Study of micro-signals: proposed analysis methodology based on data from the Lille Poison Control and Toxicovigilance Center

This project aimed to enhance the Lille Poison Control and Toxicovigilance Center health surveillance and proactive response to intoxication-related issues by integrating a micro-signal analysis methodology. We developed a methodology employing Poisson distribution and historical limits to identify unusual signals marked by significant variations, often obscured by case age, continuous information flow (due to the 24/7 operational nature of the service), or team rotations. A severity score was created to prioritize micro-signals based on case seriousness to facilitate rapid and appropriate interventions. Additionally, a dashboard was developed to visualize detected micro-signals as a heat map, improving accessibility and decision-making for poison center professionals. Testing the methodology with historical Lille Poison Control and Toxicovigilance Center data demonstrated its viability and the medical relevance of the severity score calculations. It highlighted the effect of national measures on the use of an anxiolytic that had not been detected by the teams at the time. In addition, when the values returned by the severity score are interpreted by class of substance, rather than individually, similar orders of magnitude are observed (e.g., mushrooms, anxiolytics in our case study). Our approach shows potential for improving patient care and responsiveness in toxicovigilance, particularly within the French Lille Poison Control and Toxicovigilance Center network. One limitation of the current dashboard is that it only carries out analysis by product. Future enhancements include the integration of a thesaurus which will also allow analysis by class. This study integrated a micro-signal analysis method into a health surveillance system to detect hidden trends in poison center data, improving emergency response and substance management.

本研究旨在通过集成微信号（micro-signal）分析方法，提升里尔毒物控制与毒理警戒中心（Lille Poison Control and Toxicovigilance Center）的健康监测能力，以及对中毒相关问题的主动响应水平。我们开发了一套基于泊松分布（Poisson distribution）与历史限值（historical limits）的分析方法，用于识别存在显著变异的异常信号——这类信号常因病例时效、服务全天候运转带来的持续信息流，或团队轮换而被掩盖。我们构建了严重程度评分（severity score）体系，可根据病例严重程度对微信号进行优先级排序，以助力快速且恰当的干预措施。此外，我们开发了可视化仪表盘（dashboard），可将检测到的微信号以热图形式呈现，提升了毒物中心专业人员的信息获取效率与决策能力。利用里尔毒物控制与毒理警戒中心的历史数据对该方法进行测试，结果验证了其可行性，同时证实了严重程度评分计算的医学相关性。该方法还揭示了当时团队未及时察觉的国家政策对某抗焦虑药物使用的影响。此外，若按物质类别而非单个物质对严重程度评分的输出值进行解读，可观察到相近的数量级（如本案例研究中的蕈类、抗焦虑药物）。本研究方法有望提升毒理警戒（toxicovigilance）领域的患者护理质量与响应效率，尤其适用于法国里尔毒物控制与毒理警戒中心网络。现有可视化仪表盘的局限性在于仅支持按产品开展分析，未来的优化方向将包括集成同义词典，以实现按物质类别进行分析。本研究将微信号分析方法集成至健康监测系统中，用于挖掘毒物中心数据中的隐藏趋势，从而提升应急响应与物质管理水平。