Examples of How LOINC Fails Nursing

The way LOINC encodes medical data does not support nursing well. The two mental models are only partially compatible.

In particular, patient assessments are something nurses do many times on every shift. The data that is thus generated needs to be captured in the EHR. To store a nurse's observation in an EHR, the observation must be encoded. Often, there is no appropriate encoding mechanism in LOINC. Here are some examples.

• Pack Years. In some cases, observations commonly made by nurses are completely absent from LOINC. One such example regards smoking, an important health factor often observed and recorded by nurses. One common measure of smoking history is Pack Years, that is, for how many years has the patient smoked a pack of cigarettes a day? In LOINC, there is no code under which this datum can be captured.
• Pain Goal. Another example of a common nursing assessment that is absent from LOINC is Pain Goal. Pain is often termed the "fifth vital sign" and is commonly assessed by nurses on a 0-10 scale; the patient is asked to name a number in that scale. Often, the patient is also asked to name a number in the scale that is the pain goal, sometimes described as what the patient can tolerate. For example, they might say, "My pain is at an 8. I could live with a 4." LOINC has no way to record the 4.
• Polypharmacy. In other cases, the datum is present in LOINC but the way it is coded does not match the way nurses conceive of it. For example, the Norton Pressure Ulcer Risk Assessment (at how much risk is the patient for developing a bedsore?) includes a polypharmacy question: is the patient taking five or more medications? The answer is yes or no. LOINC offers code=28192-3, name=Polypharmacy [CCC], category=observation.survey. The possible values that can be stored under this code are Improved (LA65-8), Deteriorated (LA66-6), or Stabilized (LA6635-2). "Improved" is sort of like "no" and "deteriorated" is sort of like "yes" but not exactly. As a result, we don't know how to encode the Norton Assessment.
• Sputum color. Sometimes, LOINC makes assumptions about the way data are collected that do not apply to nursing. For example, sputum is a specimen often but not exclusively analysed or described by clinical laboratories. Various aspects of the sputum produced by patients with respiratory ailments are also observed and recorded by nurses. The color of the sputum is encoded in LOINC as code=86243-3, name=Color of Sputum, category=observation.lab. That means that sputum color is a datum produced by the clinical laboratory whereas it was in fact produced by a clinician (the nurse) and should thus belong to category observation.clinical. Alas, LOINC has no such code.
• Urinary Stent Location. Nurses document whether a stent(s) is present and if so, on which side (L or R). This is important for nursing care but LOINC doesn't have a way to code it.
• Gosnell Pressure Ulcer Risk Scale. One of the metrics nurses use is the Gosnell Scale to help determine whether a patient is at risk for developing pressure ulcers. It assesses six important patient characteristics: mental status, continence, mobility, activity, nutrition, and skin status. From these, it calculates a score that indicates risk for developing pressure ulcers. LOINC doesn't have codes for the Gosnell scale so these data have to be squeezed into other fields which makes them less accurate. Also, there is no LOINC data field for the calculated Gosnell score. Perhaps all these things could be added as a LOINC panel.
• Hematocrit Low. Sometimes LOINC uses a different metric for a data element. For example, we have a yes/no field for "hematocrit low" whereas LOINC has +/- for hematocrit test results. In cases like that, we must either translate our datum format to theirs before uploading it, or modify ours to match theirs. We are reluctant to do that because our data is designed for the way nurses think about it whereas LOINC is derived from a clinical lab context; doing it their way would force our nurse users to think in unfamiliar ways.

What can be done to solve these problems? What must we do to make EHRs more nursing-supportive? I do have some ideas about this and look forward to discussions with interested parties, especially the Regenstrief folks who are the keepers of the LOINC standard.