Analysis Normal Form
There currently exists both a potential for a lack of consistency with representing clinical statements with current detailed clinical modeling efforts and variation in how the data is entered into information
systems by end-users. This has a direct impact on patient safety if a clinical statement is
recorded and displayed differently across the continuum of care. Vendors may compete on usability which may result in proprietary data models or they may use clinical statements based on standards-based models (e.g. CIMI, openEHR archetypes).
Analysis Normal Form (ANF) is a logical model intended to represent a normalized view of aggregate clinical statements recorded during treatment for analysis, research, clinical decision support, and other purposes. At a high-level an ANF statement defines the topic (WHAT happened, was observed, requested, measured, asserted, etc.) and under what circumstances (HOW, WHY, WHEN, and with what RESULT). ANF requires an ability to classify the topic of a statement using standard terminology expressions.
The Topic is the central component of clinical statements capturing WHAT is being requested or WHAT was performed. A pre-coordinated or postcoordinated “procedure” concept is required to sufficiently capture the action, which is either requested or performed. The topic:
defines the action that is being requested, measured, or observed
has to be able to exist on its own and still retain the original intent and clarity of meaning
each clinical statement may only have one topic, but can be associated with other clinical statements
Performances may be observational performances, e.g. the observation of a clinical finding or disorder being present or absent. They can also be a procedure or intervention which has been performed on the subject of record in the past, e.g. “a procedure using a 12-lead electrocardiogram”. Performances can – but do not have to – include quantitative or qualitative results, e.g. “Measurement of temperature using forehead thermometer” or “Detected SARS coronavirus 2 RNA [Presence] in Respiratory specimen by NAA with probe detection”.
Each ANF Statement has one Circumstance which can describe HOW, WHY, WHEN, and with what RESULT a requested or performed action will be or was carried out. ANF promotes a normalized representation of observation or intervention results where all results are reduced to a "measure". This approach reduces data retrieval difficulties by eliminating the potential for multiple differing representations of the same clinical statement. For example, with coded results there are multiple potential methods to represent eye color that complicate data retrieval. The Topic could be a Finding refined by an Observable (Iris finding->Interprets = Color of iris) or a Finding with no refinement (Finding of color of iris). In both of these cases the Result would be a qualifier of Blue color. The ANF Statement would represent Eye color using the Blue iris Finding as the Topic and the Result would be Present, represented as interval.lowerBound=1, interval.upperBound=INF.
A Request for Action (Circumstance) clinical statement describes a request made by a clinician where the object of the request (e.g., lab test, medication order) may or may not be fulfilled by someone other than the clinician (e.g., lab technician, pharmacist) making the request. All detailed information about the request will be documented in this clinical statement, such as patient must fast for 12 hours before having a lipids blood test.
A Performance of Action (Circumstance) clinical statement is used to represent an action or observation that was performed or placed in a status other than being requested (e.g., on hold, cancelled) and specifies the result of intervention using both measure and coded status.
Insert example of topic vs circumstance
Currently there is no tooling available to assist with the transformation of clinical data/statements into analysis normal form. As ANF is a new Informational Ballot there are no integrations for ANF based data stores and VHA EHR/health IT systems. There is some documentation within the ANF informational ballot on best practices for implementing ANF modeling.
In order to develop ANF models you must have:
Terminologist – who understands and can work with the US core meaningful use standards (e.g., SNOMED CT, LOINC, RxNorm)
Clinical Informaticist – who understands the clinical area that is being automated through business processes (aka BPM+)
ANF modeler – who understands and can adhere all needed clinical models to the ANF specification