Introduction

Developers creating clinical decision support applications often face the same challenge: the model might recite clinical guidelines and calculate the Number Needed to Treat (NNT), but what the user actually asks is, “Should this patient undergo this test? Should this medication be prescribed? Should this referral be made?” If allowed to generate responses freely, the model’s output varies in quality, and it struggles to articulate evidence strength and uncertainty. The “Clinical Decision Making & Evidence-Based Medicine” Skill introduced below addresses this specific gap. It utilizes the PICO framework to structure clinical questions, employs GRADE to grade evidence, performs quantitative trade-offs using NNT/NNH, and preserves the physician’s ultimate diagnostic authority regarding key metrics. For developers, its value lies in solidifying the evidence-based decision-making workflow, eliminating the need to reason from scratch every time.

What is this

“Clinical Decision Making & Evidence-Based Medicine” is a community-published Skill located at @user_054f2eea/clinical-decision-making. The current version is 1.0.0, categorized as “professional,” and sourced from the community. It is published by user_054f2eea and requires no API key. In one sentence, it performs PICO question construction, evidence retrieval and GRADE grading, NNT/NNH decision analysis, and uncertainty management, ensuring that every clinical decision is evidence-based and measurable. Official URL: https://api.skillhub.cn/user_054f2eea/clinical-decision-making ; SkillHub Page: https://www.skillhub.cn/skills/user_054f2eea/clinical-decision-making .

It solves the problem of “uncontrollable model output in clinical decision scenarios.” Unlike directly asking the model to answer freely, this Skill requires constructing a structured clinical question first, then retrieving and grading evidence, performing decision analysis, and finally clarifying uncertainty. The process is reusable, and the results are explainable.

Core Functions

There are four verified capabilities corresponding to a complete chain of evidence-based decision-making:

  1. PICO Question Construction. It breaks down clinical questions into four elements: Population (P), Intervention (I), Comparison (C), and Outcome (O). This prevents overly broad questions and provides a handle for subsequent retrieval and grading.

  2. Evidence Retrieval & GRADE Recommendation Grading. It grades the retrieved evidence according to the GRADE system to provide recommendation strength, rather than offering vague “strong/weak recommendations.”

  3. Clinical Decision Analysis. It performs quantitative trade-offs using NNT (Number Needed to Treat), NNH (Number Needed to Harm), and decision thresholds. The resource provides two core guidelines: an NNT < 10 is usually worth doing; caution is required if NNH < NNT.

  4. Uncertainty Management. When evidence is insufficient or disputed, it provides management pathways such as watchful waiting, therapeutic trials, or referral timing, rather than forcing a conclusion.

There are three core criteria that run through the entire Skill: Evidence is not a rule (it must be combined with experience and patient preference); NNT < 10 is usually worth doing, but caution is required if NNH < NNT; Patient preference is a necessary input for decision-making.

Furthermore, the Skill incorporates cautionary mechanisms: specific NNT/NNH values are not finalized decisions (referred to as the L5 caution mechanism in the documentation) and includes counter-evidence review and confidence labeling (H/M/L). This is a crucial safety boundary for developers: quantitative results are for reference only and are not a basis for clinical management.

Installation and Activation

It is important to note: the retrieved documentation does not provide installation commands, directory structures, or configuration fields for this Skill, nor does it detail differences between Cursor, Codex CLI, and Claude Code. Therefore, no commands will be fabricated here. Please refer to the official pages for acquisition methods:

Official URL:

https://api.skillhub.cn/user_054f2eea/clinical-decision-making

SkillHub Page:

https://www.skillhub.cn/skills/user_054f2eea/clinical-decision-making

Attributes confirmed from the documentation: it does not require an API key (tagged as requires_api_key=false); it is categorized under healthcare, within the professional health skills category. Please check the official page for the current installation method and runtime environment requirements before installing.

Typical Usage

The retrieved documentation does not provide reproducible example scripts or call samples for this Skill. Therefore, this section only describes its trigger methods and typical question forms without fabricating inputs or outputs.

Trigger Methods: The Skill is activated when a user asks questions related to clinical decisions or evidence-based medicine. Trigger words listed in the documentation include: clinical decision, whether to do this test, whether to use this drug, whether to refer, evidence-based medicine, evidence strength, GRADE. In other words, when a user asks, “Should this patient use this medication?” or “What is the strength of this evidence?”, the Skill’s workflow is engaged.

Typical Question Forms can be categorized into four types, corresponding to the four links of the core functionality:

  • Question Construction: “How should a clinical question be framed regarding a specific population, intervention, comparison, and outcome?”
  • Evidence Grading: “What is the evidence strength and recommendation strength of a specific study or guideline?”
  • Decision Analysis: “What is the NNT/NNH for a specific intervention, and is it worth doing?”
  • Uncertainty: “When evidence is insufficient, should we wait and see, conduct a therapeutic trial, or refer?”

Because the Skill includes counter-evidence review and confidence labeling, actual outputs will provide the conclusion, counter-evidence, and an H/M/L confidence rating, rather than a single conclusion. Note for usage: Quantitative values (NNT/NNH) are for decision-making reference only and must not be used directly as a basis for clinical management.

Use Cases and Precautions

It is suitable for two types of users. First, developers of clinical decision support tools who need to solidify evidence-based decision workflows into reusable capabilities rather than reasoning from scratch every time. Second, users who need to verify evidence strength and decision thresholds, such as in scenarios requiring quantitative references for treatment selection, diagnostic testing, or referral timing.

Limitations and Pitfalls: There are three points, derived from the disclaimer and core criteria in the documentation, which must be honestly communicated to users:

  1. It does not replace physician diagnosis or clinical decision-making, nor does it constitute medical practice. This is the official positioning of the Skill and reflects the principle that the final diagnostic authority rests with the physician.
  2. Evidence grading and decision analysis are based on public literature and guidelines, which may suffer from update lag and individual variations. This means that values like NNT/NNH may differ from the latest research and must be judged in the context of the patient’s actual situation.
  3. Quantitative values such as NNT/NNH are for decision-making reference only and must not be directly used as a basis for clinical management. The L5 caution mechanism built into the Skill is designed to reduce the risk of misuse: specific values do not yield a final decision but are used by qualified physicians to make decisions based on the patient’s actual condition.

Additionally, since this Skill requires integrating patient preferences (one of the core criteria), simply inputting a disease name while ignoring patient preference will weaken the quality of the decision. Patient preference should be provided as a necessary input during use.

Conclusion

In summary: this Skill breaks evidence-based decision-making down into four components—question construction, evidence grading, decision analysis, and uncertainty management—while preserving the physician’s ultimate authority regarding quantitative results. For scenarios requiring “every clinical decision to be evidence-based and measurable,” it acts as a process-oriented tool rather than an automatic conclusion machine. Official URL: https://api.skillhub.cn/user_054f2eea/clinical-decision-making ; SkillHub Page: https://www.skillhub.cn/skills/user_054f2eea/clinical-decision-making . Please refer to the official pages for specific installation and activation instructions.