Quantitative Microbial Risk Assessment Model in R

We have developed an open-source implementation of the FDA-iRISK comparative risk assessment framework using the R programming language. This model performs quantitative microbial risk assessment (QMRA) for foodborne hazards by simulating pathogen behavior from processing through consumption, ultimately estimating public health burden in disability-adjusted life years (DALYs). Model Structure The model follows a modular, stage-based process: Initial conditions: User-defined prevalence and concentration distributions for the pathogen at the starting point of the supply chain Process stages: Sequential nodes representing steps such as partitioning, storage (growth or die-off), and handling, each with mathematically defined effects on pathogen prevalence and concentration Exposure assessment: Monte Carlo simulation combining final pathogen concentration with serving size distributions to generate ingested dose distributions Hazard characterization: Pathogen-specific dose-response models (Beta-Poisson for Salmonella, exponential for Listeria) that convert dose to probability of illness Risk characterization: Multiplication of per-serving risk by annual eating occasions and DALY weights to estimate total annual burden Implementation The model is implemented in R using the mc2d package for multivariate Monte Carlo simulation, enabling separation of variability and uncertainty. All code is provided as an executable R Markdown file with embedded functions for: Three validated case studies (Salmonella in peanut butter; Listeria in soft cheese and cantaloupe) Intervention scenario testing (e.g., contamination reduction, temperature control) Uncertainty analysis for dose-response parameters Enhanced visualizations using ggplot2 Key Features Transparency: Fully open-source code enables independent verification and peer review Flexibility: Modular design simplifies adding new food-hazard pairs, modifying process stages, or incorporating new data Reproducibility: Fixed random seed ensures identical results across runs Advanced analytics: Built-in uncertainty analysis and customizable visualization functions Validation The model successfully replicated results from Chen et al. (2013), with outputs closely matching published estimates (e.g., 63.05 vs. 63.5 annual DALYs for Salmonella in peanut butter). Applications This tool supports evidence-based food safety decision-making by enabling users to compare risks across food-hazard pairs, evaluate intervention effectiveness, and communicate uncertainty in risk estimates. The model is freely available for researchers, regulators, and industry professionals.我们基于R编程语言，开发了FDA-iRISK比较风险评估框架的开源实现。该模型可针对食源性危害开展定量微生物风险评估（quantitative microbial risk assessment, QMRA），通过模拟病原体从加工到消费的全流程行为，最终以伤残调整生命年（disability-adjusted life years, DALYs）估算公共健康负担。 ### 模型结构 该模型遵循模块化、分阶段的流程： 1. 初始条件：用户自定义供应链起点处的病原体流行率与浓度分布 2. 加工阶段：一系列代表各环节的序贯节点，包括分切、储存（病原体增殖或灭活）、处理等，每个节点均通过数学方法定义对病原体流行率与浓度的影响 3. 暴露评估：采用蒙特卡洛模拟，将最终病原体浓度与食用分量分布相结合，生成摄入剂量分布 4. 危害表征：针对特定病原体的剂量-反应模型（沙门氏菌（Salmonella）采用Beta-Poisson模型，李斯特菌（Listeria）采用指数模型），可将剂量转化为患病概率 5. 风险表征：将每份食用量的患病风险乘以年度食用频次与伤残调整生命年权重，以估算年度总健康负担 ### 实现方式 本模型基于R语言开发，使用mc2d包实现多变量蒙特卡洛模拟，可区分变异与不确定性。所有代码以可执行R Markdown文件形式提供，内置以下功能： - 三项经过验证的案例研究（花生酱中的沙门氏菌、软奶酪与哈密瓜中的李斯特菌） - 干预场景测试（如污染削减、温度控制等） - 剂量-反应参数的不确定性分析 - 基于ggplot2的增强型可视化功能 ### 核心特性 1. 透明度高：完全开源的代码支持独立验证与同行评议 2. 灵活性强：模块化设计简化了新增食源-危害配对、修改加工阶段或引入新数据的流程 3. 可复现性：固定随机种子可确保每次运行结果完全一致 4. 高级分析功能：内置不确定性分析与可自定义的可视化函数 ### 验证结果 本模型成功复现了Chen等人（2013）的研究结果，输出结果与已发表的估算值高度吻合（例如花生酱中沙门氏菌的年度DALYs估算值为63.05，与发表值63.5基本一致）。 ### 应用场景 本工具可帮助用户对比不同食源-危害配对的风险、评估干预措施的有效性，并可视化风险估算中的不确定性，从而支持基于证据的食品安全决策制定。本模型面向研究人员、监管机构与行业从业者免费开放。