MLA-C01 Practice Question 52: PHI Detection Clinical Notes
AWS Certified Machine Learning Engineer – Associate (MLA-C01) · Test 6 · question 2 of 10
A hospital network wants to train a model that suggests billing codes from free-text clinical notes. The notes are already stored as UTF-8 text files in Amazon S3. Before the notes enter the training pipeline, every piece of protected health information (patient names, ages, dates, addresses, and medical record numbers) must be located in the text so that a script can mask it.
Which AWS service should the pipeline call to locate the PHI?
Show answer and explanation
Answer:
Amazon Comprehend Medical extracts information from unstructured clinical text, and its DetectPHI operation is the one to use when you only need the protected health information. It returns each PHI entity, such as a name, age, date, address, or ID, with its type, a confidence score, and its begin and end offsets in the text. A masking script can use those offsets directly. AWS recommends adding human review or other checks for strict compliance cases (option A).
- B is wrong because Textract extracts text, forms, and tables from scanned documents and images. These notes are already text, and Textract does not recognize which words are PHI. - C is wrong because Transcribe Medical converts medical *speech* to text. There is no audio here to transcribe. - D is wrong because Kendra is an enterprise search service. It returns passages for a query, not a complete, entity-level list of every PHI mention in a document.
Reference: Detect PHI - Amazon Comprehend Medical