Columbia Open Health Data (COHD) for COVID-19 Research provides access to counts and visit prevalence (i.e., prevalence from electronic health records) of conditions, procedures, drug exposures, and the co-occurrence frequencies between them. Count and frequency data were derived from the Columbia University Irving Medical Center's OHDSI database including inpatient data. Counts are the number of visits with the concept, e.g., diagnosed with a condition, exposed to a drug, or a procedure was performed. Frequencies are the number of visits with the concept divided by the total number of visits in the dataset. Clinical concepts (e.g., conditions, procedures, drugs) are coded by their standard concept ID in the OMOP Common Data Model. To protect patient privacy, all concepts and pairs of concepts where the count ≤ 10 were excluded, and counts were randomized by the Poisson distribution.
Datasets from three primary cohorts are available:
Both hierarchical and non-hierarchical datasets are available for each cohort. In the hierarchical datasets, the counts for each concept include the visits from all descendant concepts. For example, the count for ibuprofen (ID 1177480) includes visits with Ibuprofen 600 MG Oral Tablet (ID 19019073), Ibuprofen 400 MG Oral Tablet (ID 19019072), Ibuprofen 20 MG/ML Oral Suspension (ID 19019050), etc.
COHD-COVID was developed at the Columbia University Department of Biomedical Informatics as a collaboration between the Weng Lab, Tatonetti Lab, and the NCATS Biomedical Data Translator program (Red Team). This work was supported in part by grants: NCATS 1OT2TR003434, NLM R01LM012895, NCATS OT3TR002027, NLM R01LM009886-08A1, and NIGMS R01GM107145.
COHD can be used to retrieve the patient prevalence of individual concepts. In this example, information for the condition "Atrial fibrillation" (OMOP concept ID 313217) is retrieved from the 5-year non-hierarchical data set. The results show the concept definition (concept ID, concept name, concept domain, concept class, source vocabulary for the OMOP concept, and concept code in the source vocabulary) and the occurrence data (patient count and patient prevalence).
COHD can be used to retrieve association data calculated from the co-occurrences of pairs of concepts. In this example, the relative frequency of drugs associated with the condition "Atrial fibrillation" (OMOP concept ID 313217) is retrieved from the 5-year non-hierarchical data set. The relative frequency informs investigators that among patients who have taken the drug identified in each row, this proportion of patients experienced atrial fibrillation at some time within the 5-year inclusion period of the data set. Note that the relative frequency can exceed the upper limit of 100% due to the Poisson randomization.
Similar to Example 2, in this example, the relative frequency of conditions associated with the drug "Albuterol 0.83 MG/ML Inhalant Solution" (OMOP concept ID 19123989) is retrieved from the 5-year non-hierarchical data set. The relative frequency informs investigators that among patients who have had the condition identified in each row, this proportion of patients have taken "Albuterol 0.83 MG/ML Inhalant Solution" at some time within the 5-year inclusion period of the data set. Note that the relative frequency can exceed the upper limit of 100% due to the Poisson randomization.
The COHD RESTful SmartAPI provides a more powerful method to interact with the COHD data. The web user interface provides access to a few of the most frequently used endpoints in the COHD API. The COHD API additionally provides access to metadata, methods to explore the OMOP concept hierarchy, and cross-referencing between OMOP and external ontologies.