Findings or other clinical data collected about a patient during care.

Data Element
Observation Subject
Description

The patient, or group of patients, location, or device this Observation is about and into whose record the observation is placed. Note: Additional structures are needed to handle situations where the actual focus of the Observation is different from the subject (or a sample of, part, or region of the subject). For example, a measurement on a fetus that is placed in the mother's record. Data Type: Reference (Patient, Group, Device, or Location). This data element is typically a pointer into a record in another table/structure that contains more metadata about the subject. Note: This Data Element (Observation.subject) would typically point to a record/instance of the Patient Demographics Data Class, though Observations can be recorded for other “units of analysis” (such as a geographic area, group of subjects, etc). The exact mechanism for specifying this linkage is not prescribed, but the purpose of this Data Element is to establish that the ability to communicate “who the observation is about” must be supported.
Submitted By: Daniel Vreeman / RTI International
Data Element Information
Rationale for Separate Consideration To establish that the ability to communicate “who the observation is about” must be supported.
Use Case Description(s)
Use Case Description Clinical observations are an essential structure for recording many kinds of health information. A laboratory test result, vital sign measurement, blood type, pain scale rating, or recording of exercise activity type that a person engaged in (e.g. running, walking, swimming, etc.) are all considered observations.

Billions of clinical observations have been exchanged in the current health information exchange ecosystem. All of the prevailing exchange (e.g. messages, documents, APIs) and common data model (e.g. OMOP, PCORnet, Sentinel, etc) paradigms represent these core data elements (code, value, date/time, subject) pertaining to observations.

For many reasons, it would be ill-advised for USCDI to enumerate the tens of thousands of specific observation “variables” (eye color, gait speed, shoulder flexion range of motion, etc) as separate Data Elements. Rather, USCDI needs a common template structure for specifying observations and then mechanisms for creating subsets of key domains where there is value in doing. Here we have proposed the essential data elements. Using this common pattern of essential data elements, key domain subsets (like the existing labs and vital signs classes) can be defined using the shared template with further specified through bindings to appropriate terminology standards and/or additional data elements.
Estimate the breadth of applicability of the use case(s) for this data element Most health IT systems with more than a few clinical variables use an entity-attribute-value (or name value pair) model similar to this proposed approach for storing and exchanging observations.
Healthcare Aims
  • Improving patient experience of care (quality and/or satisfaction)
  • Improving the health of populations
  • Reducing the cost of care
  • Improving provider experience of care
Maturity of Use and Technical Specifications for Data Element
Applicable Standard(s) Vocabulary Standard:
No single vocabulary covers the content necessary for representing all aspects of observation reporting (observation identifiers, coded observation values, computable units of measure, etc). Yet, the collection of core vocabularies needed are sufficiently mature.

[Observation: Code] LOINC was specifically designed to fulfill the need for a freely available standard for identifying observations. More than 25 years later, it now contains a rich catalog of variables, answer lists, and the collections that contain them.

[Observation: Value.units] UCUM was specifically designed to fulfill the need for a freely available standard of computable units of measure.

[Observation: Value.code]
• If the observation represents a validated assessment instrument (e.g. survey) item and the value is a CodeableConcept, then Observation.value.code SHALL be drawn from LOINC.
• If the observation pertains to clinical genetics, then other terminologies or syntaxes may be used as appropriate, e.g. HGVS, ClinVar, etc.
• If the observation represents a human phenotype, then the Observation.value MAY be drawn from the Human Phenotype Ontology
• If the observation has a nominal or ordinal scale value and the Observation.value exists in SNOMED CT, then SNOMED CT MAY be used

Exchange and Analytic CDM specifications:
HL7 Version 2: Chapter 7
C-CDA: Results Section (entries required): 2.16.840.1.113883.10.20.22.2.3.1
FHIR: Observation
OMOP: Measurement, Observation
PCORnet CDM: LAB_RESULT_CM, PRO_CM, OBS_CLIN, OBS_GEN
Additional Specifications There are several implementation guides (e.g. for HL7 v2 and CDA) and FHIR profiles on Observation. Examples include those in the main FHIR spec (vital signs, genetics) and in US Core (smoking status) etc.
Current Use This data element has been used at scale between multiple different production environments to support the majority of anticipated stakeholders
Supporting Artifacts A few examples of documentation include:
Epic FHIR Documentation: https://fhir.epic.com/
Cerner FHIR Documentation: https://fhir.cerner.com/millennium/r4/diagnostic/observation/
Allscripts FHIR Documentation: https://developer.allscripts.com/FHIR?PageTitle=Resources
Extent of exchange 5 or more. This data element has been tested at scale between multiple different production environments to support the majority of anticipated stakeholders.
Supporting Artifacts Three examples among many:
• the INPC clinical repository contains billions of discrete observations from about 100 institutions. https://www.ihie.org/popcare/#careweb
• Apple Health uses the FHIR Observation resource for labs and vital signs to interface with many institutions: https://support.apple.com/en-us/HT208647
• participating PCORNet institutions store billions of discrete observations represented in the PCORnet CDM structure
Potential Challenges
Restrictions on Standardization (e.g. proprietary code) Some, but not all observations in operational health IT systems are linked to vocabulary standards. So local codes still persist. At present, mappings to vocabulary standards are much more common for observation identifiers than for coded observation values.
Restrictions on Use (e.g. licensing, user fees) Licensing for some reference terminology standards is available at no cost worldwide: LOINC, UCUM, HPO.

SNOMED CT is available at no cost for use within the US via an affiliate license available from NLM (https://www.nlm.nih.gov/healthit/snomedct/snomed_licensing.html).

The FHIR exchange specification is available without copyright (i.e. it is declared to be Public Domain). Other HL7 exchange specifications are available under the terms described here http://www.hl7.org/legal/ippolicy.cfm?ref=nav.
Privacy and Security Concerns Some observations may carry sensitive information. Appropriate protections for addressing privacy and security concerns should be addressed in the surrounding technical infrastructure and policies.
Estimate of Overall Burden Unknown.

Comment

Log in or register to post comments