Cohort selection for clinical trials using multiple instance learning

摘要: Abstract Identifying patients eligible for clinical trials using electronic health records (EHRs) is a challenging task usually requiring comprehensive analysis of information stored in multiple EHRs patient. The goal this study to investigate different methods and their effectiveness identifying that meet specific eligibility selection criteria based on patients’ longitudinal records. An unstructured dataset released by the n2c2 cohort track was used, each which included 2–5 manually annotated thirteen pre-defined criteria. Unlike other studies, we formulated problem as instance learning (MIL) compared performance with rule-based single instance-based classifiers. Our official best run achieved an average micro-F score 0.8765 ranked one top ten results track. Further experiments demonstrated MIL-based classifiers consistently yield better than single-instance counterparts require overall comprehension distributed among all patient’s EHRs. Rule-based approaches exhibited don’t consideration several factors across This patient can be MIL problem. exhibit supplement provide comparison approaches.