Representing Negation

Return to: Patient Care

Goto: FHIR Clinical Resources - PCWG

Goto: Clinicians on FHIR

Background

There are a number of clinical concepts (and hence clinical FHIR resources) to which negation are commonly applied. Examples include

• Allergy/Intolerance
• Observation
• Condition
• Procedure
• Medication

There does not appear to be a consistent way in applying negation to these (and other similar) resources.

PCWG, FHIR and OO agree that a consistent approach to this matter is highly desirable to ensure interoperability. These groups agree to collaborate in order to determine an agreed approach to achieve this goal.

Contributors

• Stephen Chu
• Russ Leftwich
• Rob Hausam
• Elaine Ayres
• Laura Heermann Langford
• Craig Parker
• Emma Jones
• Michelle Miller
• Lloyd McKenzie
• ...

Discussions

Altanta WGM October 2014

• Discussions on "negation" representation (Note - "negation" will be changed to some term more appropriate to avoid confusion with HL7 v3 implementation of negation)

- vigorous discussions on how negation should be represented in FHIR clinical resources including condition, procedure,, observation, medication, allergy/intolerance

- general agreement was reached to represent negation by a flag with a set of values:

~ yes

~ no

~ unknown

Example: representing patient did not have concussion

~ Condition = concussion

~ "flag" = no

- also agreement on restricting double negatives causing confusion

Example:

~ Condition = no concussion

~ "flag" = yes ("no" value should not be used)

- In case of precoordinated values, they should be represented as two clinical statements

Example: concussion with no loss of consciousness

~ this should be two clinical statements

- Note:

~ There is still a need to identify a set of standard terminology-based value to support negation

~ Examples:

- No known allergy

- No known medication

- No known procedure

Examples/Use Cases

EMS

A few years ago, the EMS Patient Care Report faced this issue. Finding the discussion nowhere near completion, we adopted the following approach:

• Answering a question about a collection of things that might not exist (e.g., allergies) results in answers of two different kinds: lists of things (e.g,. allergies) and statements about whether there are any allergies. We found the RIM semantics provided one way to address these needs, but they did so in a way that is abstract and prone to misprision.
• In order to make the semantics clear and unambiguous, we adopted the pattern of distinguishing two separate questions. In the case of EMS drug allergies,
  • The Boolean 67794-8 Does the patient have any medication allergies NEMSIS (yes, no, unknown), and, if yes,
  • The collection of RxNorm values 67850-8 Medication allergies - RxNorm code NEMSIS
• (We found the use of the null flavor in an observation value to be less ambiguous than its use on a class.)
• We also suggested implementers follow the CDA/RIM guidance for using negation indicators on nugatory allergy classes instantiated for the purpose of asserting their absence, in the interest of mutual comprehensibility with any organization the followed that guidance. We would be delighted to learn that this is no longer necessary.