Introduction

Oral nirmatrelvir has been shown to reduce the risk of progression to severe coronavirus disease 2019 (Covid-19) in high-risk, unvaccinated patients.1 Reports from the manufacturer reveal a rapid increase in prescriptions months after approval, with as many as 80,000 patients treated in one week of April of 2022.2 The drug may reduce racial and ethnic disparities as people of color are at a higher risk of complications from severe COVID-19 and are less likely to be vaccinated.3 A study of nationwide dispensing data revealed lower dispenses in higher vulnerability zip codes, but the impact of race could not be evaluated.4 We sought to determine whether nirmatrelvir prescriptions’ rate differed along racial and ethnic lines by leveraging data from a large electronic health record (EHR) data aggregation collaborative.

Methods

We queried data from Cosmos, a national EHR data aggregation collaborative with data from over 140 million patients cared for at more than 960 hospitals and 20,000 clinics in the USA.5 Contributing systems include academic organizations, community hospitals, pediatric hospitals, inpatient facilities, ambulatory-only clinics, and safety-net systems. As the data were aggregated and de-identified, institutional review board approval was not required. For our study cohort, we included patients aged 12 years and older with a least 1 health encounter between January 1, 2020, and December 31, 2021. For the primary outcome, we ascertained population-level frequencies of patients with any nirmatrelvir prescriptions from January 1, 2022, to May 31, 2022, stratified along the largest three racial and ethnic categories (White non-Hispanic, Black, or Hispanic). Counts of patients with encounters associated with a diagnosis of COVID-19 and positive lab-based SARS-CoV-2 testing were enumerated as well. As a sensitivity analysis, we conducted a similar analysis in a subset of patients who were older than 65 years of age or had diagnostic coding for at least one medical comorbidity known to increase the risk of developing severe COVID-19.

Results

There were 44.4 million patients who met criteria for study inclusion. The mean age was 49 ± 21 years, and 57% of patients were female. White, non-Hispanic patients made up 64% of the cohort, Black patients made up 13%, and Hispanic patients comprised 10%. A total of 107,589 patients (2.4%) had at least one prescription for nirmatrelvir during the study period (see Table). Background rates of SARS-CoV-2 positive tests or encounters with a diagnosis of COVID-19 were similar between White non-Hispanic, Black, and Hispanic patients, but the overall rates of prescriptions per group differed substantially (Fig. 1). In May, 24 per 10,000 White patients received nirmatrelvir prescriptions compared to 9 per 10,000 Black patients (p<0.001) and 10 per 10,000 Hispanic patients (p<0.001 when compared to White non-Hispanic patients). These differences persisted after restricting the population to higher risk patients (Table).

Figure 1
figure 1

Number of patients with COVID-19 diagnosis or test positivity per month (lines, left y-axis) and overall nirmatrelvir prescriptions per month (bars, right y-axis)

Full size image

Discussion

Even though rates of SARS-CoV-2-positive tests did not differ between major racial and ethnic groups, nirmatrelvir prescriptions where more than twice as common in White patients as they were in Blacks or Hispanics. These differences persisted in a higher risk subset, despite historically lower vaccination rates in people of color. Our findings highlight potentially worrisome racial and ethnic disparities in the prescription of nirmatrelvir during a time when COVID-19 cases were rising. Why this gap exists cannot be determined form our data set, but possibilities include disparities in access, provider behavior, and racial attitudes towards COVID-19 therapeutics. Unfortunately, the absence of prescription fill data makes it difficult to detect potential differences in post-prescription treatment follow-through. Our results are strengthened by the inclusion of a large and diverse sample from the EHRs of many healthcare systems across the USA. The limitations of our analysis include our reliance on diagnostic codes for our sensitivity analysis, which are prone to underdiagnosis as well as misdiagnosis, and the absence of home test results (as these are not likely routinely entered as discrete data points in the EHR). Due to limitations in the Cosmo query interface, we were unable to adjust for individual demographic and comorbid conditions. Finally, our results may be affected by evolving guidance and attitudes in prescribing nirmatrelvir, particular with mounting concerns of potential rebound symptoms. In conclusion, our findings should prompt closer attention to detecting and understanding disparities in nirmatrelvir prescriptions, particularly for those most at risk.6