起步顺序Order of work 先给每个模型写清目标与代理指标,再按患者群体核对,最后才扩展到全院患者流。Name target and proxy for each model first, then check by patient group, then widen to hospital flow.
公开事实Public facts 已核对Checked
- Mar 2025SKAI 启动,SickKids 2030 战略旗舰倡议SKAI launched, a flagship of its 2030 strategy含服务、研究、社区三部分Service, Discovery and community arms
- 12.2%医生首次评估时间改善Better time to first physician assessment同比;2026 年质量改进计划Year on year, 2026 quality improvement plan
- 5,1812024 年手术等候名单人数People on the surgical waiting list, 2024从 5,776 降下来,降幅约 10%Down from 5,776, about 10%
- 9,000+2025 年全院员工Staff across the hospital, 2025病床约 350 张;口径因来源而异About 350 beds; counts vary by source
扩大之前的三道核对Three checks before widening 判断Judgment
- 目标Target:每个孩子的结局更好、照护安全及时Better outcomes and safe, timely care for each child(写下来了吗?Is this written down?)
- 代理指标Proxy:模型实际度量的:等待时间、提前检查What the model measures: wait time, early tests(它与目标差多远?How far is it from the target?)
- 人群Groups:按年龄、语言与病情复杂度分开看结果Results split by age, language and complexity of illness(对所有人都有效吗?Does it work for all?)
专家组裁决Expert panel 依据公开观点的推演Inference from public views
Ziad Obermeyer
每个模型写清目标与代理指标;按患者群体核对。Write target and proxy for each model; check by patient group.
Ethan Mollick
与临床人员一起,测出分诊 AI 在真实病例上的失败点。Test where triage AI fails on real cases, with clinicians.
Thomas Davenport
患者流是跨科室的流程;点名各环节负责人。Patient flow is a process across wards; name each owner.
Erik Brynjolfsson
给 22% 的预期设基线;培训与流程再造也要计入。Baseline the 22% claim; count training and redesign too.
Andrew Ng
选一个流程(如急诊等待),数月内出数字。One flow, such as emergency waits, with numbers in months.
Foteini Agrafioti
SKAI 服务、临床与运营的权责写成一页纸。One page of ownership: SKAI Service, clinics and operations.
Andrew Ng:尽快从急诊扩大。Widen from emergency soon.
Ziad Obermeyer:先核对目标与人群。Check target and groups first.
→ 预测走快速通道;临床走完整审查。Fast lane for forecasts; full review for clinical.
六位专家是谁、怎样思考、有什么铁律 →Who the six experts are, how they think, their rules →
三个目标Three priorities 建议Recommendation
核对每个模型的目标Check every model's target
SKAI 服务下,每个模型都写明目标、代理指标与分群结果。Target, proxy and group results for each model under SKAI Service.
预测患者流Forecast patient flow
从急诊等待扩展到住院、手术与出院。Extend from emergency waits to admissions, surgery and discharge.
输出治理模式Share the governance model
政策、服务与评估,可输出给其他儿童医院。Policy, service and evaluation, offered to other children's hospitals.
场景组合Scenarios 建议Recommendation
| 场景Scenario | 做法Approach | 状态Status |
|---|---|---|
| 全院患者流预测Whole-hospital patient flow 提效Efficiency体验Experience | 逐个流程预测住院、手术、出院的容量与人力。Forecast admissions, surgery and discharge capacity and staffing, one flow at a time. | 建议 · 基于急诊工具Proposed · builds on emergency tools |
| 按人群核对的急诊工具Emergency tools, checked by group 降风险Risk体验Experience | 按患者群体复核分诊检查工具与等待时间预测。Review the triage test tool and wait predictions by patient group. | 建议 · 临床治理Proposed · clinical governance |
| 面向家属的助手Family-facing assistant 体验Experience降风险Risk | 多语言问答与就诊准备;安全规则与临床治理确认。Multilingual Q&A and visit preparation; confirm safety rules with clinical governance. | 建议Proposed |
| 向外输出 SKAI 治理SKAI governance for others 增收Revenue降风险Risk | 把 AI 采用政策与评估步骤提供给其他儿童医院。Offer the adoption policy and evaluation steps to other children's hospitals. | 建议Proposed |
12 个月路线12-month roadmap 建议Recommendation
奠基Foundation
- 盘点所有模型Inventory every model
- 为每个写明目标与代理Target and proxy for each
- 分群结果基线Group results baseline
试点Pilot
- 一个流程预测One flow forecast
- 临床人员测边界Clinicians test the edges
- 基线与对照Baseline and control
上线Go live
- 流程预测上线Flow forecast live
- 上线即监控Monitoring from day one
- 员工培训Staff training
扩展Extend
- 第二个流程立项Second flow scoped
- 分享方法Share the method
- 重排下年优先级Re-rank next year
治理与合规Governance 建议Recommendation
- 所有 AI 先经 SKAI 服务评估All A I goes through SKAI Service review first
- 患者数据:规则与隐私顾问确认Patient data: confirm rules with privacy counsel
- 影响诊疗的输出由临床人员确认A clinician confirms any care-affecting output
三层指标Three-layer metrics 建议Recommendation
| 采用Adoption | 经 SKAI 服务评估的模型数Models reviewed under SKAI Service · 每周使用预测的临床人员数Clinicians using forecasts weekly |
|---|---|
| 过程Process | 预测准确度对照基线Forecast accuracy against baseline · 临床人员改判率Clinician override rate · 首次评估时间Time to first assessment |
| 结果Outcome | 按患者群体的等待与结局Waits and outcomes by patient group · 等候名单长度对照基线Waiting-list length against baseline |
目标值由企业自定。Targets are the company's to set.
风险Risks 判断Judgment
更快,但对谁更快Fast, but for whom
工具可能提高平均速度,却让某些群体落后。A tool can speed up the average while leaving some groups behind.
证据之前先有承诺Claims before proof
类似 22% 的预期收益,引用前需要基线。Projected gains, such as twenty-two percent, need a baseline before they are quoted.
预警疲劳Alert fatigue
预警太多会让员工习惯性忽略;与临床人员共同测试。Too many warnings teach staff to ignore them; test with clinicians.
前 90 天First 90 days 建议Recommendation
| 第 1–2 周Weeks 1–2 | 为每个模型点名目标负责人Name a target owner for each model |
|---|---|
| 第 3–8 周Weeks 3–8 | 为每个模型写明目标与代理Write target and proxy for each model |
| 第 9–12 周Weeks 9–12 | 定分群基线;选一个流程Set group baselines; choose one flow |
| 第 13 周Week 13 | 首次复盘First review |
我们可能错在:We could be wrong if: 基线显示急诊工具对各群体本来就同样有效,或流程预测带来的增益很小。the baselines show emergency tools already work equally well for every group, or flow forecasts add little.
出处Sources
本页是基于公开资料的分析,与所述企业没有合作关系,不代表其官方立场或内部计划;企业数字多来自其自述或媒体转述;「专家组」是依据六位专家公开观点的推演,不是他们的原话,也不代表他们或其雇主的立场。资料整理日期:2026-11-14。This page is an analysis based on public information. It has no affiliation with the company and does not represent its official position or internal plans. Company figures come mostly from its own statements or media reports. The expert panel is an inference from six experts' public views, not their words, and does not represent them or their employers. Compiled 2026-11-14.
想给你的公司也做一份?Want one for your company?
同样的方法:公开事实、专属判断、12 个月路线和前 90 天。公司大小都可以聊,中文沟通。Same method: public facts, a company-specific judgment, a 12-month roadmap and the first 90 days. Any size of business; we can talk in Chinese or English.