This is a retrospective, observational study utilizing real-world data. For our study, we used Cerner Real-World Data™ that was provided through Cerner's HealtheDataLab research tool.6 The COVID-19 dataset in HealtheDataLab contained de-identified patient data of one hundred and seventeen thousand (117K) patients from 62 contributing health systems after a database refresh in July 2020. The dataset contained all patients tested for COVID-19 at some point during their visits to one of the 62 health centers. The database contained tables with names like condition, demographics, COVID-19 labs, encounters, and medication that contained information for each of the de-identified patients. Note that the database undergoes a frequent refresh to keep the patients' data up to date.
To begin with, all patients that had Myasthenia Gravis were identified from the condition table using the ICD-9-CM codes (i.e., 358.0 and 358.01), ICD-10-CM codes (i.e., G70.00 and G70.01), and SNOMED-CT codes (i.e., 91637004, 230686005, 193207007, 230685009, 77461000119109, 77471000119109, 31839002, 55051001, 80976008) irrespective of their COVID-19 test result. We chose an exhaustive list of codes to avoid missing any Myasthenia Gravis patients in the database. The database only yielded G70.00 and G70.01, and in total, 91 patients had Myasthenia Gravis. A SQL join was done on the condition and COVID labs table to extract Myasthenia Gravis patients (with or without exacerbation) that had received a COVID-19 test done along with the test date. The patients were then divided into two sets: one with a positive COVID-19 test and the other set that contained patients that always tested negative for COVID-19. There were a total of 27 hospitalized Myasthenia Gravis patients (from November 2019 to July 2020) that tested positive for COVID-19. To find patients who always tested negative for COVID-19, we computed the set difference between all patients with Myasthenia Gravis (91) and those patients who had this condition and tested positive for the COVID-19 (27), and then manually verified the result. This set of non-COVID-19 patients contained 64 patients. For each patient in both the sets, we extracted information such as race, ethnicity, and gender from the demographics table. We extracted information on patient complications, comorbidities, and first-reported-date-of-condition from the condition table. We also extracted start date and end date of medications that were prescribed after the COVID-19 test result from the medication table. Finally, we extracted the discharge disposition and length of stay after the COVID-19 test result from encounter table. The number of deceased patients were obtained from the demographics table and were then verified using the latest discharge disposition from the encounter table. We report the data using means, range, prevalence rates in these two sets of patients. The p-values were calculated using the two-sample t-test and Pearson’s chi-squared test. The p-values are also reported for statistical significance (< 0.05).