摘要：David Shaywitz，在健康、科技等领域有深入见解，其文章常发表于专业媒体，对新兴的健康科技趋势如个人健康操作系统等有独特的分析，从数据、理念等多方面探讨健康领域的发展，为读者提供了关于健康与科技融合的新视角。
江晓彤：Health Deserves A Vision More Capacious Than Dashboard Metrics
（健康理应拥有比仪表盘指标更具广度的愿景）
大卫・谢维茨（David Shaywitz）
David Shaywitz，在健康、科技等领域有深入见解，其文章常发表于专业媒体，对新兴的健康科技趋势如个人健康操作系统等有独特的分析，从数据、理念等多方面探讨健康领域的发展，为读者提供了关于健康与科技融合的新视角。
2025 年 5 月 9 日
消费者健康与保健领域正经历着一系列的活跃发展。
实验室检测公司 Function（口号：“是时候掌控你的健康了”）收购了全身核磁共振成像公司 Ezra，Ezra 承诺提供 “世界上最先进的长寿扫描”。
热门智能戒指的制造商 Oura 最近增加了持续血糖测量功能，以及通过照片计算膳食营养的能力。Oura 还聘请了瑞奇・布卢姆菲尔德博士（Dr. Ricky Bloomfield）担任其首位首席医疗官；布卢姆菲尔德博士此前曾在苹果公司担任临床和健康信息学负责人，并以其在健康数据互操作性方面的专业知识而闻名。
与此同时，Oura 的竞争对手、智能手环制造商 Whoop 刚刚宣布推出其设备的最新版本，具备监测血压、心电图的能力，并能够评估其所谓的生物年龄指标，他们称之为 “Whoop 年龄”。Whoop 现在表示，它旨在 “解锁人类的表现和健康寿命”，用 “全面了解你的健康状况” 这一宣传语吸引用户。
迈向个人健康操作系统（OS）
注意到其中的规律了吗？
正如行业通讯《Fitt Insider》（FI）最近所观察到的，将这些方法以及许多其他方法联系在一起的是，它们反映了一种生成 “个人健康操作系统” 的尝试，旨在 “赋予个人对自身健康的掌控权”，更广泛地说，是从一个常被（尤其是年轻人）认为要么毫无用处、要么具有阻碍性的医疗系统手中夺回控制权。
FI 引用了最近的埃德尔曼调查，报告称：
…… 近一半的年轻人认为消息灵通的人可以和医生一样知识渊博，三分之二的人将生活经验视为专业知识，61% 的人将机构视为获得医疗服务的障碍。
受够了被动式医疗，许多人已经在可穿戴设备、生活方式应用程序、直接面向消费者的诊断工具等方面收集数据，但大多数数据都是孤立的。总的来说，Function 正在构建一个统一的平台，能够从原始输入中生成具有临床相关性的见解。
FI 指出，像 Bright OS、Gyroscope 和 Guava Health 这样专注于 “日常数据管理” 的公司大量涌现，还有像 Superpower（“提供礼宾级指标，无需初级保健医生”）和 Mito Health（一个 “袖珍人工智能医生”，“通过合并实验室数据、医疗记录、家族病史、生活方式信息等生成全面的数字健康档案”）这样的初创公司。
人工智能似乎在许多这类公司中有望发挥越来越核心的作用。
FI 推测：
更进一步，端到端的大语言模型可以闭环，将因果联系起来，将见解转化为行动，与初级保健医生同步，并为人工智能驱动的医疗未来奠定基础。
这是一个需要深呼吸的时刻 —— 同时也需要更仔细、更批判性地审视这种以消费者为中心、以数据为支撑的健康愿景。
一个强大的愿景
毫无疑问，这里有很多值得接受的东西，尤其包括：
个人有机会从更多样化的来源收集更丰富的健康数据，特别是可穿戴设备；
从这些数据中获得相关见解的可能性增加（这是早期 “量化自我” 努力的一个关键缺陷）；
围绕个人集中管理健康数据（Superpower 的标语是 “健康数据，集中一处”），这是一个长期以来被承诺但在实践中常常令人沮丧地难以实现的医疗保健目标。如今，仍然有许多患者发现自己不得不乞求才能有效地获取自己的健康信息，而医疗系统往往将这些数据视为竞争优势，并不愿意放手。
一种由科技赋能的健康方法，即你拥有更多关于自己的数据，这些数据明确处于你的控制之下，并可能导致更健康的行为，这代表了一种值得庆祝的进步。
与此同时，当我审视许多这些健康方法时，我看到了两大类担忧。
担忧一：脆弱数据的堆砌可能无法带来深刻见解
第一个，也许是更具体的担忧是，套用喜剧演员丹尼斯・米勒（Dennis Miller）的话，“一堆垃圾加一堆垃圾还是垃圾”，仅仅收集大量的数据（其中许多数据可能是不可靠的），即使热切地调用人工智能的神奇力量，也不一定能转化为卓越的见解。
埃里克・托普博士（Dr. Eric Topol）在一篇特别深刻的 “基本事实” 博客文章中，专注于 “促进长寿和健康寿命的业务”，他写道：“在一个人身上获取数百个生物标志物结果和成像测试，大大增加了假阳性结果的可能性”，这是一个令人担忧的可能性。
我在这里讨论过假阳性的挑战，并在这里深入探讨了围绕贝叶斯定理（用于评估）的一些细节。在这个领域的原始参考文献可能是扎克・科汉（Zak Kohane）及其同事 2006 年的这篇论文，他们在论文中引入了 “偶发组学” 这个术语。
公平地说，至少一些支持广泛检测的人认识到假阳性的挑战，但他们认为随着时间推移收集个人密集数据的机会能够观察到重要的变化，彼得・阿蒂亚博士（Dr. Peter Attia）在《超越寿命》（Outlive）中明确强调了这一点；我在这里讨论了他的 “风险管理” 心态。
同样，巴克研究所（Buck Institute）的教授、Thorne 公司的首席科学官内森・普莱斯（Nathan Price）认为，（在人工智能的辅助下）仔细检查丰富的个人数据可以确定（例如）补充剂干预的机会。这些干预措施在人群层面可能没有太大作用（因此，正如托普博士在他的最新著作《超级老人》（Super Agers）中指出的那样，缺乏有说服力的补充剂临床试验数据 —— 我在《华尔街日报》上对该书的评论见这里），但在特定个体中可能会有作用。（我也在这里和这里讨论了普莱斯）
“个人健康操作系统” 的支持者也可能强调有利因素 —— 随着测量技术不断改进、更密集的数据可用以及人工智能工具功能不断增强，改善预测的可能性增加。倡导者可能会争辩说，也许我们还没有完全实现我们想象中的未来，但我们已经足够接近，可以开始看到它可能的样子。
担忧二：对健康的狭隘看法
关于我们似乎正在走向的健康模式，一个可以说是更深层次的担忧是，它在多大程度上受到一种严格的简化思维模式的影响。在这种有限的、经典的管理（或咨询）观点中，健康仅仅变成了仪表盘上的指标，一系列不断扩展的参数，必须不断地进行测量、量化和优化。
夸梅・安东尼・阿皮亚（Kwame Anthony Appiah）在《纽约时报杂志》上发表的一篇关于我们对幸福的不断演变的理解和方法的优美文章提醒我们可能会错过什么。
阿皮亚写道，在新千年伊始，我们进入了
由蒂姆・费里斯（Tim Ferriss）等优化大师主导的生活黑客、自我量化、习惯堆叠的时代，他的第一本书《每周工作 4 小时》于 2007 年出版，用他的话说，这是一个 “用于最大化每小时产出的工具包”。
阿皮亚继续说，因此，幸福的概念被分解为 “模块化升级”，因为我们在完善我们的 “个人操作系统”。
然而，正如阿皮亚所写，至关重要的是要认识到 “幸福不是一个优化问题”，而是更深刻、更实质的东西。
我在 2018 年的一篇题为《我们不是一个仪表盘》的文章中也表达了类似的观点。
我观察到 “仪表盘已经成为我们这个时代的一个有力象征”，我写道：“大数据的意识形态已经有了自己的生命力，假定了一种必然性和自我正当性。”
我继续写道：“从为了人而进行测量，我们越来越像是为了数据而进行测量，建立起这样的系统和组织，在其中不断的测量常常似乎本身就是目的。”
我想起了凯特・克劳福德（Kate Crawford）的《人工智能地图集》（Atlas of AI）（我在《华尔街日报》上对该书的评论见这里）中我最喜欢的一句话：“工具的便利性成为了真理的地平线”，在这种情况下，这提醒我们，即使我们被能够测量和分析健康数据的工具所包围，我们也必须确保我们对健康的理解超越这些工具的局限性。
当然，重点不是走向另一个极端，完全拒绝指标。
正如精彩著作《指标的暴政》（Tyranny of Metrics）的作者杰里・穆勒（Jerry Muller）教授所解释的：“我无法想象有能力的专家会忽视指标。问题在于他们评估指标重要性的能力，以及认识到未测量因素的作用。”（重点已添加）
我在 2011 年一篇题为《硅谷对医学的误解》的文章中也谈到了这一需求，我写道：“一个新颖的技术平台，如果忽视了患者的综合需求，或者低估或没有考虑到疾病实际发生和患者（以及与患者最亲近的人）所经历的复杂性和混乱性，将不可避免地存在不足。”
向前发展
为了最有效地满足患者的需求 —— 包括预防或预先防范疾病这一至关重要的目标，这样人们就不会成为患者 —— 至关重要的是要接受新兴技术的力量和前景，包括那些能够实现 “个人健康操作系统” 概念的技术，同时不要将这个概念地图误认为实际情况（正如阿尔弗雷德・科日布斯基（Alfred Korzybski）的名言所教导的那样）。
与每位患者合作确定优先事项，并确定一些关键的健康参数作为关注重点至关重要；大卫・布卢门撒尔博士（Drs. David Blumenthal）和 J. 迈克尔・麦金尼斯（J. Michael McGinnis）在 2015 年《美国医学会杂志》（JAMA）的 “观点” 栏目中深入讨论了 “核心指标” 这一话题。
与此同时，我们必须坚持一种健康与保健的愿景，这种愿景远远超出仪表盘的限制，追求超越指标递归优化的目标（正如我最近在这里讨论的那样）。我们的方法必须足够包容，真实地重视并切实培养健康、繁荣生活的其他组成部分，这可能包括智力上的吸引、对目标的追求以及与家人、朋友和社区的社交互动。
（马丁・塞利格曼（Martin Seligman）的 PERMA 模型 —— 积极情绪 / 快乐、投入 / 心流、人际关系 / 与他人的联系、意义 / 目标和成就 —— 为扩展我们的思维提供了一个潜在有用的框架 [见这里，这里]）
尽管将健康的一些最重要、最深刻的组成部分简化为一个易于理解的数字即使不是完全不可能，也是非常困难的，但我们必须继续重视并追求这些组成部分。
即使我们努力利用新兴技术来构建和完善健康仪表盘，我们也要决心朝着一个更广泛、更持久、更有意义的健康愿景努力，这个愿景超越了行、列和数字的枯燥语法。
May 09, 2025：
Consumer health and wellness is experiencing a flurry of activity.
The lab testing company Function (motto: “It’s time to own your health”) acquired Ezra, a whole body MRI company promising “the world’s most advanced longevity scan.”
Oura, maker of the popular smart ring, recently added an integration for continuous glucose measurement as well as the ability to calculate meal nutrition based on a photo. Oura also hired Dr. Ricky Bloomfield as its first Chief Medical Officer; Dr. Bloomfield had previously served as Clinical and Health Informatics Lead at Apple, and is known for his expertise in health data interoperability.
Meanwhile, Oura competitor Whoop, maker of a smart band, just announced the latest versions of its device, with the ability to monitor blood pressure, ECG, and to assess what it describes as a measure of biological age, which it calls “Whoop Age.” Whoop now says it seeks to “unlock human performance and healthspan,” enticing users with the pitch, “Get a complete picture of your health.”
Towards a Personal Health Operating System (OS)
Notice a pattern yet?
What unites these approaches and so many others, as the industry newsletter Fitt Insider (FI) recently observed, is they reflect an attempt to generate a “personal health OS,” intended to “give individuals agency over their well-being,” and more generally, wrest control back from a health system that’s often perceived (especially by young adults) as somewhere between useless and obstructive.
Citing a recent Edelman survey, FI reports,
…nearly half of young adults believe well-informed people can be as knowledgeable as doctors, two-thirds see lived experience as expertise, and 61% view institutions as barriers to care.
Fed up with reactive care, many already collect data across wearables, lifestyle apps, DTC diagnostics, and more, but most are siloed. Rolling up, Function is architecting a unified platform capable of generating clinically relevant insights from raw inputs.
FI points to the proliferation of companies like Bright OS, Gyroscope, and Guava Health focused on “day-to-day data management,” as well as startups like Superpower (“Delivering concierge-level metrics minus the PCP”) and Mito Health (a “pocket-sized AI doctor” that “generates comprehensive digital health profiles by merging labs, medical records, family history, lifestyle info, and more.”)
AI seems poised to play an increasingly central role in many of these companies.
FI speculates,
A step further, end-to-end LLMs could close the loop, linking cause and effect, turning insights into actions, syncing with PCPs, and laying the foundation for an AI-powered medical future.
This is a good time to take a deep breath – as well as a closer, more critical look at this vision of consumer-empowered, data-fortified health.
A Powerful Vision
Unquestionably, there’s a lot to embrace here, including in particular:
The opportunity for individuals to gather more and richer health data from a greater variety of sources, including in particular wearables;
The increased possibility of relevant insights (a key deficiency of early “Quantified Self” efforts) from these data.
The explicit centralization of your health data around you (Superpower’s tagline is “Health Data, In One Place”), a long-promised but often frustratingly elusive healthcare goal in practice. Today, still, (still!), so many patients find themselves having to beg and plead for efficient access to their own health information, data that health systems tend to view as a competitive advantage and aren’t eager to let go.
A tech-enabled approach to health where you have more abundant data about you, that are explicitly in your control, and which could lead to healthier behaviors represents the sort of progress that deserves to be celebrated.
At the same time, when I look at many of these approaches to health, I see two broad categories of concerns.
Concern One: Plural of Fragile Data May Not Be Insight
The first, perhaps more concrete worry, is that, to paraphrase comedian Dennis Miller, “two of [crap] is [crap],” and simply the collection of a lot of data, much of which may be fragile, isn’t sure to translate into brilliant insight, even if the magical power of AI is fervently invoked.
In an especially incisive “Ground Truths” blog post focused on “The business of promoting longevity and healthspan,” Dr. Eric Topol writes that “getting hundreds of biomarker results and imaging tests in an individual greatly increases the likelihood of false-positive results,” a concerning possibility.
I’ve discussed the challenge of false positives here, and get into some of the details around Bayes Theorem (which informs the assessment) here. The OG reference in this space may be this 2006 paper by Zak Kohane and colleagues, in which they introduce the term “incidentalome.”
To be fair, at least some of the proponents of extensive testing recognize the challenge of false positives but feel that the opportunity to collect dense data on individuals over time enables important inflections to be observed, a point Dr. Peter Attia explicitly emphasizes in Outlive; I discuss his “risk-management” mindset here.
Similarly, Nathan Price, a professor at the Buck Institute and the CSO of Thorne, has argued that close inspection (assisted by AI) of rich individual data could identify (for example) opportunities for supplement intervention. These interventions may not make much of a difference on the population level (hence the paucity of persuasive clinical trial data for supplements, as Dr. Topol notes in his latest book, Super Agers – my WSJ review here), but could in selected individuals. (I also discuss Price here, here).
Proponents of the “personal health OS” also might emphasize the presence of tailwinds – the likelihood of improved predictions as measurement technologies continue to get better, denser data become available, and the AI tools become ever-more capable. Perhaps we’re not quite at the point of realizing the future we imagine, advocates might argue, but we’re close enough to start to see what it might look like.
Concern Two: A Constricted View of Health
What’s arguably a deeper concern about the model of health we seem to be moving towards is the degree to which it seems to be informed by a rigidly reductive mindset. In this limited, classically managerial (or consultant) view, health becomes simply metrics on a dashboard, an ever-expanding series of parameters that must constantly be measured, quantified, optimized.
A recent, beautiful essay about our evolving understanding of and approach to happiness in the New York Times Magazine by Kwame Anthony Appiah reminds us what we may be missing.
Around the start of the new Millenium, Appiah writes, we entered
the life-hacking, self-quantifying, habit-stacking era of optimization gurus like Tim Ferriss, whose first book, published in 2007, was “The 4-Hour Workweek” — “a toolkit,” in his words, “for maximizing per-hour output.”
Consequently, Appiah continues, the concept of flourishing was decomposed into “modular upgrades” as we refine our “personal operating system.”
Yet it’s essential to recognize, Appiah writes, that “happiness is not an optimization problem,” but something deeper and more substantial.
I reached for a similar point in 2018, in a piece entitled, “We Are Not a Dashboard.”
Observing that the “dashboard has become a potent symbol of our age,” I wrote that “the ideology of big data has taken on a life of its own, assuming a sense of both inevitability and self-justification.”
I continued, “From measurement in service of people, we increasingly seem to be measuring in service of data, setting up systems and organizations where constant measurement often appears to be an end in itself.”
I’m reminded of a favorite phrase from Kate Crawford’s Atlas of AI (my WSJ review here): “The affordances of the tools become the horizon of truth,” a reminder, in this context, that even if we’re awash in tools enabling the measurement and analysis of health data, we must ensure our understanding of health transcends the limits of these tools.
Of course, the point isn’t to go the other way, and reject metrics completely.
As Professor Jerry Muller, author of the brilliant book Tyranny of Metrics, explains, “I can’t see how competent experts could ignore metrics. The question is their ability to evaluate the significance of the metrics, and to recognize the role of the unmeasured.” (emphasis added).
I also spoke to this need in a 2011 piece entitled “What Silicon Valley Doesn’t Understand About Medicine,” writing, ”a novel technology platform that overlooks the integrated needs of patients or underestimates or fails to account for the complexity and messiness of illness as it actually occurs and is experienced by patients (and those closest to them) will inevitably fall short.”
Moving Forward
To most effectively meet the needs of patients – including the vitally important goal of preventing or preempting disease so people don’t become patients – it’s essential to embrace the power and promise of emerging technologies, including those enabling the conceptualization of “personal health OS,” while not mistaking this map for the territory (as Alfred Korzybski famously instructed).
It will be essential to establish priorities – in partnership with each patient – and identify a handful of key health parameters on which to focus on; Drs. David Blumenthal and J. Michael McGinnis discuss the topic of “core metrics” thoughtfully in this 2015 JAMA “Viewpoint.”
At the same time, we must hold fast to a vision of health and wellness that expands far beyond the confinement of a dashboard and aspires to something beyond the recursive optimization of metrics (as I recently discussed here). Our approach must be capacious enough to include, authentically value, and meaningfully cultivate other components of a healthy, flourishing life, which might include intellectual captivation, the pursuit of purpose, and social engagement with family, friends, and community.
(Martin Seligman’s PERMA model — positive emotion/joy, engagement/flow, relationships/connection with others, meaning/purpose, and accomplishment — represents a potentially useful framework [see here, here] for expanding our thinking.)
Despite the difficulty, if not utter impossibility, of reducing some of the most important and profound components of health to an easily digested number, we must continue to value and pursue them.
Even as we diligently leverage emerging technology to construct and refine health dashboards, let’s resolve to work towards a more expansive, durable, and meaningful vision of health that exists beyond the sterile syntax of rows, columns, and digits.
来源：非常道
