Managing Your COVID-19 Data

How to encode accurate COVID-19 patient data to improve data analysis and decision support, transforming a healthcare organization into a learning organization.

The COVID-19 pandemic has shown how important timely, accurate, and detailed clinical data is so that providers can learn from data, share those learnings with others, and use those learnings to provide education, clinical care, and effectively manage resources. In essence, it allows providers to be a learning healthcare organization where science, informatics, incentives, and culture are aligned for continuous improvement and innovation, with best practices seamlessly embedded in the delivery process and new knowledge captured as an integral by-product of the delivery experience

“The use of accurate, real-time data to inform decision make is essential…In the ongoing COVID-19 pandemic there is an overwhelming amount of data, including many indicators that can be misleading if not considered correctly.”

A first step in becoming a learning healthcare organization is to capture data as an integral by-product of delivery. In our example, Mind Computing’s KNOWLEDGE web application shows how a health organization can encode data by integrating meaningful use data encoding standards, such as SNOMED, LOINC, and RxNorm. KNOWLEDGE enables a medical organization to extend those data encoding standards with concepts of local interest or of emerging importance, as described below. This integrated and extensible system can then be used to encode data capture forms, to collect and normalize data, and to analyze that data so that decisions about how to best deliver care can be encoded into defined processes.

Encoded Data Capture Forms

Healthcare organizations create electronic data capture forms to collect healthcare information. For example: to collect COVID-19 related healthcare information, a healthcare organization might create an electronic data capture form. These forms are encoded to enable analysis and decision support.

Most healthcare organizations use some combination of local terminology and meaningful use standards to encode the information captured in their electronic data capture forms. KNOWLEDGE imports the local terminology, integrate it with meaningful use standards, and distribute the mapped data in standard electronic formats to preserve the meaning of patient information when sharing patient data with healthcare partners. This integration maximizes the analytic value of the data collected.

“Improve the efficacy of clinical care by extending meaningful use standards.”

Healthcare organizations are being asked to provide the CDC with COVID-19 data.

To encode the response in an electronic data capture form– “subjective fever,” “yes,” “no,” and “unknown” these values must exist in the meaningful use standards or their local terminology. A search in the meaningful use standards does not return “subjective fever.” There are several solutions to add “subjective fever” to the meaningful use standard. Figure 1: Adding a new description-subjective fever in KNOWLEDGE, demonstrates adding a description– “subjective fever” to the existing SNOMED concept “Feels hot/feverish”. Adding this description is an example of extending the meaningful use standard SNOMED CT. The ability to extend meaningful use standards enables a healthcare organization to create content needed in the timeframe needed. It also maintains their local terminology the same way that organizations such as SNOMED develop and maintain their terminology systems. In addition, extending meaningful use standards enables an organization to share their extension with other healthcare partners to preserve the meaning of patients’ information.

Figure 1: Adding a new description-subjective fever in KNOWLEDGE

Typically, when a patient has been tested for COVID-19 they are asked about underlying health conditions, such as diabetes, high blood pressure, lung disease, cardiovascular disease, and/or severe obesity. This data is used to analyze the susceptibility and expected outcome of patient, track their outcomes, and to identify the best way to manage the delivery of care. It is important to encode these data elements consistently to ensure that the data collected is represented in a standardized way for decision support rules and patient care.

The following sections describe two example questions that are asked when screening for COVID-19, “Do you have a cough?” and “Do you have a fever?”

Do you have a cough?

Figure 2: Cough taxonomy. The cough taxonomy view shows the 37 children of “Cough” and the descendants of “Cough.” The children of “Cough” are the first level concepts beneath “Cough.” The 7 children of “Finding related to ability to cough” are descendants of “Cough” and “Finding related to ability to cough” has 7 children.

To encode “cough” in an electronic data entry form, a form developer would search for “cough” in the meaningful use standards. Figure 2 Cough taxonomy displays the taxonomy of “cough” using the KNOWLEDGE taxonomy view of the meaningful use standards. The cough taxonomy is reviewed to determine if all the concepts under “cough” are accurate and representative of the patient’s condition. The green and orange hexagons are the children of “cough”. Once “cough” and its children are validated, the concept “cough” will be used to constrain the values that are displayed to the provider in the electronic data entry form.

Figure 3: Fever and feels hot/feverish taxonomy. This figure shows that “Feels hot/feverish” has a lineage of “Observation of sensation” and “Fever.”

Do you have a fever?

To encode the terminology “fever” in the question “Do you have a fever?”, the form developer will search for “fever” in meaningful use standards. This search will return concepts located in two different hierarchies of the meaningful use standards. The concept “Fever” has parents in the “Body temperature” hierarchy and the “Feels hot/ feverish” concept parents are in the “Observation of sensation” hierarchy, see Figure 3. This is an error in the current release of SNOMED CT. This is a problem if a healthcare organization would like to perform analytics on all patients who tested positive for COVID-19 and presented with a fever because the results of this search would return patients that presented with the children of “Fever” but not patients that presented with “Feels hot/feverish” since “Feels hot/feverish” has parents in a hierarchy that is not connected to “Body temperature.” To correctly return all patients who tested positive for COVID-19 and presented with a “Fever” including “Feels hot/feverish,” the concept “Fever” needs to be added to the definition of “Feels hot/feverish” as well. Figure 4 New Parent shows the addition “Fever” as a parent in the “Feels hot/feverish” definition.

Figure 4: New parent - A concept’s definition can be modified to correct errors in meaningful use standards such as adding a parent to an existing concept. The parent “Fever” was added to ensure that “Feels hot/ feverish” would also be included in the Fever lineage.

"Timely Curation of Terminology will improve your data analytics.”

Adding a New Parent

With KNOWLEDGE, a new parent is added and classification is initiated. Once classification is complete, “Feels hot/ feverish” will display with updated relationships that alter the taxonomy. The updated lineage displayed for “Feels hot/feverish” is now present under “Fever” as well as “Observation of sensation” as show in Figure 5 New Taxonomy. The next taxonomy shows “Feels hot/ feverish” as a child of fever and with parents “Fever” and “Observation of sensation”. As a result of this change, a search for patients who tested positive for COVID-19 will return all patients who presented with “Fever” including patients that presented with “Feels hot/feverish.”

Figure 5: New taxonomy. The orange boxes show the parents of “Feels hot/feverish” to be “Fever” and “Observation of sensation.”

Analyze Data

There are many reasons to collect patient data: to determine prevalence of a cough to patients who test positive for COVID-19, to determine the number of underlying health conditions a patient suspected of COVID-19 has, or to determine the reproductive number (R0)3 value for a community (city, county, state etc.) for COVID-19.

KNOWLEDGE allows our organizations to extend meaningful use standards along with integrating local terminology to provide traceable means to fix errors that may be identified in the process of defining encoded forms, or in the process of analyzing patient data.

As described above, “Fever” can be fixed using KNOWLEDGE to manage an organization’s local terminology. Now all patients with a fever — either measured or subjective are found. KNOWLEDGE enables an EHR to perform class-based queries on the data collected with high precision and recall.

With correct data definitions and ability to query data with high precision and recall, business process automation becomes an attainable capability, so that organizations can automate workflows and optimize the care-delivery processes.

Become a learning organization

Integrating local terminology and meaningful use standards enable organizations to encode accurate patient data. This enables clinical workflows that ensure the right information on COVID-19 is available to the right people as quickly as possible. This also helps clinicians make informed decisions for patient care, and guarantees that the right data is collected in the most efficient manner for analysis to provide optimal patient care, and support organizational learning.

To learn more about KNOWLEDGE, please visit https://mindcomputing.com/products/