Data Challenge: Implementing change-point detection algorithms across time-series resistance

Expression of Interest – AMR Data Challenge 2025
We are excited to express our interest in participating in the 2025 AMR Data Challenge, building on our 2024 ResistAI platform. By leveraging a robust data foundation—including Pfizer ATLAS, Paratek KEYSTONE, and EARS surveillance datasets—we aim to integrate high-impact innovations that transform antimicrobial resistance (AMR) data into actionable insights across clinical, stewardship, public health, and health system levels.
1. Helping Improve Patient Outcomes
We will integrate Explainable AI (XAI) into our predictive models using SHAP values, enabling clinicians to understand the specific factors—such as geography, age, or drug usage—that drive a resistance prediction. This transparency empowers evidence-based prescribing decisions, reduces treatment failure, and tailors care to patient subgroups at greatest risk of resistance.
2. Strengthening Antimicrobial Stewardship
Our enhanced platform will include change-point detection algorithms (via Python’s ruptures library) to identify sudden shifts in resistance trends. These early alerts allow hospitals and stewardship teams to adjust formularies or prescribing behavior proactively—before resistance becomes entrenched.
3. Informing Public Health Practice
By stratifying resistance forecasts by region, age group, and pathogen, and highlighting temporal changes, we provide public health stakeholders with timely insights into emerging AMR threats. These tools can support surveillance programs, prioritize interventions, and align with WHO’s goals for early-warning AMR systems.
4. Strengthening Health Systems
Our user-friendly, web-based platform democratizes access to advanced analytics, even in low-resource settings. Scalable, interpretable, and responsive, it offers health systems a powerful decision-support tool that integrates forecasting, explainability, and surveillance into one accessible interface
Through these innovations, we aim to not just predict AMR—but to help understand, act, and intervene in meaningful ways. We look forward to contributing significantly to the 2025 challenge and supporting global efforts to combat AMR.

参赛意向——2025年抗菌药物耐药性（Antimicrobial Resistance, AMR）数据挑战赛

我们十分荣幸地表达参与2025年AMR数据挑战赛的意向，相关工作将基于我方2024年的ResistAI平台开展。依托包括辉瑞ATLAS数据库、Paratek KEYSTONE数据库以及EARS监测数据集在内的坚实数据基础，我们计划整合高影响力创新技术，将AMR数据转化为可落地的洞见，覆盖临床诊疗、抗菌药物管理、公共卫生以及医疗系统等多个维度。

1. 助力改善患者预后
我们将借助SHAP值将可解释人工智能（Explainable AI, XAI）融入预测模型，使临床医师能够明晰驱动耐药性预测的具体因素，例如地域分布、年龄层或药物使用情况。这种透明度可为循证处方决策提供支撑，降低治疗失败风险，并为耐药性高风险患者亚群定制个体化诊疗方案。

2. 强化抗菌药物管理
升级后的平台将集成变点检测算法（基于Python的ruptures库实现），以识别耐药性趋势的突发变化。此类早期预警可帮助医疗机构及抗菌药物管理团队在耐药性根深蒂固之前，主动调整药品目录或处方行为。

3. 支撑公共卫生实践
通过按地域、年龄组及病原体类别对耐药性预测结果进行分层，并展示时间维度上的变化趋势，我们可为公共卫生相关方提供针对新发AMR威胁的及时洞察。这些工具可助力监测项目、优先部署干预措施，并契合世界卫生组织（WHO）构建AMR早期预警系统的目标。

4. 赋能医疗系统建设
我们打造的易用型网页端平台将高级分析技术的应用门槛降低，即使在资源匮乏的环境中也可使用。该平台具备可扩展性、可解释性与响应性，能够为医疗系统提供一款集成预测、可解释性分析及监测功能的一体化决策支持工具，操作界面便捷易用。

通过上述创新举措，我们的目标不仅是预测AMR，更是以有意义的方式助力理解、应对并干预该问题。我们期待为2025年挑战赛作出实质性贡献，并支持全球AMR防控工作。