RR:C19 Evidence Scale rating by reviewer:
Potentially informative. The main claims made are not strongly justified by the methods and data, but may yield some insight. The results and conclusions of the study may resemble those from the hypothetical ideal study, but there is substantial room for doubt. Decision-makers should consider this evidence only with a thorough understanding of its weaknesses, alongside other evidence and theory. Decision-makers should not consider this actionable, unless the weaknesses are clearly understood and there is other theory and evidence to further support it.
This manuscript analyzes the electronic health record (EHR) database from the NYC Health and Hospital system, the NYC public system, to assess SARS-CoV-2 positivity and aggressiveness in Asians overall and in Asian subpopulations between March 1 and May 31, with follow up through August 15, 2020. The results add to the data on disparities in COVID-19 outcomes and contributes to the literature in an understudied area of focus, Asian populations. For this reason, the manuscript is significant and potentially actionable.
Some design issues need to be addressed: although the population was selected based on having a COVID-19 test performed, we don’t know if these subjects came to the hospital with a COVID-related complaint, or they were tested as routine practice while undergoing some other non-COVID-19 related medical procedure. These two population samples are very different and so are their COVID-19 hospitalization rates and outcomes. There could be an association to being tested because of a COVID-19-related complaint and that could introduce a significant bias in the study.
The definition of Asian and Asian sub-population was performed based on last names when race/ethnicity was not present in the EHR. The variable race/ethnicity is the key measure in this study, thus we need to know more on its construct: how much data was missing from EHR and how many patients were assigned race/ethnicity based on last name, for example? How reliable is the last name in determining ethnicity in Asian patients? A simple way to go about this would be to select those patients who have race/ethnicity data available and reclassify them according to their last name, and then compare the two definitions to assess how well the last name capture the race/ethnicity reported in the EHR.
Despite the attempt to assign race/ethnicity to these patients, still in roughly 10% of them the information was missing, and the distribution of missing race/ethnicity appeared not to be random: younger patients, patients with commercial insurance, and self-paying patients were missing race/ethnicity more often, for example. This is a bias that could drive the associations in unknown directions. I suspect that restricting the analysis to those who came to the hospital for COVID-19 related complaints could partially solve this problem.
Because of these limitations, the strength of evidence is potentially informative, unless major restructuring of the sample selection and data analysis is performed. There is an opportunity to revise this manuscript and to resubmit a new version that could be more actionable.