Early Detection of Fabry Disease (FD): Using Real-World Data for the Development of Advanced Natural Language Processing Methods: Retrospective Database Analysis
试验速览
- 阶段
- 不适用
- 状态
- 已完成
- 发起方
- 入组人数
- 50
- 试验地点
- 3
- 主要终点
- Percentage of Participants With Positive Predictive Value (PPV) at Different Cut-off Values (top 10, 20, 50, 100 and 200)
研究概览
简要总结
The main aim of this study is early detection of FD using real-world data for the development of advanced natural language processing methods and to develop a predictive algorithm and to measure the performance of the algorithm in identifying participants with FD.
This study is about using data from hospital Electronic Health Record database from the last 10 years to describe the ranking of participants with FD using multilevel likelihood ratios and to validate the algorithm using positive controls. No investigational medicinal product or device will be tested in this study. Hospital electronic health record data will be analyzed for a period of up to 6 months.
研究设计
- 研究类型
- Observational
- 观察模型
- Other
- 时间视角
- Retrospective
入排标准
- 性别
- All
- 接受健康志愿者
- 否
入选标准
- •In-patient or out-patient datasets of the participating hospital in the last 10 years
- •Participants at any age Positive controls: a subset of all participant hospital records that includes the participants with confirmed FD.
排除标准
- 未提供
研究组 & 干预措施
Retrospective Database Analysis
Data from patient's hospital records of the last 10 years will be collected/extracted retrospectively using epidemiological methods to test the forecasting power of the algorithm.
干预措施: No intervention (Other)
结局指标
主要结局
Percentage of Participants With Positive Predictive Value (PPV) at Different Cut-off Values (top 10, 20, 50, 100 and 200)
时间窗: Up to End of the study (approximately 6 months)
PPV is a clinically relevant statistical measure that indicates how likely participants that screen positive are to be affected by the condition assessed. Thus, the PPV can be considered as the percentage of participants which are identified as FD candidates by the ranking algorithm who are indeed FD participants. As FD predictive algorithm, we will use (multilevel) likelihood ratios (LRs) as this method permits a good use of clinical test results to establish diagnoses for the individual participant. LR is calculated, defined as the probability of a patient who has FD to present with this feature divided by the probability of a participant who not has FD to present with the feature: Likelihood ratio= features the participant/Fabry divided by features the participant/not Fabry. Positive predictive value of the algorithm at several cutoffs (top 10, top 20, top 50, top 100, top 200) will be reported.
次要结局
- Percentage of Participants Based on Ranking With Known FD Using Multilevel Likelihood Ratios For Algorithm Validation Purposes(Up to End of the study (approximately 6 months))