ISO 13485 里与检测相关的条款怎么落地

结论：体系条款的很多要求，最终落点是一份数据

ISO 13485 的条款读起来是管理语言，但其中相当一部分的最终证据形式是检测数据。写了程序文件、做了培训记录，但拿不出对应的数据，条款就没有真正落地。

审核时被追到底的，往往就是这个环节：制度有、流程有、记录有，唯独缺了证明结论成立的那份测量结果。

主要条款与对应的检测证据

验证与确认的区别要分清

这是落地时最常混淆的一对。

验证回答的是「做得对不对」——设计输出是否满足设计输入。输入写了某项性能指标，验证就要测这项指标看达没达到。

确认回答的是「做的是不是对的东西」——产品在预期使用条件下能不能满足用户需求。它关注的不是单项指标，而是整体使用效果。

实际中常见的问题是：验证做得很充分，确认只有一份用户满意度调查。确认同样需要客观证据，比如模拟使用条件下的性能表现、人因可用性的评价数据。

过程确认的数据要求

过程确认针对的是那些输出无法通过后续检验完全验证的过程，比如灭菌、焊接、注塑、粘接。

这类过程的确认需要建立过程参数与产品特性之间的关系：参数在什么范围内，产品特性能稳定落在合格区间。这个关系要用数据建立，不能靠经验声明。

落地不到位的典型表现是：做了三批验证，结论写「过程稳定」，但没有说明参数的允许波动范围。这样的确认在参数发生漂移时无法判断是否仍在有效范围内。

测量设备控制这一条常被低估

所有检测数据的可信度都依赖测量设备。这一条落地不到位，前面所有数据的说服力都打折。

要做到的包括：设备有校准且在有效期内；校准的量程覆盖实际使用范围；两次校准之间有期间核查；设备状态异常时能追溯受影响的数据。

最后一项最常缺失。 如果发现设备失准，需要能回答「从上次确认正常到现在，用这台设备测过哪些产品」。没有这个追溯能力，只能扩大范围处理，代价很大。

体系文件与检测报告怎么衔接

建议建立一张对应表，把体系要求的验证活动与实际的检测报告对应起来：某项设计输入，对应哪份报告的哪个项目，报告编号是多少。

这张表的作用有三个：自查时能发现哪些要求没有证据支撑；审核时能快速定位；产品变更时能判断哪些验证需要重做。

做这张表的过程本身就会发现缺口。 实际操作中，把设计输入逐条对应下来，常会发现有两三条从来没有被验证过——不是故意漏，而是在层层文件传递中被遗忘了。

委托外部检测时的体系要求

检测委托给外部机构，不等于责任外包。体系上仍然要做到：对外部机构的能力做过评价并有记录；委托的项目和要求有明确的文件；收到报告后有确认和接收的记录；报告纳入相应的质量记录管理。

第一项在审核中经常被问。 选择这家实验室的依据是什么、评价了哪些方面、记录在哪里。临时找一家做了就用，评价记录缺失，这一条就站不住。

记录保存期限的确定

检测记录要保存多久，体系文件里应当有规定，而规定的依据是产品寿命和法规要求中较长的那个。

实际操作中常见的问题是一刀切：所有记录统一保存若干年。这样处理简单，但对于寿命较长的产品可能不够——产品还在使用，支撑其符合性的记录已经销毁了。

建议按产品分类设定保存期，并在销毁前做一次确认，避免机械执行造成证据缺失。

数据完整性的基本要求

无论纸质还是电子，检测数据应当满足几项基本属性：可归属到人、清晰可读、同步记录、原始形态保留、准确。

这几项里「同步记录」最容易被忽略——先在草稿纸上记，事后誊抄到正式记录上，这个做法在审核中会被质疑，因为誊抄环节引入了出错和修改的可能。正确的做法是直接记录在受控的记录表上。

供方评价怎么做才站得住

对外部检测机构的评价，常见做法是收集一份资质证书存档。这在审核中通常不够，因为它只证明对方有资质，没有证明你评价过它是否适合你的项目。

比较扎实的做法包含几项：确认认可范围覆盖你要委托的具体项目；了解对方在该类产品上的经验；确认报告形式满足你的申报用途；对合作过程中的响应与质量做定期回顾。

最后一项是动态评价，一次性评价之后长期不更新，也会被质疑。建议按年度做一次简单回顾并留记录。

检测能力的内外分工

企业要决定哪些项目自己做、哪些委外。判断依据除了成本，还有三点：该项目的频次（高频适合自建）、设备投入与维护成本、以及该项目出具的报告是否需要第三方资质。

自建能力意味着相应的体系负担：设备管理、人员培训、方法确认、期间核查。这部分持续投入常被低估，只算设备购置费用做决策，往往低估了实际成本。

我们的做法

我们向委托方提供的报告，会把试验条件、方法依据和设备信息写清楚，原因之一就是这些内容会进入委托方的体系记录。报告上信息不全，委托方在体系审核时就要额外解释。

对于有体系审核需求的委托方，我们也可以提供实验室的资质文件与认可范围，供其完成供方评价的记录要求。这类材料通常在委托阶段一并提供比较省事。

如果你正在梳理体系里的检测证据，或者审核中被指出检测记录不足，可以把具体条款和现有记录发过来一起看，或者直接联系：132 4819 8029。能力范围见服务介绍，流程见检测流程，更多内容见知识库。

English version

Conclusion. The clauses of ISO 13485 read as management language, but for a substantial number of them the ultimate evidence is measurement data. Having a procedure, training records and forms does not satisfy a clause if the data demonstrating the conclusion are missing. In audits, this is usually where questioning goes deepest: the system exists, the process exists, the records exist, and what is absent is the measurement result supporting the conclusion.

Clauses and their evidence. Design verification requires test data addressing each design input. Design validation requires data obtained under simulated or actual use conditions. Process validation requires data linking process parameters to product characteristics. Verification of purchased product requires incoming inspection records. Monitoring and measurement of product requires in-process and final inspection records. Control of monitoring and measuring equipment requires calibration certificates and intermediate checks. Control of nonconforming product requires re-inspection data and review records.

Verification versus validation. This pair causes the most confusion. Verification asks whether the design output meets the design input: if the input specifies a performance figure, verification measures it. Validation asks whether the product meets user needs under intended use conditions, which is about overall effect rather than individual figures. A common shortfall is thorough verification accompanied by validation consisting only of a user satisfaction survey. Validation also needs objective evidence, such as performance under simulated use conditions or usability data.

Process validation data. This applies to processes whose output cannot be fully verified by subsequent inspection, such as sterilisation, welding, moulding and bonding. Validation must establish the relationship between process parameters and product characteristics: within what parameter range does the characteristic remain reliably inside the acceptance range. That relationship must be established with data rather than asserted from experience. A typical shortfall is three validation batches concluding that the process is stable, with no statement of the permitted parameter window, leaving no way to judge whether a later drift remains acceptable.

Control of measuring equipment is underrated. The credibility of every test result rests on it. Requirements include calibration within validity, calibration points covering the actual working range, intermediate checks between calibrations, and the ability to trace which products were measured with an instrument found out of tolerance. That last capability is the one most often missing, and without it the only option is to widen the scope of the response, which is expensive.

Linking system documents to test reports. Build a table mapping each required verification activity to the specific report and item that provides the evidence, with report numbers. It serves three purposes: finding requirements with no supporting evidence, locating evidence quickly during audit, and judging which verifications must be repeated when the product changes. Building the table itself reveals gaps; mapping design inputs one by one commonly uncovers two or three that were never verified, lost in the handover between documents rather than deliberately omitted.

Outsourced testing. Commissioning an external laboratory does not outsource responsibility. The system still requires a documented evaluation of that laboratory's capability, a clear specification of what was commissioned, a record of receipt and acceptance of the report, and inclusion of the report in quality records. The first of these is frequently questioned in audit: what was the basis for selecting this laboratory, what was evaluated, and where is the record. Using a laboratory found at short notice, with no evaluation record, does not satisfy the clause.

How we handle it. Our reports state test conditions, method basis and equipment information, partly because that content enters the client's quality records. An incomplete report means the client has to explain the gap during audit. For clients with audit requirements we can also provide accreditation documents and scope to support their supplier evaluation records, which is simplest to arrange when the work is commissioned.

Send us the clauses in question and your existing records and we will work through it. Phone or WeChat: +86 132 4819 8029.