信息流先改，责任流后动 · AI 与医疗健康产业 · 2026-07Information first, responsibility last · AI & Healthcare · Jul 2026

AI 会深刻改写医疗健康产业——但它先重写的是信息流、工作流、支付流，而不是责任流、身体接触与人际信任 AI will deeply rewrite the healthcare industry — but it rewrites the flows of information, work and payment first, not the flows of responsibility, physical touch and human trust

决定你健康的因素里，医疗只占 8–10%，生活方式占 40–60%——睡眠、运动、饮食、不吸烟胜过绝大多数晚期医疗干预。但资本与 AI 仍不可避免地涌向那 8%，因为只有医疗节点有清晰的支付路径。Of what decides your health, medicine is only 8–10%; lifestyle is 40–60% — sleep, exercise, diet and not smoking beat most late-stage interventions. Yet capital and AI still pour into that 8%, because only the medical node has a clear path to payment.

AI 已通过主任医师笔试（夸克 67.7%），却无法坐牢、无法被吊销行医执照——只要签字责任不可转移，AI 在医疗体系里永远是「辅助驾驶 Copilot」，成不了「自动驾驶 Auto-pilot」。AI has passed the chief-physician written exam (Quark 67.7%), yet it cannot go to jail or lose its medical license — as long as the signature can't transfer, AI in medicine is forever a 'Copilot,' never an 'Auto-pilot.'

主脊是健康生命周期九节点（预防→体检→问诊分诊→诊断→治疗→康复→慢病→养老→安宁），每节点标注传统 vs AI + 落地强弱；叠加支付方轨（DRG/集采控费）与替代医学 / 安慰感经济影子轨两条平行轨。另含不对称地图（已重写 vs 攻不下）、六块硬骨头。这是一张批判性行业解剖，不是诊疗建议。相邻议题见姊妹图：生活方式杠杆→growth、医美颜值→beauty、心理疗愈→mind。 The spine is the nine-node health lifecycle (prevention → screening → triage → diagnosis → treatment → rehab → chronic care → elder care → hospice), each tagged traditional vs AI + landing strength; overlaid with the payer rail (DRG / bulk-buy cost control) and the alternative-medicine / placebo-economy shadow rail. Plus the asymmetric map (rewritten vs uncrackable) and six hard bones. A critical industry dissection, not medical advice. Adjacent topics on siblings: the lifestyle lever → growth, cosmetic medicine → beauty, mental healing → mind.

传统医疗节点Traditional node

AI 诊疗 · 已重写AI care · rewritten

责任流 · 签字权 · 硬骨头Responsibility · uncrackable

支付方 · 控费Payer · cost control

替代医学 · 安慰感Alt-medicine · placebo

8–10%

医疗对健康结果的贡献（WHO 8% / McGinnis 10%）；生活方式占 40–60%。CKB 研究：五项健康生活方式者全因死亡风险降约 68%（HR=0.32）Medicine's contribution to health outcomes (WHO 8% / McGinnis 10%); lifestyle is 40–60%. The CKB study: five healthy habits cut all-cause mortality ~68% (HR=0.32)

67.7%

夸克健康大模型 2025-07 首个通过中国 12 门核心学科主任医师笔试（人类约 60%、DeepSeek-R1 38.7%、GPT-o3-mini 35.5%）。⚠️考试≠真实临床Quark's health LLM (Jul 2025) was the first to pass China's 12-subject chief-physician written exam (humans ~60%, DeepSeek-R1 38.7%, GPT-o3-mini 35.5%). ⚠️Exams ≠ real clinics

3.0万亿

2025 基本医保基金支出（30009 亿元）——谁付钱决定一切。DRG/DIP 打包付费 2024 底全覆盖、2025 切 2.0，医院从「创收中心」变「成本中心」2025 basic-insurance spending (¥3.0T) — the payer decides everything. DRG/DIP bundled payment reached full coverage by end-2024, moved to v2.0 in 2025; hospitals shift from 'profit center' to 'cost center'

500万

养老护理员缺口（失能老人约 3500 万、持证护理员仅约 50 万）——养老照护是「需求最刚性、AI 最无能为力」的最大错位The elder-care worker gap (~35M disabled seniors, only ~500k certified) — elder care is the biggest mismatch: 'most rigid demand, where AI is most powerless'

口径警告：本页是批判性行业分析，非诊疗 / 用药 / 投资建议。三份研究文档聚焦中国医疗健康产业。健康决定因素两套口径须并陈：WHO（医疗 8%）与 McGinnis（医疗 10%）框架不同（McGinnis「10%」原指「医疗缺陷对早死的贡献」），不择一；CKB / 哈佛队列为关联非严格因果。医保基金 2024（3.49 万亿）与 2025（35873.11 亿元）是不同年份口径非矛盾。过度诊断比例分癌种不可混用（甲状腺 4–36% / 肺癌女性 85–89% / 前列腺 PSA 20.7–50.4%）。凡「降本 / 提效 / 分诊准确率」类厂商自述打 D 级，已逐条标注；市场规模含预测（AI 中医 1001 亿、生成式 AI 医疗 509 亿美元为单一机构预测）。每张卡片右上角 A/B/C/D＝证据强度。 Basis warning: a critical industry analysis, not medical / prescribing / investment advice. The three reports focus on China's healthcare industry. Two frameworks for health determinants are shown together: WHO (medicine 8%) and McGinnis (medicine 10%) differ (McGinnis's '10%' originally meant 'shortfalls in care as a contributor to early death'); neither is picked. CKB / Harvard cohorts are association, not strict causation. Insurance-fund figures for 2024 (¥3.49T) and 2025 (¥3.587T) are different years, not a contradiction. Overdiagnosis rates can't be mixed across cancers (thyroid 4–36% / female lung 85–89% / prostate PSA 20.7–50.4%). 'Cost cut / efficiency / triage accuracy' claims are grade D (vendor), flagged individually; market sizes include forecasts (AI-TCM ¥100B, generative-AI healthcare $50.9B are single-firm forecasts). Each card's top-right A/B/C/D = evidence strength.

◆ 诚实层 · 地基（最重）The honesty layer · foundation

医疗，只是健康的最后 8–10%Medicine is only the last 8–10% of health

看懂整张图前，先接受一个反直觉的坐标系：决定健康结果的因素里，医疗只占小头。两套权威框架数字必须并陈——但结论一致：生活方式才是大杠杆。这解释了全图最深的反讽：「最赚钱的环节」（治疗、药械）未必是「对健康贡献最大的环节」（预防、生活方式），而后者恰恰最难商业化、最难被 AI 替代。Before reading the map, accept a counter-intuitive frame: medicine is only a small share of what decides health. Two authoritative frameworks must be shown together — but they agree: lifestyle is the big lever. Hence the map's deepest irony: 'the most profitable node' (treatment, drugs, devices) is not 'the node that contributes most to health' (prevention, lifestyle) — and the latter is exactly the hardest to commercialize and to replace with AI.

McGinnis 框架The McGinnis frameworkHealth Affairs 2002 · B

行为 40%Behavior 40% 遗传 30%Genes 30% 社会 15%Social 15% 医疗 10%Care 10% 环境 5%Env 5%

WHO 口径The WHO framingB / C

生活方式 60%Lifestyle 60% 环境 17%Env 17% 遗传 15%Genes 15% 医疗 8%Care 8%

杠杆排序：睡眠 / 运动 / 饮食 / 不吸烟＞绝大多数晚期医疗干预。CKB（中国 51 万人队列，中位随访 10.2 年）：五项健康生活方式全占者全因死亡风险 HR=0.32＝降约 68%；哈佛两大队列：采纳五项低风险生活方式者 50 岁时预期寿命女性延长 14.0 年、男性延长 12.2 年。⚠️ 队列研究为关联非严格因果，McGinnis 比例为专家综合估计。Leverage order: sleep / exercise / diet / not smoking ＞ most late-stage interventions. CKB (510k Chinese, 10.2-yr median follow-up): all five healthy habits → all-cause mortality HR=0.32 = ~68% lower; Harvard cohorts: five low-risk behaviors → life expectancy at 50 +14.0 yrs for women, +12.2 for men. ⚠️ Cohort studies are association, not strict causation; McGinnis shares are expert estimates.

▸ 张力二 · 医疗 vs 生活方式Tension 2 · care vs lifestyle

为何钱与 AI 仍涌向那 8%？Why do money & AI still pour into that 8%?

因为生活方式极难被标准化、规模化地定价（「如何让人坚持睡够、动起来」没有清晰支付路径），而医疗节点——疾病诊断、高值耗材、靶向药——有清晰的支付路径（医保 / 商保买单）。杠杆在 growth，支付在这里。Because lifestyle is extremely hard to standardize and price at scale ('how to make people sleep and move' has no clear payment path), while medical nodes — diagnosis, high-value implants, targeted drugs — do have a clear payment path (insurance foots the bill). The lever is in growth; the payment is here.

▸ 张力一 · 循证 vs 安慰感Tension 1 · evidence vs placebo

体系内外，买的是同一种「确定感」Inside and outside, buying the same 'certainty'

体系内的过度医疗（无效抢救、大处方大检查）与体系外的保健品 / 中医 / 身心灵，共享同一心理机制：买的不是可证伪的疗效，而是「被照顾的感觉 + 对不确定性的掌控感」。研究：ChatGPT 答复被评「有同理心」的比例是医生的近 10 倍。循证轨里 AI 做效率工具，影子轨里 AI 做需求放大器。In-system overtreatment (futile rescue, over-prescription) and out-of-system supplements / TCM / wellness share one psychology: buying not falsifiable efficacy, but 'the feeling of being cared for + control over uncertainty.' A study rated ChatGPT's replies 'empathetic' ~10× more often than physicians'. On the evidence rail AI is an efficiency tool; on the shadow rail it is a demand amplifier.

从这张图看到的五条规律Five patterns this map makes visible