引導式測驗 · 中等 等級
合成病患資料測驗
測試合成資料定義、隱私風險、特定任務的實用性、來源以及醫療保健臨床驗證的限制。
問題 1 的 8
Which description matches synthetic patient data?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- Which description matches synthetic patient data?
- A diagnosis predicted by a clinical model
- Generated records that mimic selected features of health data
- A de-identified copy guaranteed to have no privacy risk
- A real chart with names removed
- Does synthetic data automatically guarantee privacy?
- Yes, when records look realistic
- Yes, when names are absent
- No privacy test can be meaningful
- No; a generator may memorize or reveal information from source data
- Why is utility task-specific?
- Utility means only that files open successfully
- Clinical validation never requires real data
- A dataset useful for software testing may not support clinical model validation
- Every generator preserves all clinically relevant patterns
- According to the Synthea paper, which output does its patient generator produce?
- It produces a synthetic billing file from one hospital’s real claims
- It produces simulated patient lifespans and structured synthetic health records
- It replaces EHR data with verified patient records
- It generates only de-identified copies of actual patient charts
- Which validation plan checks both synthetic-data fidelity and downstream utility?
- Compare only average feature values and ignore rare events
- Check relationships, missingness, rare events, subgroup fidelity, and task performance
- Measure downstream accuracy on training records alone
- Check visual realism and treat that as proof of utility
- Which clinical conclusion cannot be established by synthetic data alone?
- That the model performs safely and effectively on real patients
- That generated data can help test data pipelines
- That synthetic examples can support software development
- That utility should be evaluated for the intended task
- Why document generator and source data?
- To prove all records are clinically equivalent to real people
- To certify that privacy assessment is unnecessary
- To establish subgroup representativeness without evaluation
- To understand provenance, transformations, and limitations
- Which workflow best prevents a synthetic record from being mistaken for a real patient history?
- Mark only the README while leaving individual records unmarked
- Describe generated records as de-identified real charts
- Label records as synthetic wherever they are presented or used
- Remove generator metadata and mix records into live charts without notice