是否保持一致」的可计算检测任务。
3. 行级 β 轴漂移的检测协议
阶段 1(仿真):在 P3-EXP-049 已部署的 1,000 例心气虚 OMOP 行级数据上,注入四版(1993/2002/2017/2023)模拟扰动(增条目/改权重/移切点),观察同一患者的 CONCEPT_RELATIONSHIP 边集从 v5.0 → v5.X 的哈希稳定性。哈希漂移 > 0.20 即触发 F1 证伪(互转公式失效),需重训 P3-EXP-036 互转映射。
阶段 2(多中心盲法):在 P3-EXP-049 阶段 2 部署的 10,000 行 OMOP 数据上,跨中日韩三国同时切换 v5.0/v5.X 词表版本,记录 CONCEPT_RELATIONSHIP 边集漂移率与 P3-EXP-048 α+β+γ 三轴分解系数 α̂、β̂、γ̂。
阶段 3(反事实干预):若 β̂ > 0.5(β 轴漂移主导),触发 F2 证伪(互转公式不充分),进入”第四轴”识别协议(P3-EXP-150 预留):可能引入文化语言漂移、监管政策漂移或证候病机哲学差异。
4. 与 P3-EXP-040 / 036 / 049 / 050 的方法学衔接
P3-EXP-040 提供 α 轴校正(设备信号),P3-EXP-036 提供 β 轴互转公式草案,P3-EXP-049 提供三国部署基线数据,P3-EXP-050 提供四版→concept_id 原子化映射。本研究在四者之间架设 CONCEPT_RELATIONSHIP 嵌入层,使三轴漂移归因从”摘要级”提升到”行级可追溯”。
证据与局限
支持证据:
– OMOP CONCEPT_RELATIONSHIP 是 OHDSI 标准版词表 v5.0 核心表,三类核心边(Maps to / Is a / Mapped from)在 TKM OMOP 实证(PMID 41840573)已被用于跨源词表对齐
– ICD-11 TM 章 26 条目在 SNOMED CT 仅 1/260 覆盖率(PMID 42793550)确认了 OMOP CONCEPT_RELATIONSHIP 在传统医学的语义缺口
– TCM 四版辨证(1993/2002/2017/2023)跨年代 Meta 分析已证标准漂移(PMID 14727501, 19134453),与本研究 β 轴假设一致
– OMOP 词表版本可复现性已有方法学先例(PLOS ONE 2022 ETL 框架),支持 custom relationship_id 嵌入
局限:
– PubMed 直接命中「OMOP CONCEPT_RELATIONSHIP + β 轴 + TCM 互转公式」=0;本研究为方法学提议
– custom relationship_id 嵌入需 OHDSI 词表工作组审批,无传统医学先例
– CONCEPT_RELATIONSHIP 哈希稳定性阈值(0.20)为草案参数,依赖 P3-EXP-048 阶段 1 仿真校准
– OMOP 行级数据在三国部署差距(PMID 41840573 仅 TKM 端实证)限制阶段 2-3 的样本多样性
– 不支持儿童/孕产/重症;不替代 P3-EXP-036/040/041 的独立研究价值
相关阅读
- P3-EXP-036 互转映射公式草案
- P3-EXP-040 ComBat 仿射校正(设备 α 轴)
- P3-EXP-048 三轴漂移归因(DDC 框架)
- P3-EXP-049 OMOP CDM 三国部署基线
- P3-EXP-050 TCM 四版辨证→concept_id 原子化映射
-
147 三轴归因 dashboard(Shiny 横向工具)
免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。
OMOP Concept Relationship Embedding Protocol for Standard Version to Inverse Formula in Beta Axis Drift Attribution
Evidence level: B
Introduction
P3-EXP-048 decomposes cross-cultural external drifts of TCM syndrome differentiation into three axes: α (device/signal drift), β (diagnostic standard/label drift), and γ (pathomechanism concept drift). This study establishes a row-level connection layer between P3-EXP-040 (device correction chain) and P3-EXP-036 (interconversion formula). Specifically, it embeds the “Standard Version → Interconversion Formula” from P3-EXP-036 into the OMOP CONCEPT_RELATIONSHIP table as custom CONCEPT_RELATIONSHIP entries across the three relationship types: “Maps to”, “Is a”, and “Mapped from”. This enables row-level traceable attribution of β-axis drift.
Main Body
1. OMOP Landing Design for Three-Axis Drifts
α-axis drift (device signals) has already been processed via the MEASUREMENT table plus ComBat affine correction in P3-EXP-040. β-axis drift occurs during semantic boundary migration between OBSERVATION.value_as_concept_id and CONCEPT_RELATIONSHIP. When mapping four versions of TCM syndrome differentiation (1993/2002/2017/2023) to OMOP concepts, the same syndrome entry points to different concept_ids under different OMOP ontology versions (v5.0 / current quarterly version). However, the CONCEPT_RELATIONSHIP table itself records “Maps to” and “Is a” edges, allowing the construction of a normalization link from “Four Versions → Unified concept_id”. This serves as a natural anchor for β-axis attribution.
2. “Maps to” and “Mapped from” Edges as Carriers for Interconversion Formulas
OMOP standards define relationship_id = 44818821 (“Maps to”) as the direction from “source-to-standard concept”, while relationship_id = 44818823 (“Mapped from”) records the reverse. After completing the atomic mapping of four-version TCM syndrome differentiation OMOP concepts (P3-EXP-050, stages 1-2), custom relationships can be appended to the CONCEPT_RELATIONSHIP table (e.g., relationship_id = 99990001 “TCM version 1993/2002/2017/2023 alignment”). Each weight parameter of the P3-EXP-036 interconversion formula can then be stored as edges in the format concept_id → concept_id. This transforms β-axis drift attribution into a computable detection task: “whether the set of CONCEPT_RELATIONSHIP edges remains consistent when switching across OMOP ontology versions at the row level in OMOP data.”
3. Row-Level β-Axis Drift Detection Protocol
Stage 1 (Simulation): On the 1,000 row-level OMOP datasets of Heart Qi Deficiency already deployed in P3-EXP-049, inject simulated perturbations from four versions (1993/2002/2017/2023) by adding entries, modifying weights, or shifting cut-points. Observe the hash stability of the patient’s CONCEPT_RELATIONSHIP edge set from v5.0 → v5.X. A hash drift > 0.20 triggers F1 falsification (interconversion formula failure), requiring retraining of the P3-EXP-036 interconversion mapping.
Stage 2 (Multi-center Blinded): On the 10,000 row-level OMOP datasets deployed in Stage 2 of P3-EXP-049, simultaneously switch between v5.0 and v5.X ontology versions across China, Japan, and Korea. Record the drift rate of CONCEPT_RELATIONSHIP edge sets alongside the P3-EXP-048 three-axis decomposition coefficients $\hat{\alpha}$, $\hat{\beta}$, and $\hat{\gamma}$.
Stage 3 (Counterfactual Intervention): If $\hat{\beta} > 0.5$ (β-axis drift dominates), trigger F2 falsification (interconversion formula insufficient) and enter the “Fourth Axis” identification protocol (reserved in P3-EXP-150): potentially introducing cultural linguistic drift, regulatory policy drift, or philosophical differences in syndrome pathomechanism.
4. Methodological Integration with P3-EXP-040 / 036 / 049 / 050
P3-EXP-040 provides α-axis correction (device signals); P3-EXP-036 provides the β-axis interconversion formula draft; P3-EXP-049 provides baseline data from three countries; P3-EXP-050 provides the atomic mapping from Four Versions → concept_id. This study establishes a CONCEPT_RELATIONSHIP embedding layer among these four, elevating three-axis drift attribution from “abstract level” to “row-level traceability.”
Evidence and Limitations
Supporting Evidence:
– The OMOP CONCEPT_RELATIONSHIP table is a core table in the OHDSI standard ontology v5.0; the three core edges (Maps to / Is a / Mapped from) have been used for cross-source ontology alignment in TKM OMOP empirical studies (PMID 41840573).
– Only 1/260 coverage of ICD-11 TM Chapter 26 entries in SNOMED CT (PMID 42793550) confirms semantic gaps in OMOP CONCEPT_RELATIONSHIP for traditional medicine.
– Cross-decade meta-analyses of TCM four-version syndrome differentiation (1993/2002/2017/2023) have confirmed standard drift (PMID 14727501, 19134453), consistent with this study’s β-axis hypothesis.
– Reproducibility of OMOP ontology versions already has methodological precedents (PLOS ONE 2022 ETL framework), supporting custom relationship_id embedding.
Limitations:
– Direct PubMed hits for “OMOP CONCEPT_RELATIONSHIP + β-axis + TCM interconversion formula” = 0; this study is a methodological proposal.
– Custom relationship_id embedding requires OHDSI ontology working group approval; no precedents exist in traditional medicine.
– The hash stability threshold for CONCEPT_RELATIONSHIP (0.20) is a draft parameter dependent on calibration from P3-EXP-048 Stage 1 simulation.
– Row-level OMOP data disparities across three countries (PMID 41840573 only validated on TKM side) limit sample diversity in Stages 2-3.
– Does not support pediatric/pregnancy/critical care populations; does not replace independent research value of P3-EXP-036/040/041.
Related Reading
- P3-EXP-036 Interconversion Mapping Formula Draft
- P3-EXP-040 ComBat Affine Correction (Device α-axis)
- P3-EXP-048 Three-Axis Drift Attribution (DDC Framework)
- P3-EXP-049 OMOP CDM Three-Country Deployment Baseline
- P3-EXP-050 TCM Four-Version Syndrome Differentiation → concept_id Atomic Mapping
-
147 Three-Axis Attribution Dashboard (Shiny Horizontal Tool)
Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.
Leave a Reply