OMOP Measurements and Devices

Bridge2AI Standards Module

Jared Houghtaling

Software Developer/Analyst

Tufts Clinical and Translational Science Institute (CTSI)

Office Hours 18 May 2024

An overview of conventions for the MEASUREMENT and DEVICE domains

11 September 2025

Content

• Intro to the MEASUREMENT table
• Domain constraints and mapping approaches
• Importance of units in OMOP CDM
• A note about Observations
• Intro to the DEVICE table

​

​

​

​

Structure of OMOP

OMOP Vocab Content

OMOP Vocab Statistics:

• 140+ vocabularies
• 40+ domains
• 10M+ concepts
• ~4M standard concepts
• ~1M classification concepts
• ~80M Concept relationships
• ~100M Ancestral relations
• ~3M Concept synonyms

OMOP Vocab Content

OMOP Vocab Statistics:

• 140+ vocabularies
• 40+ domains
• 10M+ concepts
• ~4M standard concepts
• ~1M classification concepts
• ~80M Concept relationships
• ~100M Ancestral relations
• ~3M Concept synonyms

OMOP Vocab Content

Relevant Domains for MEASUREMENT Events:

• MEASUREMENT
• MEAS VALUE
• UNIT
• OPERATOR
• TYPE CONCEPT

​

* Note that Measurements have the most domain-constrained concept_id fields of any event type

​

Measurements

(MOST) LOINC Codes are Standard in OMOP

• Look-up LOINC code (e.g. 28803-5) in concept_code field of CONCEPT table; make sure you filter on LOINC vocab & MEASUREMENT domain!
• Join associated concept_id for LOINC code (e.g. 3005795) on CONCEPT_RELATIONSHIP table’s concept_id_1, where relationship_id = ‘Maps to’
• If code is standard, will map to itself. If non-standard, will map to replacement
• Join associated concept_id_2 from CONCEPT_RELATIONSHIP back on CONCEPT to get its details and ensure it is STANDARD
• Place this concept_id in your measurement_concept_id field (see demo)

Measurement Values & Operators

IF NUMERIC (or Numeric along with <, >, =, etc.)

• (Optional) Verify that LOINC code does not have expected categorical responses for number (e.g. Has Answer)
• Extract numeric-only portion from value and place in value_as_number
• If relevant, extract operator and MAP to operator_concept_id (+ source_value)

​

IF CATEGORICAL (e.g. “POSITIVE”)

• Map categorical value to concept_id in the MEAS VALUE domain, favoring LOINC Answers where duplicated
• E.g. “POSITIVE” = 45884084
• Note: while OBSERVATION has a ‘value_as_string’ field where you can place the literal value, MEASUREMENT does not.
• Note: be careful with value_source_value fields, as they can contain free text with MRNs and dates

Measurement Range

Range Low and Range High are useful in contextualizing value

• Be sure to check for consistent units
• If unit mismatch, update the range values to match meas value unit, not the other way around
• Non-required fields, and some measurements only have one side of the range (not a problem)

​

Introduction to Units in OMOP CDM

• Units provide context for the combination of measurement and value
• Important within a single site
• VERY IMPORTANT when investigating measurements across a consortium
• LOINC codes (almost always) provide a combination of laboratory test and associated unit
• Many EHR systems, EPIC included, do not enforce the embedded units in laboratory measurements recorded in LOINC

Challenges with Units

1. The unit value was simply omitted from a laboratory measurement
• Units are rarely required to accompany a measurement recording during input
2. The unit provided for a given measurement is nonsensical
• The technician selects ‘K’ (Kelvin) for a concentration measurement reported in K/L (thousand/liter).
3. A single measurement has multiple unit possibilities depending on measurement device or source system
• Body weight taken in the ICU as KG, in other dept’s as LBS

​

​

​

​

Simplest Option: DQD Plausible Units

Simplest Option: DQD Plausible Units

• At Tufts, we’ve converted the DQD CSV file to an expanded table of measurement concept + unit concept
• Filtering on possible_units = 1 allows us to infer units for those measurements (~20%) where only 1 unit is plausible

ONCOLOGY Extension Columns

1. Measurement_event_id and meas_event_field_concept_id

​

​

​

​

​

​

Devices

Devices

OMOP Vocab Content

Relevant Domains for DEVICE Events:

• DEVICE
• UNIT
• TYPE CONCEPT

​

* Note that Devices have other fields that capture device-specific info (unique_device_id, production_id) WARNING – these fields are free text and should be scrutinized for any potential PHI.

​

Devices

(MOST) Standard Devices are in SNOMED (+ other regional vocabularies)

• Look-up SNOMED code (e.g. 28803-5) in concept_code field of CONCEPT table; make sure you filter DEVICE domain!
• Join associated concept_id for SNOMED code (e.g. 467994009) on CONCEPT_RELATIONSHIP table’s concept_id_1, where relationship_id = ‘Maps to’
• If code is standard, will map to itself. If non-standard, will map to replacement
• Join associated concept_id_2 from CONCEPT_RELATIONSHIP back on CONCEPT to get its details and ensure it is STANDARD
• Place this concept_id in your device_concept_id field

Device Quanity and Units� (most relevant for transfusions)

• If there is a record of device exposure in the source but no quantity value, then set to 1.
• There is no standardization requirement for units associated with DEVICE_CONCEPT_IDs, however, it is the responsibility of the ETL to choose the most plausible unit. If the source unit is NULL (applicable to cases when there’s no numerical value or when it doesn’t require a unit), keep unit_concept_id NULL as well.
• Using the blood transfusion example, blood transfusion is represented by the DEVICE_CONCEPT_ID and the unit (mL) would be housed in the UNIT_SOURCE_VALUE and mapped to a standard concept in the unit domain.

Indirect Mapping – Discussion

• If you have flowsheet measurements related to a ventilator, should you also populate a row for the ventilator as a device?