Between December 22, 2021 and August 12, 2023, Walgreens dispensed 3,517,404 nirmatrelvir/ritonavir prescriptions to 2,971,056 individuals. Our analysis included 2,103,570 unique prescriptions to 1,985,990 individuals (eFig. 1). Overall, nearly all individuals prescribed nirmatrelvir/ritonavir were ≥ 18 years old (99%). Most were ≥ 50 years old (66%) and did not have a documented high-risk condition (70%). The vast majority (95%) had only one nirmatrelvir/ritonavir prescription, and 5% had > 1 prescription ordered. Among those with only one dose prescribed, 95% were prescribed the standard dose (Table 1).

Figure 1 plots the number of nirmatrelvir/ritonavir prescriptions ordered at Walgreens pharmacies and the percent filled over time for those with only one prescription. The number of prescriptions ordered was low immediately following EUA, but increased in April 2022, with two peaks in July 2022 and December 2022, corresponding with a summer wave and rising infections between Thanksgiving and the new year. Prescriptions then declined through mid-April 2023, stabilized through the end of June 2023, and then increased again. The proportion of prescriptions filled was initially low but increased from 49% during the first full week of prescribing to 61% in the second full week. The proportion of prescriptions filled continued to rise through the end of May 2022 and then stabilized (average 85%; range 81–88%).

Nirmatrelvir/ritonavir prescriptions ordered weekly and percent of nirmatrelvir/ritonavir prescriptions filled within 90 days of being ordered among patients who had a single nirmatrelvir/ritonavir prescription during the study period

Among those with only one nirmatrelvir/ritonavir prescription, 88% filled their prescription (Table 1). Consistent with our overall findings, in the subset of patients with a self-reported symptom onset date (n = 21,553), 88% of those prescribed nirmatrelvir/ritonavir filled it and 76% of those who filled it did so within 5 days of symptom onset (overall, 67% of those prescribed nirmatrelvir/ritonavir filled within 5 days of symptom onset).

For individuals with > 1 prescription, 77% filled only one prescription and 13% filled > 1. Only 3% of those with > 2 prescriptions filled > 2. Only 2% of those with > 3 prescriptions filled > 3 (Fig. 2).

Percent of nirmatrelvir/ritonavir prescriptions filled within 90 days of being ordered by the number of nirmatrelvir/ritonavir prescription orders patients received

Among the subset of patients with a self-reported symptom onset date, < 1% of filled prescriptions were filled on the day symptoms started, 11% were filled 1 day after, 47% were filled 2–3 days after, and 17% were filled 4 days after symptom onset. After the tenth day, the percent filling a prescription on any day was < 1% (Fig. 3). The proportion of individuals filling a prescription within 5 days of symptom onset was initially low (16%) but increased to 70% by June 2022. Between June 2022 and August 2023, the proportion filling a prescription within 5 days of symptom onset fluctuated between 61% and 71%. Similar trends were observed when the number of days between symptom onset and prescription fill date were graphed for different time intervals within the first 5 days of symptom onset (eFig. 2).

Days between symptom onset and nirmatrelvir/ritonavir prescription order fill

For individuals with only one nirmatrelvir/ritonavir prescription, the likelihood of filling a prescription was ≥ 84% for all age groups but increased slightly with age (84% for 12–17-years-olds, 85% for 18–49-years-olds, 89% for 50–64-year-olds, and 90% for those ≥ 65 years old; all P < 0.0001 vs. those aged 12–17 years). Likelihood of filling a prescription was also slightly higher for those with vs. without a documented high-risk condition (90% vs. 87%; P < 0.0001). Those pregnant or recently pregnant (vs. not) were less likely to fill a prescription (81% vs. 88%; P < 0.0001) (Table 1). In adjusted models, most effect sizes were very small (i.e., < 5% difference) and not clinically meaningful, with the exception of age group. Compared with those 12–17 years old, likelihood of filling a prescription increased with age: 18–49 years old (adjusted relative risk [aRR] 1.01; 95% confidence interval [CI] 1.01–1.02), 50–64 years old (aRR 1.06; 95% CI 1.06–1.07), and ≥ 65 years old (aRR 1.07; 95% CI 1.06–1.08) (Table 2). In an exploratory analysis modeling individual high-risk conditions rather than the number of high-risk conditions, effect sizes for all statistically significant conditions were small (i.e., < 5% difference) and not clinically meaningful (eTable 2).

For individuals with > 1 prescription during the 90-day infection period, the proportion filling > 1 prescription increased with age (10% of 12–49-years-olds, 13% of 50–64-year olds [P = 0.003 vs. those aged 12–17 years], and 14% of those ≥ 65 years old [P < 0.0001 vs. those aged 12–17 years]), and was higher among those with ≥ 1 documented high-risk condition vs. none (14% vs. 12%; P < 0.0001). Individuals with the following high-risk conditions were significantly more likely to fill > 1 treatment course: primary immunodeficiencies (33%), cancer (25%), chronic lung disease (16%), asthma (15%), heart conditions (14%), and those taking corticosteroids or immunosuppressive medications (17%). Those filling > 1 prescription (vs. 0 or 1) were more likely to use pharmacies in tracts with lower levels of deprivation (mean ADI 41.6 vs. 44.4, respectively; P < 0.0001) and in trade areas ≥ 25% non-Hispanic white (13% vs. 10% in trade areas < 25% non-Hispanic white; P < 0.0001) (Table 1).

Results from multivariable models were similar, with the largest effects observed for age, number of high-risk conditions, and proportion of the pharmacy trade area characterized as non-Hispanic white. Compared with 12–17-years-olds, the likelihood of filling multiple prescriptions was higher for 50–64-years-olds (aRR 1.29; 95% CI 1.08–1.55) and those ≥ 65 years old (aRR 1.44; 95% CI 1.20–1.72). The likelihood of filling > 1 prescription increased with the number of high-risk conditions (aRR 1.16; 95% CI 1.12–1.21 for those with one and aRR 1.27; 95% CI 1.21–1.34 for those with ≥ 2) and with the proportion of the pharmacy’s trade area characterized as non-Hispanic white (aRR 1.18; 95% CI 1.12–1.25 for pharmacy trade areas characterized as 25–75% vs. < 25% non-Hispanic white and aRR 1.23; 95% CI 1.15–1.31 for pharmacy in trade areas characterized as > 75% vs. < 25% non-Hispanic white) (Table 2). In an exploratory analysis modeling individual high-risk conditions rather than the number of high-risk conditions (eTable 2), those with cancer (aRR 1.69; 95% CI 1.30–2.20), asthma (aRR 1.17; 95% CI 1.11–1.23), or taking corticosteroids or other immunosuppressive medications (aRR 1.21; 95% CI 1.09–1.36) were more likely to fill > 1 prescription.

In the subanalysis restricted to those with a symptom onset date, differences in the likelihood of filling a single nirmatrelvir/ritonavir prescription within 5 days of symptom onset were observed by self-reported race/ethnicity, with those identifying as non-Hispanic white most likely to fill prescriptions (P < 0.0001 for all comparisons except vs. those identifying as non-Hispanic Native [P = 0.93]; eTable 3); this disparity remained significant in the multivariable model. Compared with those identifying as Non-Hispanic white, those identifying as Hispanic, Non-Hispanic Asian, and non-Hispanic Black or African American were 7%, 11%, and 16% less likely to fill a prescription within 5 days of symptom onset, respectively (eTable 4). In the primary analysis, where self-reported demographics were not available, the proportions filling one and multiple prescriptions were higher among those using pharmacies in areas with a larger non-Hispanic white population (Tables 1 and 2). Unlike the primary analysis, statistically significant differences were not observed by age (eTable 3). In contrast to the primary analysis, which used ICD-10 codes and GPIs from the past 12 months to identify high-risk conditions, the supplementary analysis used self-reported conditions. In the multivariable model, self-reporting being overweight or obese was associated with a 5% higher likelihood and self-reported diabetes was associated with a 3% lower likelihood of filling a prescription within 5 days of symptom onset. Self-reported fever was associated with an increased likelihood of filling within 5 days of symptom onset (aRR 1.06; 95% CI 1.04–1.08), while cough (aRR 0.94; 95% CI 0.92–0.96), shortness of breath (aRR 0.92; 95% CI 0.89–0.95), recent loss of sense of smell or taste (aRR 0.80; 95% CI 0.75–0.84), and diarrhea (aRR 0.89; 95% CI 0.86–0.93) were all associated with a decreased likelihood of filling a prescription within 5 days of symptom onset. Compared with those unvaccinated, those vaccinated with the primary series (aRR 1.07; 95% CI 1.02–1.12) or both the primary and booster series (aRR 1.12; 95% CI 1.07–1.17) were more likely to fill a prescription within 5 days of symptom onset (eTable 4).