pharmacovigilance-guide

本条目面向分析数据与指标归因场景,提供“pharmacovigilance-guide”技能的目录入口。公开标签:learning、药物警戒、不良事件、安全信号。请通过固定上游版本核对实际功能、依赖与权限;此页为 Skillπ 独立编写的用途介绍和原站链接,不代表运行安全保证。原文与附件依各自许可开放。

仓库
brycewang-stanford/Auto-Empirical-Research-Skills
仓库最近推送
Stars
5,023
Forks
578
Issues / PRs
0

安装方式

先用 Prompt 核对来源,也可切换到安装命令或下载可用的安装包。

安装方式

用 AI 助手安装

复制提示词,让助手先核对固定版本和依赖。

查看 Prompt收起 Prompt
先检查当前 Agent 是否已安装 SkillPI Skill Finder。如果尚未安装,请按官方仓库的说明安装并启用:https://github.com/Findskills/skillpi。然后使用 Skill Finder 打开此 SkillPI 详情页并定位 "pharmacovigilance-guide":https://skillpi.com/zh/skills/deepskill-bfd19e091fbe953db0d7a1d42a7dafb8d7eb9ee9c08c31cbe35be4851d2fc794。安装前核对上游来源和版本:https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/blob/bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba/skills/43-wentorai-research-plugins/skills/domains/pharma/pharmacovigilance-guide/SKILL.md。来源不一致时停止并说明差异。核对通过后,用 Skill Finder 安装目标 Skill,并报告安装路径和来源版本。

此命令从 GitHub 来源仓库获取指定技能。运行前请先查看源文件。

运行工具
npx skills add https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/tree/bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba/skills/43-wentorai-research-plugins/skills/domains/pharma/pharmacovigilance-guide --skill pharmacovigilance-guide
许可协议: —

来源与配置

查看 GitHub 源码

brycewang-stanford/Auto-Empirical-Research-Skills固定提交 bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba

配置方式待核实

请查看原始来源中的适用工具、依赖和权限要求。

安装实测待补充

本条目面向分析数据与指标归因场景,提供“pharmacovigilance-guide”技能的目录入口。公开标签:learning、药物警戒、不良事件、安全信号。请通过固定上游版本核对实际功能、依赖与权限;此页为 Skillπ 独立编写的用途介绍和原站链接,不代表运行安全保证。原文与附件依各自许可开放。

↓ 0 下载 ↗ 0 分享点击

文件浏览

在原站浏览文件 ↗

原始 SKILL.md

打开原文 ↗
---
name: pharmacovigilance-guide
description: "Adverse drug event detection, safety signal mining, and drug monitoring"
metadata:
  openclaw:
    emoji: "⚠️"
    category: "domains"
    subcategory: "pharma"
    keywords: ["pharmacovigilance", "adverse-events", "drug-safety", "faers", "signal-detection", "disproportionality"]
    source: "wentor"
---

# Pharmacovigilance Guide

A skill for computational pharmacovigilance research, covering adverse drug event (ADE) databases, signal detection algorithms, disproportionality analysis, and safety surveillance methods used in post-market drug monitoring.

## Adverse Event Data Sources

### Key Databases

| Database | Operator | Coverage | Access |
|----------|----------|----------|--------|
| FAERS (FDA Adverse Event Reporting System) | FDA | US spontaneous reports | Free quarterly downloads |
| EudraVigilance | EMA | European reports | Research access via application |
| VigiBase | WHO-UMC | Global (150+ countries) | Research license |
| VAERS | CDC/FDA | US vaccine adverse events | Free download |
| MAUDE | FDA | Medical device reports | Free download |

### Loading FAERS Data

```python
import pandas as pd
import zipfile
import os

def load_faers_quarter(data_dir: str, year: int, quarter: int) -> dict:
    """
    Load FAERS quarterly data files into DataFrames.
    Downloads available from: fis.fda.gov/extensions/FPD-QDE-FAERS/FPD-QDE-FAERS.html
    Returns dict of DataFrames for each file type.
    """
    prefix = f"faers_ascii_{year}Q{quarter}"
    tables = {}

    file_map = {
        "DEMO": "demographics",    # Patient demographics
        "DRUG": "drugs",            # Drug information
        "REAC": "reactions",        # Adverse reactions (MedDRA terms)
        "OUTC": "outcomes",         # Patient outcomes
        "INDI": "indications",     # Drug indications
        "THER": "therapy",          # Therapy dates
        "RPSR": "report_sources",   # Report source
    }

    for suffix, name in file_map.items():
        filepath = os.path.join(data_dir, f"{suffix}{year}Q{quarter}.txt")
        if os.path.exists(filepath):
            tables[name] = pd.read_csv(
                filepath, sep="$", encoding="latin-1",
                low_memory=False, on_error="warn"
            )

    return tables

# Example: Load and inspect
faers = load_faers_quarter("./faers_data", 2024, 3)
print(f"Reports: {len(faers['demographics']):,}")
print(f"Drug-reaction pairs: {len(faers['reactions']):,}")
```

## Signal Detection Methods

### Disproportionality Analysis

Disproportionality measures compare the observed frequency of a drug-event pair against the expected frequency under independence:

```python
import numpy as np
from scipy.stats import chi2

def compute_disproportionality(a: int, b: int, c: int, d: int) -> dict:
    """
    Compute disproportionality measures from a 2x2 contingency table:

              Event+    Event-
    Drug+       a          b
    Drug-       c          d

    a: reports with both the drug and the event
    b: reports with the drug but not the event
    c: reports with the event but not the drug
    d: reports with neither
    """
    n = a + b + c + d
    expected = (a + b) * (a + c) / n if n > 0 else 0

    # Reporting Odds Ratio (ROR)
    ror = (a * d) / (b * c) if b * c > 0 else float("inf")
    ln_ror = np.log(ror) if ror > 0 and ror != float("inf") else 0
    se_ln_ror = np.sqrt(1/a + 1/b + 1/c + 1/d) if min(a, b, c, d) > 0 else float("inf")
    ror_lower = np.exp(ln_ror - 1.96 * se_ln_ror)

    # Proportional Reporting Ratio (PRR)
    prr = (a / (a + b)) / (c / (c + d)) if (a + b) > 0 and (c + d) > 0 else 0

    # Information Component (IC, Bayesian shrinkage)
    ic = np.log2((a + 0.5) / (expected + 0.5)) if expected > 0 else 0

    # Chi-squared with Yates correction
    chi2_val = (n * (abs(a * d - b * c) - n / 2) ** 2) / (
        (a + b) * (c + d) * (a + c) * (b + d)
    ) if min(a + b, c + d, a + c, b + d) > 0 else 0

    return {
        "a": a, "b": b, "c": c, "d": d,
        "expected": round(expected, 2),
        "ROR": round(ror, 3),
        "ROR_lower_95": round(ror_lower, 3),
        "PRR": round(prr, 3),
        "IC": round(ic, 3),
        "chi2": round(chi2_val, 3),
        "signal": ror_lower > 1 and a >= 3 and chi2_val > 3.84,
    }
```

### Multi-Item Gamma Poisson Shrinker (MGPS)

The MGPS method (used by FDA) applies empirical Bayesian shrinkage to stabilize estimates for rare events:

```python
def empirical_bayes_geometric_mean(observed: np.ndarray,
                                     expected: np.ndarray) -> np.ndarray:
    """
    Simplified EBGM computation.
    Shrinks observed/expected ratios toward the overall mean,
    reducing false positives from small counts.
    """
    # Raw ratio
    rr = observed / np.maximum(expected, 0.01)

    # Empirical Bayes shrinkage (simplified two-component mixture)
    # Full implementation uses EM algorithm to fit mixture of gammas
    global_mean = np.mean(rr)
    shrinkage = expected / (expected + 1)  # more shrinkage for small expected
    ebgm = shrinkage * rr + (1 - shrinkage) * global_mean

    return ebgm
```

## MedDRA Terminology

### Medical Dictionary for Regulatory Activities

MedDRA provides the standardized terminology for adverse event coding:

```
Hierarchy (5 levels):
  System Organ Class (SOC)      -- e.g., "Cardiac disorders"
    High Level Group Term (HLGT) -- e.g., "Cardiac arrhythmias"
      High Level Term (HLT)      -- e.g., "Supraventricular tachyarrhythmias"
        Preferred Term (PT)      -- e.g., "Atrial fibrillation"
          Lowest Level Term (LLT) -- e.g., "Auricular fibrillation"
```

### Standardized MedDRA Queries (SMQs)

Pre-defined search strategies for known safety topics:

- **Anaphylactic reaction (SMQ)**: Broad and narrow search terms
- **Drug-induced liver injury (SMQ)**: Hy's Law criteria
- **Torsade de pointes / QT prolongation (SMQ)**: Cardiac safety signals
- **Rhabdomyolysis (SMQ)**: Muscle-related adverse events

## Temporal Pattern Analysis

### Time-to-Onset Analysis

```python
def time_to_onset_analysis(drug_start_dates: pd.Series,
                            event_dates: pd.Series) -> dict:
    """
    Analyze time-to-onset distribution for a drug-event pair.
    Useful for distinguishing causal signals from coincidental reports.
    """
    ttp = (event_dates - drug_start_dates).dt.days
    ttp = ttp[ttp >= 0]  # exclude negative (data quality issue)

    return {
        "n_reports": len(ttp),
        "median_days": ttp.median(),
        "mean_days": ttp.mean(),
        "q25_days": ttp.quantile(0.25),
        "q75_days": ttp.quantile(0.75),
        "within_30_days_pct": (ttp <= 30).mean() * 100,
        "within_90_days_pct": (ttp <= 90).mean() * 100,
    }
```

## Causality Assessment

Standard frameworks for evaluating whether a drug caused an adverse event:

| Method | Type | Key Criteria |
|--------|------|-------------|
| WHO-UMC | Algorithmic | Temporal, dechallenge, rechallenge, alternative causes |
| Naranjo Score | Scoring scale | 10 questions, score 0-13 (definite/probable/possible/doubtful) |
| Bradford Hill | Principles | Strength, consistency, specificity, temporality, biological gradient |

## Tools and Resources

- **openFDA API**: Direct access to FAERS data via REST
- **OHDSI / OMOP CDM**: Standardized observational health data for pharmacoepidemiology
- **PhViD (R package)**: Pharmacovigilance signal detection methods
- **EHRtemporalVariability**: R package for temporal data quality in EHR
- **VigiRank**: WHO-UMC signal prioritization algorithm
- **AEOLUS**: Standardized and cleaned version of FAERS data

版本历史

  1. v2026.07 GitHub ↗
    原版更新说明

    ## Highlights - **First tagged release.** Every gate is green: `make check` (validate + 183 unit tests + eval-harness + numeric benchmarks), validate-catalog, quality-evals, and OpenSSF Scorecard on `main`. - **Rigor double loop at 13 method families.** Each family closes taxonomy tag → eval scenario (does the agent avoid the method's known traps?) → deterministic numeric benchmark (does it recover the true effect?). New this cycle: heterogeneous treatment effects (CATE) and quantile treatment effects (QTE). - **Paper-WorkFlow flagship at 29/29 executable rigor gates**, now merged to its `main` with a drift-gated README badge — the credibility artifact is checked by CI, so it can never silently go stale ([RIGOR.md](https://github.com/brycewang-stanford/Paper-WorkFlow/blob/main/RIGOR.md)). - **1,150 vendored skills / 69 collections**, all license-audited, security-scanned, tracked in a machine-readable catalog with CI-enforced freshness, and importable as a single whole-repo router skill. - **Hardened weekly upstream syncs**: synced trees must pass the full gate before merge; this cycle both sync PRs were verified, patched (StatsPAI's SkillOpt execution-gate card restored), and merged. - **First-party plugin marketplace** (Python / R / Stata empirical-analysis plugins + guide) shipped in the previous cycle, now under a tagged version. ## Known Follow-Ups - Ecosystem benchmark comparison (AERS vs econometrics agents on the shared benchmark suite) — [PLAN-2026-07 week 3](https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/blob/main/docs/PLAN-2026-07.md). - Decomposition (Oaxaca–Blinder) and shift-share (Bartik) method families (weeks 3–4). - GitHub Pages catalog site, README demo screenshots, five-language README consistency script (week 4). ## Catalog - Top-level collections: **69** - `SKILL.md` files: **1150** - Cataloged agent tools: **335** ([`tools/CATALOG.md`](../tools/CATALOG.md)) ## Methodological rigor - Method families in the coverage map: **13** ([`RIGOR_COVERAGE.md`](RIGOR_COVERAGE.md)) - Eval scenarios: **26** (138 rubric items, 122 auto-checkable) ([`../eval-harness/`](../eval-harness/README.md)) - Benchmark tasks: **13** (70 gold checks, 58 required) ([`../benchmark/`](../benchmark/README.md)) ### Method families (skills tagged) - `iv`: 27 - `rdd`: 25 - `panel-fe`: 19 - `staggered-did`: 17 - `bayesian`: 13 - `did`: 10 - `event-study`: 10 - `matching`: 10 - `synthetic-control`: 10 - `cate`: 8 - `dml`: 8 - `survival`: 3 - `quantile`: 2 ### Eval scenarios by category - `causal-identification`: 14 - `citation-hygiene`: 1 - `reproducibility`: 2 - `research-integrity`: 1 - `runtime-safety`: 1 - `statistical-validity`: 2 - `writing-compliance`: 3 - `writing-style`: 2 ## License buckets - AGPL-3.0: 1 - Apache-2.0: 1 - CC-BY-NC-4.0 (non-commercial): 1 - CC-BY-SA-4.0: 1 - CC-BY-SA-4.0 (repository default): 4 - GPL-3.0: 1 - MIT: 35 - MIT Non-Commercial: 1 - Mixed (per-skill; see folder): 2 - UNKNOWN - check upstream: 22 ## Quality gates - `make check` — catalog/provenance/audit/eval freshness, repo hygiene, Paper-WorkFlow gates, stdlib unit tests, eval-harness lint + graded smoke, and the strict numeric benchmark (reference-gold drift fails the build). - External links and OpenSSF Scorecard run on their scheduled workflows.

当前收录版本

0.0.1

核对来源版本 ↗

质量评测

满分 100。前四项来自 DeepSkill 收录的质量报告。

场景价值88/100
评分依据

提供药物警戒领域完整技术栈:从 FAERS/EudraVigilance 等权威数据库接入,到 ROR/PRR/IC/MGPS 等核心信号检测算法实现

触发精准92/100
评分依据

触发词精准覆盖 pharmacovigilance、adverse-events、drug-safety、faers、signal-detection 等专业术语

结构规范90/100
评分依据

文档结构清晰,分模块讲解数据来源、信号检测、术语体系、时间分析和因果关系

执行可靠89/100
评分依据

所有代码示例均包含合理的异常处理(如除零保护、数据质量过滤)

证据可验 (前四项均值)90/100
计算方式

取前四项原始分值的算术平均,四舍五入取整。可通过收录报告核对各项评分与说明,运行记录需单独查证。

权限与使用前检查

根据当前主文档自动整理,命中内容可能是示例或防护说明;不是实际权限声明,也未覆盖附件、外部依赖或运行行为。

未从主文档识别到上述常见操作,不能据此判断无需权限或没有风险。

本站展示提示不会授予权限。涉及文件改动、密钥、外部发送或费用时,仍由你使用的工具按需确认。

安全评级与访问范围

risk · 静态证据核验,不代表运行安全保证。

使用前请由宿主核对并授权实际文件读写、外部服务和联网目的地;本站的核验不会授予任何运行权限。

来源与使用声明 SkillPi 的部分技能资料来自公开来源或创作者提交。第三方内容的权利归各自权利人所有。安装或运行前,请查看原站的许可、权限要求和使用风险。

来源与使用

请按详情页的作者、原始来源和许可证核对使用条件,运行前检查技能需要的权限。目录收录不等于已验证安装或安全性。