为什么预防效果数字无法被支付方接受Why Prevention Data Is Rejected by Payers
想象这样一个场景：一家企业为1000名员工推行了心血管健康管理项目，一年后，参与者的平均心血管风险评分下降了12%，而没有参与的员工只下降了3%。差异显著。Imagine an enterprise runs a cardiovascular health program for 1,000 employees. After one year, participants' average cardiovascular risk score dropped 12%, versus 3% for non-participants. The difference looks significant.
但当企业拿着这份数据去找保险公司谈费率优惠，精算师会问一个让人哑口无言的问题：你怎么证明这12%的改善是你的干预造成的，而不是因为参与项目的员工本来就更注重健康？But when the enterprise takes this data to negotiate insurance discounts, the actuary asks a devastating question: How do you prove this 12% improvement was caused by your intervention, rather than by the fact that health-conscious employees self-selected into the program?
这个问题，就是选择偏倚。This is selection bias.
选择偏倚：相关性证据的致命漏洞Selection Bias: The Fatal Flaw of Correlation Evidence
主动参与健康管理项目的人，往往在参与之前就已经比不参与的人更健康。他们有更强的健康意识，更规律的生活习惯。换句话说，两组人在参与干预之前就已经不可比了。People who actively join health management programs tend to already be healthier before joining — stronger health awareness, more regular habits. In other words, the two groups were not comparable before the intervention even started.
核心问题：Core Issue:相关性可以告诉你"参与者比不参与者健康"，但无法告诉你"是因为参与才更健康"。支付方需要的是后者——因果证据。Correlation can tell you "participants are healthier than non-participants," but cannot tell you "they are healthier because they participated." Payers need the latter — causal evidence.
PSM：跨越方法论鸿沟的工具PSM: The Tool to Bridge the Methodological Gap
倾向评分匹配（PSM）的核心思路是：通过统计方法，在对照组中为每一个干预组成员找到一个"基线特征高度相似"的匹配对象。匹配之后，两组在可观测特征上高度可比，消除了可观测的选择偏倚。在这个匹配后的配对样本上计算出的效果差异，才具有因果解释力。PSM's core idea: use statistical methods to find a control group member with highly similar baseline characteristics for each treatment group member. After matching, the two groups become comparable on observable characteristics, eliminating observable selection bias. The effect difference calculated on matched pairs then has causal interpretability.
相关性和因果性的区别，不是统计技术问题，而是支付体系能否接受证据的根本前提。没有因果证据，预防干预的商业价值无法被支付方正式认可。The difference between correlation and causation is not a statistical technicality — it is the fundamental prerequisite for whether payers will accept the evidence. Without causal evidence, prevention's commercial value cannot be formally recognized.