Skip to main content

Privacera Documentation

Table of Contents

Preview: Scan Electronic Health Records with NER Model

Based on Natural Language Processing (NLP), the Electronic Health Record Named Entity Recognition (EHR NER) model detects entities like medical test, medical problem and medical treatment while scanning documents. This model is intended only for use with unstructured text files or unstructured fields present in structured files as it works with contextual information present in the text surrounding the target tags.

If a structured file has fields with long sentences for which predictions are needed via NLP, you can set the UNSTRUCTURED_FIELD_IN_STRUCTURED_FILE_ENABLED parameter to true. However, setting this parameter to true might result in reduced speed for classification. The time required for classification depends on the number of unstructured field records with five or more words.

Supported tags

EHR_NER_ML_MODEL supports the following tags:

  • MEDICAL_TEST

  • MEDICAL_PROBLEM

  • MEDICAL_TREATMENT

Tags

By default, tags supported by EHR_NER_ML_MODEL are not present on the portal UI. If you want scans to detect and showcase these tags on the user portal, you need to add them explicitly under the Tags tab. For more information, see Tags

  • MEDICAL_TEST

  • MEDICAL_PROBLEM

  • MEDICAL_TREATMENT

Table 64. Parameters

Parameter

Data Type

Default

Description

UNSTRUCTURED_FIELD_IN_STRUCTURED_FILE_ENABLED

Boolean

False

Setting this parameter to true enables scanning of unstructured fields or columns within structured files.

NLP_WORD_PROXIMITY_LENGTH

Integer

10

This parameter sets the total length of words to be considered for contextual information around PII information.

NLP_LOG_LEVEL

String

INFO

This parameter sets the log level in the background process used for NLP.



.