The Benefit of Standardized Clinical Statements in the Fight Against COVID-19

How the Health Community Can Prepare for the Next Pandemic

The Current State of COVID-19 Data Collection

The COVID-19 pandemic has devastated communities across the United States. Though vaccine distribution has begun and treatments continue to improve, the country is still lacking one major tool that will help contain this virus: accurate and reliable data. The past eight months have shown COVID-19 data is not only essential for clinicians seeking to manage and care for their patients, but also for researchers who seek to better understand the behavior of the virus in order to develop strategies for containing and treating COVID-19. Unfortunately, much of the data currently collected on COVID-19 patients is ambiguous and unreliable due to variation in the format of clinical statements.

Encoded clinical statements are the primary source of data in health care, recording key patient information gathered by providers in the delivery of care. They are used by pharmacists, physicians, nurses, and researchers, all of whom need to understand the details of a patient’s condition. Because clinical statements are of interest to several different parties, variation in representation can cause communication errors that jeopardize the health of patients.

The danger of this variation in clinical statements is clearly illustrated when physicians see a patient with sepsis caused by COVID-19 (Figure 1). Electronic Health Record (EHR) systems may contain several different codes and classifications for representing the statement, “Sepsis due to COVID-19.” As a result, one clinician may record this statement using ICD-10 as “A41.89: other specified sepsis,” while another clinician records the same statement using SNOMED as, “870588003: Sepsis due to disease caused by COVID-19.” Though this difference may seem trivial, the discrepancy has major implications for the patient’s health. Using the wrong classification code may fail to trigger relevant Clinical Decision Support (CDS) alerts, potentially reducing the quality of care that the patient receives. In this case, one code does not specify COVID-19 as the cause of sepsis, so clinical reminders specific to COVID-19 procedures might not be triggered. Consequently, a lapse in memory on the part of the provider could lead to the patient not receiving optimal treatment for COVID-19. This omission may also cause future providers to be unaware that the patient previously had COVID-19, and as a result could lead to an uninformed treatment decision.

Figure 1: Breakdown of risks associated with clinical statement variation.

Additionally, researchers or clinicians that use this data to develop better future treatments may fail to capture key information from clinical statements that are not represented accurately. Inconsistent clinical statements prevent the medical community from advancing knowledge about COVID-19 and obstruct the development of improved treatments. If, however, clinical statements were standardized, discrepancies of representation would be eliminated and the quality of the data those statements produce would be dramatically improved. In short, patient care and medical research powered by reliable data becomes more dependable and effective. To achieve this goal, the Solor project developed Analysis Normal Form (ANF).

What is Analysis Normal Form (ANF)?

ANF is a standardized and machine-readable way of representing clinical statements. ANF is designed to be easily interpretable and reproducible by medical professionals. It is the Solor project’s initiative to improve clinical statements using design principles tailored to make clinical information more understandable, reproducible, and useful.[1] The ANF template is designed to represent all types of clinical information, so that any clinical statement can be adapted to the ANF format by simply populating the fields of the template. The consistent structure of ANF statements allows health care professionals to ensure they are communicating effectively, while also maintaining the integrity of patient data as it changes hands. Because of the emphasis on understandability, reproducibility, and usability, ANF empowers researchers seeking to study aggregated health data, as it is designed to facilitate accurate data capture and analysis. Most importantly, ANF enables safe and reliable care through more accurate CDS by ensuring clinical statements that trigger CDS are all represented consistently. 

ANF and COVID-19

Throughout the pandemic, unreliable and inconsistent data has been a persistent problem. ANF has immense potential to help the medical community fight COVID-19 and future health crises by enabling more accurate clinical data and syndromic surveillance. By utilizing standardized clinical statements designed to facilitate data analysis, ANF can accelerate and enhance current intervention and research efforts underway to better understand and treat the virus. The COVID-19 pandemic has accelerated the need for tools like ANF to enhance the power of accurate health data as we look towards a future of digitized medicine.

Though challenging in its current state, representing the development of sepsis in a COVID-19 patient with ANF is simple. The diagram below (Figure 2) demonstrates how the fields of an ANF statement would be populated to account for this scenario. Deconstructing the statement in this way facilitates data analysis while also making the statement clear and unambiguous. ANF statements allow for the attachment of “associated statements” that add relevant clinical information to the primary statement. In the representation below (Figure 2), the presence of COVID-19 has been recorded as an associated statement, linked to the primary statement by the semantic, “due to”. This structure eliminates the ambiguity of using a single code (as shown in Figure 1), ensuring appropriate CDS will be triggered and important data is not left out. ANF also allows the attachment of multiple codes that can be tailored for certain systems. Using this format, a researcher or clinician can easily interpret the description of a patient who developed sepsis due to COVID-19.

Figure 2: Diagram of ANF statement structure with fields populated to describe a patient with Sepsis due to COVID-19.

Another key benefit of ANF statements is that they are designed to allow a computer to easily extract and aggregate key information. In the statement above, the presence of both sepsis and COVID-19 are recorded in separate binary segments, so that a computer can accurately identify them and incorporate this statement into analysis of either or both conditions. If instead a clinician were to use the code “A41.89-other specified sepsis” and record the presence of COVID-19 as a note, a computer processing this statement might miss the note, precluding this patient from being included in analysis on COVID-19. Furthermore, there are myriad different ways that clinicians could write out “Sepsis due to COVID-19,” making it more difficult to program a computer to consistently pick up on the key information. Accurate data analysis on COVID-19 and related conditions like sepsis is essential to better understanding the virus and designing optimal treatments and interventions. ANF’s machine readability can help us strengthen that analysis, boosting our efforts to address the pandemic.

In health care, accurate data can be a matter of life or death. Clinical statements are the primary source of data in health care and must be represented accurately to protect patients and advance medical research. The COVID-19 pandemic has exposed significant challenges and weaknesses in the delivery of health care but has also demonstrated the possibilities that data standardization has to offer when faced with a crisis. By standardizing clinical statements in a precise, easily interpretable, and machine-readable format, ANF can be leveraged to produce higher quality care and safer patient outcomes.

[1] Spackman KA, Reynoso G. Examining SNOMED from the perspective of Formal Ontological Principles:

Some Preliminary Analysis and Observations. Proc. KR-MED, 2004: 72-80.