Discussion

Our findings highlight persistent differences in COVID-19 antiviral prescribing, with complex and sometimes non-linear associations across race, care setting, insurance type, and social vulnerability.

Overall antiviral treatment of COVID-19 was low with less than third of patients receiving therapy. Outpatient prescribing relies on a complex time sensitive access pathway including testing, clinical evaluation, medication reconciliation, prescription, and access to dispensing sites, which likely contributes to gaps in treatment uptake. ^{Reference Gold, Kelleher and Magid21} Treatment allocation was more commonly seen in older patients as recommended by the IDSA and CDC treatment guidelines. ^{Reference Adarsh Bhimraj, Baden and Bedimo11}

One key finding was that non-White race was independently associated with higher odds of receiving any COVID-19 treatment, even after adjustment; however, this association was not observed in the outpatient-only analysis, and only patients with unknown or missing race had higher odds of treatment. This stands in partial contrast to earlier reports suggesting that racial minorities were less likely to receive COVID-19 antiviral treatments. ^{Reference Malden, McLaughlin and Hong8,Reference Boehmer, Koumans and Skillen22} This discrepancy suggests that the observed relationship may be influenced by differences in care setting rather than reflecting a consistent pattern across populations. ^{Reference Bromley-Dulfano and Barnett23} ED encounters may capture patients with higher acuity or fewer barriers to same-day treatment, whereas outpatient prescribing is often more dependent on timely access, testing, and follow-up. Furthermore, patients receiving care in outpatient settings often have established relationships with their providers, which may facilitate trust, communication, and shared decision-making regarding treatment. ^{Reference Birkhäuer, Gaab and Kossowsky24} In addition, outpatient clinicians may have greater ability to arrange follow-up and monitor for signs of disease progression. In contrast, ED encounters are typically episodic, with limited prior knowledge of the patient and minimal ability to ensure follow-up, which may influence both clinical decision-making and perceived feasibility of initiating outpatient therapies. Although these factors were not directly measured in our study, they may represent an important and underexplored contributor to observed differences in treatment patterns across care settings. At the same time, the inconsistency across analyses highlights the need for cautious interpretation, as residual confounding, differences in healthcare utilization, may contribute to these findings. Taken together, these results underscore the complexity of assessing equity in antiviral access and suggest that disparities may manifest differently across care settings. ^{Reference Bromley-Dulfano and Barnett23}

VVI demonstrated different patterns across treatment contexts. We selected the VVI as our primary measure of social vulnerability because it was specifically designed to capture health care–relevant dimensions of socioeconomic disadvantage. While commonly used indices such as the Area Deprivation Index and the Social Vulnerability Index have been widely applied in health disparities research, they were developed for different primary purposes, mortality prediction and disaster preparedness, respectively. ^{25,Reference Abdul-Mutakabbir26} In addition, because VVI does not include race as a component variable, it enables evaluation of structural and socioeconomic drivers of disparities independent of race, which is particularly important when examining access-related inequities.

In our analysis, the relationship between social vulnerability and COVID-19 antiviral treatment was more complex than a simple linear gradient. In the combined outpatient and ED cohort, patients in the third quartile of the VVI had higher odds of receiving treatment compared with those in the least vulnerable (first) quartile, while no significant differences were observed for the most vulnerable (fourth) quartile. In contrast, in the outpatient-only analysis, patients in the second quartile had lower odds of treatment, with no significant differences observed in higher quartiles. Together, these findings suggest that patients with moderate vulnerability may be more likely to receive COVID-19 antiviral treatment, while those at the highest levels of vulnerability do not experience a proportional increase in treatment despite potentially greater need. ^{Reference Wilcock, Kissler and Mehrotra27}

In other reports, patients who are more socially vulnerable were less likely to receive treatment. ^{Reference Gold, Kelleher and Magid21} Several studies have demonstrated that patients with higher social vulnerability scores have higher risk of severe COVID-19 disease and worse outcomes such as hospitalization. ^{Reference Tipirneni, Karmakar, O’Malley, Prescott and Chopra28} Greater disease severity may result in patients presenting less frequently to the outpatient setting and more often to the ED or inpatient care. ^{Reference Tipirneni, Karmakar, O’Malley, Prescott and Chopra28,Reference Mullachery, Bilal, Li and McClure29} Patients with moderate vulnerability quartiles may have sufficient clinical risk to meet treatment criteria while retaining enough access healthcare to obtain therapy. In contrast, patients in the least vulnerable strata may have lower clinical risk and fewer indications for antivirals, while patients in the highest vulnerability quartile, despite potentially having higher risk, may face greater structural barriers to accessing timely care.

Additionally, outpatient treatment requires several steps including access to timely testing, quick appointment availability, reliable pharmacy access, and medication reconciliation for drug interactions which disproportionately disadvantage patients in socially vulnerable neighborhoods. ^{Reference Chou, Moore, Zhao, Melly, Payvandi and Buehler30,Reference Catalano, Khan, Chatzipanagiotou and Pawlik31} COVID-19 antiviral agents are of most benefit when started within the first 5–7 days on symptom onset. Delayed presentation may therefore reduce clinician consideration of antiviral therapy. ^{Reference Adarsh Bhimraj, Baden and Bedimo11}

A major strength of this study is the use of a large, diverse patient population drawn from a healthcare system encompassing both urban and community practices. This provides robust statistical power and enhances the generalizability of our findings to real-world clinical settings. Additionally, our study examines prescribing patterns during a later phase of the pandemic, a period in which antiviral access was more stable, and guidance was more consistent making it possible to identify persistent inequities.

This study has several limitations. First, prescription data do not confirm medication dispensing or adherence, and we could not determine whether prescriptions were declined, contraindicated, or unavailable at the pharmacy. Second, VVI is less widely validated and has been used in fewer studies compared with established measures of social deprivation, which may limit comparability with prior disparities research. Additionally, provider-level factors like clinician specialty or knowledge of drug–drug interactions could not be assessed. Further, our language variable reflects preferred language recorded in the medical record and cannot determine English proficiency. Lastly, our overall population, when compared to other regions in the United States, appeared less vulnerable which might limit generalizability to severely vulnerable populations seen in other areas of the country.