Predicting the Public Health Impact of Bivalent Vaccines and Nirmatrelvir-Ritonavir Against Coronavirus Disease 2019

Abstract

Background

Uptake of coronavirus disease 2019 (COVID-19) bivalent vaccines and the oral medication nirmatrelvir-ritonavir (Paxlovid) has remained low across the United States. Assessing the public health impact of increasing uptake of these interventions in key risk groups can guide further public health resources and policy and determine what proportion of severe COVID-19 is avertable with these interventions.

Methods

This modeling study used person-level data from the California Department of Public Health on COVID-19 cases, hospitalizations, deaths, and vaccine administration from 23 July 2022 to 23 January 2023. We used a quasi-Poisson regression model calibrated to recent historical data to predict future COVID-19 outcomes and modeled the impact of increasing uptake (up to 70% coverage) of bivalent COVID-19 vaccines and nirmatrelvir-ritonavir during acute illness in different risk groups. Risk groups were defined by age (≥50, ≥65, ≥75 years) and vaccination status (everyone, primary series only, previously vaccinated). We predicted the number of averted COVID-19 cases, hospitalizations, and deaths and number needed to treat (NNT).

Results

The model predicted that increased uptake of bivalent COVID-19 boosters and nirmatrelvir-ritonavir (up to 70% coverage) in all eligible persons could avert an estimated 15.7% (95% uncertainty interval [UI], 11.2%–20.7%; NNT: 17 310) and 23.5% (95% UI, 13.1%–30.0%; NNT: 67) of total COVID-19–related deaths, respectively. In the high-risk group of persons ≥65 years old alone, increased uptake of bivalent boosters and nirmatrelvir-ritonavir could avert an estimated 11.9% (95% UI, 8.4%–15.1%; NNT: 2757) and 22.8% (95% UI, 12.7%–29.2%; NNT: 50) of total COVID-19–related deaths, respectively.

Conclusions

These findings suggest that prioritizing uptake of bivalent boosters and nirmatrelvir-ritonavir among older age groups (≥65 years) would be most effective (based on NNT) but would not address the entire burden of severe COVID-19.

This modeling study, in collaboration with the California Department of Public Health, found prioritizing additional uptake of bivalent vaccines and nirmatrelvir-ritonavir among older age-groups would effectively reduce the number of severe COVID-19 cases but would not address the entire burden.

The coronavirus disease 2019 (COVID-19) pandemic, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), continues to be a public health problem in the United States (US) [1, 2]. The epidemiologic landscape of COVID-19 in the US has changed over the pandemic and is now characterized by a population with widespread vaccination with monovalent COVID-19 vaccines (including 69% fully vaccinated, and 38% with monovalent booster doses as of March 2023), high prevalence of prior infection, and emergence of increasingly infectious SARS-CoV-2 variants such as Omicron subvariants [3–5]. As social distancing and public health measures are relaxed, a key public health question is understanding the impact of increasing uptake of additional vaccination and oral medications to further mitigate hospitalizations and deaths from COVID-19 and to understand what proportion of severe COVID-19 cases are avertable with these interventions. Two key medical interventions against severe COVID-19 are bivalent vaccines and use of oral antiviral medications during COVID-19 illness, most commonly nirmatrelvir-ritonavir (Paxlovid) [6–9]; however, uptake of these interventions has been low [5, 10].

COVID-19 vaccination is a key tool to reduce severe COVID-19 [11, 12]. Bivalent messenger RNA (mRNA) vaccines, composed of components of the SARS-CoV-2 ancestral and Omicron BA.4 and BA.5 (BA.4/BA.5) variants, were made available in the US in the beginning of September 2022 [13]. Bivalent COVID-19 vaccines were recommended by the Advisory Committee on Immunization Practices within the US Centers for Disease Control and Prevention as booster doses to potentially better target the Omicron variant and subvariant waves [14]. Observational clinical data on bivalent COVID-19 vaccines suggest benefit of booster doses of this vaccine to reduce symptomatic infection [6, 15], and additional data support improved protection against COVID-19–related hospitalizations [16]. As of March 2023, uptake of the new bivalent vaccines in California is low, with only 24% of adults and 44% of those >65 years of age having received a dose [17].

The oral antiviral drug nirmatrelvir-ritonavir is another key public health tool for minimizing severe COVID-19 outcomes in high-risk patients. In December 2021, nirmatrelvir-ritonavir was given US Food and Drug Administration (FDA) Emergency Use Authorization on the basis of evidence that this medication can reduce hospitalization and death among COVID-19 patients with mild/moderate symptoms who are at high risk for progression to severe COVID-19 within 5 days of symptom onset [7–9, 18]. However, use of nirmatrelvir-ritonavir among eligible patients with COVID-19 in the US has been low, with studies estimating that only 28% of eligible persons were prescribed nirmatrelvir-ritonavir [7, 8, 10].

This article reports on the predicted public health impact of increasing uptake of bivalent vaccines and nirmatrelvir-ritonavir in key risk groups on COVID-19 cases, hospitalizations, and deaths. While both COVID-19 vaccines and nirmatrelvir-ritonavir are not a constrained resource, this study aims to prioritize time and effort by public health departments to promote uptake of these interventions in the groups most likely to benefit. Furthermore, this study aims to understand what proportion of severe COVID-19 cases are avertable with high uptake of these interventions. We use the case example of California given the large and diverse population and magnitude of COVID-19 burden.

METHODS

Data

We obtained person-level data on confirmed COVID-19 cases, hospitalizations, and deaths in California from 23 July 2022 to 23 January 2023 from the California Department of Public Health (CDPH). A COVID-19 case was defined as a person whose positive SARS-CoV-2 molecular test was reported to the state. COVID-19 hospitalizations and deaths were defined as a confirmed COVID-19 case who was either hospitalized or died with COVID-19 and reported to the state. CDPH receives reports on hospitalizations and deaths from 2 independent sources (California COVID-19 Reporting System and healthcare facility–mandated reporting). In addition, estimates for hospitalizations utilized a multiplier to account for 75% ascertainment of linkage of cases to hospitalization based on internal analysis. We used case episode dates to link the earliest date associated with a SARS-CoV-2 infection for hospitalization and death.

We obtained publicly available vaccine administration data from CDPH (23 July 2022–23 January 2023) to estimate vaccine status, defined as partially vaccinated, fully vaccinated, and boosted. Fully vaccinated referred to those who completed their primary series, defined as having received 1 dose of the Ad26.COV2.S vaccine (Janssen) or 2 doses of the BNT162b2 mRNA (Pfizer/BioNTech) or mRNA-1273 (Moderna) vaccine. Partially vaccinated referred to those who have received at least 1 vaccine dose but have not completed a primary series. Boosted referred to those who have completed the primary series and received at least 1 monovalent vaccine dose. We also estimated age-specific coverage of bivalent booster doses. Vaccine data were demographically stratified by age group (5–11 years, 12–17 years, 18–49 years, 50–64 years, ≥65 years). We excluded vaccination data with missing age information.

Study Outcomes

The primary study outcomes were COVID-19 cases, hospitalizations, and deaths. We calculated the total averted outcomes (absolute measure), the number needed to treat (NNT; relative measure), and proportion of total averted outcomes associated with each intervention strategy.

Statistical Analysis

We used a modeling approach to predict the number of COVID-19 cases, hospitalizations, and deaths over a future 6-month period to fully capture difference between strategies, and then estimated how many of these outcomes could be averted with additional uptake of bivalent vaccines (or monovalent vaccines for unvaccinated persons) and use of nirmatrelvir-ritonavir during acute illness.

Predicting COVID-19 Outcomes

In this model, we predicted the number of cumulative COVID-19 outcomes (cases, hospitalizations, deaths) over the next 6 months (January 2023–July 2023) without introduction of any additional vaccination or nirmatrelvir-ritonavir treatment aside from baseline uptake (base case scenario). Models were calibrated to data from 23 July 2022 to 23 January 2023. We used quasi-Poisson regression models to predict the number of weekly COVID-19 outcomes based on age and vaccine status, from which we estimated cumulative COVID-19 outcomes for each age and vaccine status group at the end of the 6-month period. We used this parsimonious set of predictors given their strong relation to COVID-19 outcomes [19] and given that our goal was to estimate the cumulative number of COVID-19 outcomes over a time period by relevant risk group (age, vaccination status) [20]. We defined age group as 0–17 years, 18–49 years, 50–64 years, 65–74 years, 75–84 years, or ≥85 years, informed by estimations of case fatality rate in each group (Supplementary FigureA1). We defined vaccine status as unvaccinated, primary series, primary series with 1 booster dose, or primary series with ≥2 booster doses. Partially vaccinated individuals were classified as fully vaccinated (primary series only) for simplicity as they represented a small fraction (<4%) of total COVID-19 outcomes. We fit separate regression models for each of the 3 COVID-19 outcomes (cases, hospitalizations, deaths). We accounted for the effect of prior bivalent vaccine coverage during the calibration period, and adjusted the prediction of COVID-19 outcomes based on the expected impact of prior bivalent vaccination (see Supplementary Materials). For model validation, we evaluated model performance on an alternative 6-month calibration period (Supplementary Table 1).

Predicting the Impact of Additional Vaccination

We modeled 7 vaccination strategies, which simulated administering 1 additional dose of a COVID-19 vaccine (majority bivalent) to key risk groups, based on age and vaccine status. We modeled the following vaccine strategies, which targeted (1) everyone, regardless of vaccination status; (2) previously vaccinated individuals; (3) unvaccinated individuals; (4) individuals who have completed the primary series only; (5) persons aged ≥75 years (excluding unvaccinated individuals); (6) persons aged ≥65 years (excluding unvaccinated individuals); and (7) persons aged ≥50 years (excluding unvaccinated individuals). We assumed use of bivalent vaccines for all strategies, except for the primary series, which remains as a monovalent dose. These strategies were selected with input from CDPH. While all groups are eligible by guidelines [20] to receive vaccination, this analysis is intended to provide an estimate of the impact of increasing uptake and coverage in these groups. Additional analyses were conducted, examining vaccine strategies that stratified older age groups by vaccination status (Supplementary Table 2). For modeling of vaccination, we used available data on vaccine effectiveness of bivalent vaccination by different baseline vaccination statuses against symptomatic infection [6], hospitalization [16, 21], and death [21], and extrapolated estimates on vaccine effectiveness from data on monovalent vaccines (assuming this provides a conservative estimate) [21]. We modeled durable vaccine-induced protection over the 6-month simulation period, except for waning of vaccine-induced protection against infection. We did not assume additional benefit beyond 3 booster doses to be conservative. Our assumed vaccine effectiveness estimates for bivalent vaccination are shown in Table 1. We accounted for current baseline coverage of bivalent vaccines (23.7% among eligible population; see Supplementary Materials for age-specific estimates). We estimated the impact of 70% coverage of interventions in the base case analysis. We calculated total averted outcomes for each vaccine strategy by applying vaccine effectiveness estimates to counts of predicted COVID-19 outcomes by risk groups for different vaccine strategies, subtracting out benefit from recent bivalent vaccination already given. For each vaccine strategy, we calculated the number of vaccine doses necessary for distribution across the California population. We estimated the NNT for each vaccine strategy, defined as number of individuals needed to receive an additional vaccine dose to avert 1 COVID-19 outcome. NNT was calculated as the total vaccine doses administered divided by the total number of outcomes averted.

Table 1.

Characteristics of Coronavirus Disease 2019 Cases, Hospitalizations, and Deaths Between July 2022 and January 2023 and Model Parameters Related to Vaccination and Nirmatrelvir-Ritonavir Treatment

Characteristic	Cases (N = 1 128 004)	Hospitalizations (N = 43 684)	Deaths (N = 5876)
Descriptive characteristics
Sex
Female	605 770 (53.7%)	22 360 (51.2%)	2680 (45.6%)
Male	506 918 (44.9%)	21 141 (48.4%)	3181 (54.1%)
Unknown or nonbinary	15 316 (1.4%)	183 (0.4%)	15 (0.3%)
Age group, y
0–17 y	136 014 (12.0%)	1688 (3.8%)	12 (0.2%)
18–49 y	545 796 (48.4%)	8420 (19.3%)	250 (4.3%)
50–64 y	234 546 (20.8%)	7256 (16.6%)	653 (11.1%)
65–74 y	107 977 (9.6%)	8156 (18.7%)	1143 (19.4%)
75–84 y	66 798 (5.9%)	9568 (21.9%)	1541 (26.2%)
≥85 y	36 873 (3.3%)	8596 (19.7%)	2277 (38.8%)
Race/ethnicity
American Indian	4292 (0.4%)	181 (0.4%)	18 (0.3%)
Asian	123 425 (10.9%)	3788 (8.7%)	848 (14.4%)
Black	53 722 (4.8%)	3487 (8.0%)	444 (7.6%)
Latinx	315 659 (28.0%)	13 789 (31.5%)	1337 (22.7%)
Native Hawaiian and other Pacific Islander	6767 (0.6%)	207 (0.5%)	18 (0.3%)
White	264 133 (23.4%)	17 249 (39.5%)	2964 (50.4%)
Multiracial	5462 (0.5%)	359 (0.8%)	74 (1.3%)
Othera	115 492 (10.2%)	3372 (7.7%)	138 (2.4%)
Unknown	239 052 (21.2%)	1252 (2.9%)	35 (0.6%)
Vaccination status
Unvaccinated	396 325 (35.1%)	14 853 (34.0%)	2245 (38.2%)
Primary series	236 825 (21.0%)	10 331 (23.6%)	1188 (20.2%)
Boosted (1 dose)	371 305 (32.9%)	11 863 (27.2%)	1487 (25.3%)
Boosted (≥2 doses)	123 549 (11.0%)	6637 (15.2%)	956 (16.3%)
Relative VEb (1 additional dose)
Baseline vaccination status
Unvaccinated (giving 1 monovalent dose)	See Supplementary Table A1 [22]	29% (21–34) [23]	57% (46–69) [23–25]
Primary series (giving 1 bivalent dose)	See Supplementary Table A1 [22, 26]	57% (41–69) [16]	75% (67–83) [27–29]
Boosted (1 dose) (giving 1 bivalent dose)	See Supplementary Table A1 [30, 31]	38% (13–56) [16]	62% (34–90) [30, 32, 33]
Boosted (2 doses) (giving 1 bivalent dose)	See Supplementary Table A1 [30, 31]	56% (12–78) [21]	63% (27–81) [21]
Nirmatrelvir-ritonavir effectiveness
Vaccination status
Unvaccinated	…	89% (51–100) [34]	89% (51–100) [34]
Vaccinated	…	40% (19–56) [8]	71% (29–88) [8]

Data are presented as No. (%) or % (95% CI). See Supplementary Materials for further description of VE and nirmatrelvir-ritonavir data. Estimates on VE are informed by literature from both bivalent and monovalent vaccination.

Abbreviation: VE, vaccine effectiveness.

^a“Other” refers to individuals who do not identify with the available selection of race/ethnicity categories.

^bRelative VE against a COVID-19 case included waning of protection, with complete estimates available in the Supplementary Table A1.

Predicting the Impact of Nirmatrelvir-Ritonavir Treatment

We compared 4 main nirmatrelvir-ritonavir prioritization strategies, which simulate usage of nirmatrelvir-ritonavir during acute COVID-19 in various risk groups. We modeled the following prioritization strategies, which targeted persons (1) aged ≥18 years with comorbidity or high-risk features (mainly immunocompromising conditions for 18–49 years group); (2) aged ≥50 years with comorbidity or high-risk features; (3) aged ≥65 years; and (4) aged ≥75 years. We based medical eligibility for nirmatrelvir-ritonavir as (1) receipt of a positive SARS-CoV-2 test result; (2) ≤5 days since symptom onset or positive test; and (3) belonging to a key risk group, such as ≥65 years; ≥50 years and unvaccinated; ≥50 with multiple medical comorbidities; or person with immunocompromising condition. These eligibility criteria were modeled after the FDA's nirmatrelvir-ritonavir eligibility guidelines [35]. Alternatively, in exploratory analysis, we also investigated 18 years and older and 50 years and older, which expanded the eligibility beyond the current guidance (Supplementary Table 4). In all analyses, we excluded those with contraindications for nirmatrelvir-ritonavir. To account for the true eligible population among each risk group, or the proportion of individuals who meet all the criteria, we created a nirmatrelvir-ritonavir treatment cascade (Figure 1) using published literature estimates for each step. We accounted for current age-specific baseline usage of nirmatrelvir-ritonavir (∼22% in overall population; age-specific estimates in Supplementary TableA3) and evaluated the effects of increasing prescription (step D of Figure 1). Additional analyses were conducted, examining nirmatrelvir-ritonavir uptake strategies that stratified older age groups by vaccination status (see Supplementary Table 3).

Figure 1.

Nirmatrelvir-ritonavir treatment care cascade. We estimated the probability of receiving nirmatrelvir-ritonavir in the population of all coronavirus disease 2019 (COVID-19) cases based on medication eligibility (A), those who seek medical attention (B) [7], no contraindications (C) [7], and who are prescribed therapy (D) [7, 8]. In the study model, we increased uptake at (D) to simulate higher nirmatrelvir-ritonavir usage.

In this model, we used published data on nirmatrelvir-ritonavir effectiveness to estimate the total number of COVID-19 hospitalizations and deaths averted due to these nirmatrelvir-ritonavir prioritization strategies. We extrapolated the effect size of the nirmatrelvir-ritonavir effectiveness for the exploratory analysis in treating persons 18 years and older. Our assumed nirmatrelvir-ritonavir effectiveness estimates for treatment are shown in Table 1; current estimates for effectiveness in vaccinated persons is based on available observational data and can be updated as additional data becomes available [8]. We calculated total averted hospitalizations and deaths for each strategy by applying nirmatrelvir-ritonavir effectiveness estimates to the predicted baseline outcome counts corresponding to the risk groups for different strategies. For each prioritization strategy, we also calculated the number of nirmatrelvir-ritonavir prescriptions necessary based on the number of COVID-19 cases. In addition to estimating total averted outcomes, we also estimated the NNT for each nirmatrelvir-ritonavir prioritization strategy. NNT was calculated as the total prescriptions to persons who otherwise would not have received treatment by strategy divided by the total number of outcomes averted.

Sensitivity and Uncertainty Analysis

We conducted several sensitivity analyses to evaluate the robustness of our study findings (see Supplementary Materials). We conducted sensitivity analyses modeling various uptake assumptions including perfect uptake and uptake correlated and inversely correlated with risk. We conducted a sensitivity analysis modeling higher rate of background vaccination and alternative age-stratified baseline bivalent booster coverage in older age groups, and different fractions of COVID-19 outcome reporting. We performed an analysis with combined interventions of nirmatrelvir-ritonavir and bivalent vaccine. We generated 95% uncertainty intervals (UIs) for the primary analysis based on uncertainty in COVID-19 outcome and treatment effectiveness (see Supplementary Materials).

Patient Consent Statement

This study was approved by the institutional review board at the University of California, San Francisco and Stanford University. The requirement for informed consent was waived given the analysis used anonymized secondary datasets that were collected as part of public health surveillance and deemed minimal risk. Study reporting followed relevant aspects of Consolidated Health and Economic Evaluation Reporting Standards (CHEERS) guidelines. Data requests can be made to CDPH. Analytic code is available at: github.com/hailey-park/bivalent-vaccines-paxlovid-impacts.

RESULTS

Descriptive Data

Over the period from 23 July 2022 to 23 January 2023, there were 1 128 962 confirmed COVID-19 cases reported in California (Figure 2). We excluded 958 cases (0.08%) due to missing covariate data; the final sample size was 1 128 004 COVID-19 cases. Among the COVID-19 cases included in this analysis, 43 684 were reported as COVID-19–related hospitalizations (3.9%) and 5876 were reported as COVID-19–related deaths (0.5%). An estimated 60% of reported hospitalizations and 84% of reported deaths occurred in those 65 years and older. More information on demographics of COVID-19 cases, hospitalizations, and deaths is shown in Table 1.

Figure 2.

Coronavirus disease 2019 (COVID-19) cases, hospitalizations, deaths, and vaccination over time in California. Data on COVID-19 outcomes were obtained from the California Department of Public Health for the period of 1 January 2020 to 23 January 2023. These data included weekly absolute COVID-19 cases based on a positive test reported to the state (A), COVID-19–related hospitalizations (B), and COVID-19–related deaths (C). We plotted coverage of different COVID-19 vaccination statuses (D) using publicly available data from 1 December 2020 to 23 January 2023. The boosted coverage is among the booster-eligible population. The model calibration period (23 July 2022 to 23 January 2023) is shaded.

As of 23 January 2023, we found that 82% of people in California had received at least 1 dose of a COVID-19 vaccine, 73% of people had completed the primary series, and 60% had received at least 1 monovalent booster. Uptake of a monovalent booster dose was reported to be highest among adults aged 18–49 years (43%). Coverage of bivalent doses was 23.7% in the overall eligible population and reported to be highest among those 65 years and older (36% of bivalent doses).

Model Calibration, Prediction, and Validation

The calibrated regression model predicted a total of 1 174 195 (95% confidence interval [CI], 985 150–1 403 574) COVID-19 cases, 48 893 (95% CI, 42 503–56 281) COVID-19–related hospitalizations, and 7189 (95% CI, 6239–8311) COVID-19–related deaths over a 6-month period, driven by historical data (Supplementary FigureA2). We performed a validation exercise of the model prediction by changing the calibration periods to an earlier 6-month period (23 April 2022–23 October 2022) to evaluate model performance, and estimated overall similar relative ranking of risk groups, although underestimated outcomes due to a surge during the validation period (Supplementary Table 1).

Comparison of Vaccine Strategies

The model predicted the public health impact of additional COVID-19 vaccine doses and the number needed to avert 1 COVID-19 case, hospitalization, and death in different epidemiologic groups defined by age and vaccine status (Table 2). For averting COVID-19 cases, targeting everyone with an additional vaccine dose at 70% coverage was predicted to avert 187 201 cases (95% UI, 168 691–205 233; NNT: 111), corresponding to 16.5% of total cases. In the unvaccinated group, an additional vaccine dose was predicted to avert 95 458 cases (95% UI, 87 586–101 917; 8.4% of total cases; NNT: 50). In persons aged ≥65 years, an additional bivalent vaccine dose was predicted to avert 10 794 cases (95% UI, 9439–12 333; 1.0% of total cases; NNT: 177).

Table 2.

Public Health Impact and Number Needed to Treat to Avert Coronavirus Disease 2019 (COVID-19) Cases, Hospitalizations, and Deaths With Bivalent COVID-19 Vaccine Strategies

Group/Strategy	COVID-19 Cases			COVID-19 Hospitalizations			COVID-19 Deaths
	NNT	Total Averted	% Averteda	NNT	Total Averted	% Averteda	NNT	Total Averted	% Averteda
Everyone
Strategy 1 (everyoneb)	111 (101–123)	187 201 (168 691–205 233)	16.5% (14.8–18.3)	2273 (1961–3059)	9066 (6736–10 511)	21.0% (14.7–25.4)	11 378 (8676–14 783)	1811 (1394–2375)	31.1% (24.5–38.1)
Vaccine group–based strategies
Strategy 2 (previously vaccinated)	173 (145–211)	91 743 (75 197–109 362)	8.1% (6.6–9.7)	2623 (2130–4062)	6039 (3901–7438)	14.0% (8.5–18.1)	17 310 (12 053–24 867)	915 (637–1314)	15.7% (11.2–20.7)
Strategy 3 (unvaccinatedb)	50 (47–55)	95 458 (87 586–101 917)	8.4% (7.7–9.0)	1575 (1322–2128)	3027 (2240–3606)	7.0% (5.1–8.4)	5320 (4167–6981)	896 (683–1144)	15.4% (12.0–18.9)
Strategy 4 (primary series only)	139 (133–146)	44 147 (41 911–46 226)	3.9% (3.7–4.1)	1881 (1530–2573)	3245 (2373–3990)	7.5% (5.4–9.4)	12 159 (9165–16 021)	502 (381–666)	8.6% (7.3–10.0)
Age group–based strategies
Strategy 5 (≥75 y)	144 (126–165)	5302 (4636–6061)	0.5% (0.4–0.5)	404 (336–572)	1892 (1335–2275)	4.4% (2.9–5.5)	1435 (1178–1862)	532 (410–648)	9.1% (6.5–11.7)
Strategy 6 (≥65 y)	177 (155–202)	10 794 (9439–12 333)	1.0% (0.8–1.1)	698 (581–990)	2725 (1923–3275)	6.3% (4.2–7.9)	2757 (2225–3609)	690 (527–855)	11.9% (8.4–15.1)
Strategy 7 (≥50 y)	173 (148–205)	28 734 (24 251–33 622)	2.5% (2.1–3.0)	1274 (1057–1856)	3898 (2677–4701)	9.0% (5.8–11.4)	5941 (4659–8115)	836 (612–1066)	14.4% (10.0–18.4)

All analyses compared 70% vaccine uptake to baseline coverage of bivalent vaccines. All age group–based strategies (strategies 5–7) excluded the unvaccinated population when targeting vaccines to older age groups. Data is presented with 95% uncertainty intervals shown in parentheses.

Abbreviation: COVID-19, coronavirus disease 2019; NNT, number needed to treat.

^aThe denominator of this estimate accounted for baseline coverage of bivalent vaccines.

^bFor unvaccinated persons, we assumed use of a monovalent vaccine following current clinical guidance for the primary series.

For averting severe COVID-19 (hospitalization and death), we predicted that targeting everyone with an additional vaccine dose at 70% coverage could avert 9066 hospitalizations (95% UI, 6736–10 511; NNT: 2273) and 1811 deaths (95% UI, 1394–2375; NNT: 11 378), corresponding to 21.0% and 31.1% of total hospitalizations and deaths. In persons 65 years and older, we predicted that an additional bivalent vaccine dose could avert 2725 hospitalizations (95% UI, 1923–3275; NNT: 698) and 690 deaths (95% UI, 527–855; NNT: 2757), corresponding to 6.3% and 11.9% of total hospitalizations and deaths.

Comparison of Nirmatrelvir-Ritonavir Strategies

We predicted the public health impact of additional uptake of nirmatrelvir-ritonavir and the number needed to avert 1 COVID-19 hospitalization and death in different epidemiologic groups defined by age and vaccine status (Table 3). For averting COVID-19–related severe outcomes (hospitalization and death), we predicted that increasing nirmatrelvir-ritonavir uptake to 70% in eligible populations could avert 4809 hospitalizations (95% UI, 3365–5813; NNT: 18) and 1292 deaths (95% UI, 798–1594; NNT: 67), corresponding to 11.4% and 23.5% of total hospitalizations and deaths. In persons 65 years and older, the model predicted increased uptake of nirmatrelvir-ritonavir could avert 4458 hospitalizations (95% UI, 3121–5388; NNT: 14) and 1252 deaths (95% UI, 778–1537; NNT: 50), corresponding to 10.5% and 22.8% of total hospitalizations and deaths.

Table 3.

Public Health Impact and Number Needed to Treat for Nirmatrelvir-Ritonavir During Coronavirus Disease 2019 Illness to Avert Hospitalizations and Death

Age Group–Based Strategy, Based on Current Eligibility	COVID-19 Hospitalizations			COVID-19 Deaths
	NNT	Total Averted	% Averteda	NNT	Total Averted	% Averteda
Strategy 1b,c (≥18 y)	18 (15–26)	4809 (3365–5813)	11.4% (7.7–14.0)	67 (54–108)	1292 (798–1594)	23.5% (13.1–30.0)
Strategy 2c (≥50 y)	17 (14–24)	4741 (3318–5731)	11.2% (7.6–13.8)	61 (50–98)	1289 (798–1590)	23.4% (13.1–30.0)
Strategy 3 (≥65 y)	14 (12–20)	4458 (3121–5388)	10.5% (7.1–13.0)	50 (41–79)	1252 (778–1537)	22.8% (12.7–29.2)
Strategy 4 (≥75 y)	10 (9–15)	3130 (2192–3782)	7.4% (5.0–9.1)	32 (27–51)	974 (607–1186)	17.7% (9.8–22.9)

All analyses compared 70% uptake to current uptake (30% in eligible COVID-19 cases without medical contraindications). Data on treatment effectiveness based on available evidence, although current estimates are limited for persons with varying degrees of prior infection, vaccination, and variants. Data is presented with 95% uncertainty intervals shown in parentheses.

Abbreviations: COVID-19, coronavirus disease 2019; NNT, number needed to treat.

^aThe denominator of this estimate accounted for baseline coverage of nirmatrelvir-ritonavir treatment.

^bPersons aged 18–49 years were deemed eligible for nirmatrelvir-ritonavir based on presence of immunocompromising conditions.

^cPersons aged 50–64 years were deemed eligible for nirmatrelvir-ritonavir based on presence of multiple comorbidities.

Sensitivity Analyses

When intervention strategies were stratified by age and vaccine status (Supplementary Tables 2 and 3), we identified that the highest-risk groups (eg, older groups who are unvaccinated) had the lowest NNT but were a small population. Estimates with perfect uptake of interventions is available in Supplementary Tables 7 and 8, with proportional increases in number of averted COVID-19 outcomes but stable NNT estimates. Sensitivity analyses with higher uptake in high-risk groups (eg, older age groups) and lower uptake in low-risk groups maximized overall number of averted severe COVID-19 outcomes (Supplementary Tables 9 and 10).

DISCUSSION

In this modeling study, we simulated the public health impact of increasing uptake of bivalent vaccines and nirmatrelvir-ritonavir treatment to avert COVID-19–related cases, hospitalizations, and deaths. In general, strategies that prioritized the most high-risk populations (persons ≥65 years, unvaccinated individuals) were effective prioritization strategies, with larger benefits from age-based strategies over vaccine status–based strategies. However, these targeted strategies may only avert a subset of total severe COVID-19 cases. Higher uptake of bivalent vaccines and nirmatrelvir-ritonavir had similar overall impact, although no strategy entirely averted the burden of severe COVID-19. The goal of this study is to help prioritize public health efforts to increase uptake of these interventions in the highest-risk groups.

We projected that for averting severe COVID-19 outcomes using bivalent vaccines, age group–based strategies performed better than vaccine status–based strategies based on NNT. Targeting vaccines to the unvaccinated population, which may be a challenging strategy, was an effective approach, but had a broadly comparable NNT to an age-based strategy of targeting the 50 years and older group. When comparing these 2 strategies based on absolute impact, targeting individuals 50 years and older had a similar impact compared to targeting the unvaccinated. These results suggest that age, rather than vaccination status, could be emphasized for guidance on prioritization of bivalent vaccines and may balance public health impact and feasibility.

We projected that nirmatrelvir-ritonavir treatment would be especially impactful for averting COVID-19–related hospitalizations and deaths in persons aged ≥65 years. The NNT estimate for averting COVID-19–related hospitalization or death was far lower for nirmatrelvir-ritonavir than for bivalent vaccines; this is largely because persons receiving this medication already have confirmed COVID-19 whereas vaccination is given to all persons. However, nirmatrelvir-ritonavir under this strategy of targeting the 75 years and older group exhibited higher overall impact when comparing the proportion of total outcomes averted, with bivalent vaccines averting 4.4% and 9.1% of total hospitalizations and deaths, while nirmatrelvir-ritonavir treatment averted 7.4% and 17.7% of total hospitalizations and deaths, respectively. This finding held true when we compared the other age group–based strategies, including targeting the 65 years and older group and the 50 years and older group. Our results suggest that while nirmatrelvir-ritonavir is a high-impact intervention (both in terms of NNT and absolute impacts), both bivalent vaccines and nirmatrelvir-ritonavir treatment are effective interventions. These findings are supported by other nirmatrelvir-ritonavir prioritization studies [36]. Benefit of vaccine includes potential to prevent SARS-CoV-2 infection and subsequent complications, rather than nirmatrelvir-ritonavir treatment that may just minimize severity. Furthermore, older and higher-risk adults are more likely to have contraindications to nirmatrelvir-ritonavir. We estimated that expanding eligibility for nirmatrelvir-ritonavir treatment to younger age groups had minimal benefit.

Our study has several limitations. Prospective prediction of COVID-19 outcomes is challenging and our parsimonious model relied on historical data that may not fully capture trends in future COVID-19 outcomes. However, the goal in this study was to compare treatment strategies between risk groups based on historical data; cumulative outcomes over the study period and relative difference between groups was most important to our analysis rather than trends in outcomes. The model validation analysis suggests the model may underpredict COVID-19 outcomes (if COVID-19 outcomes are lower during the calibration period, and there is a subsequent surge of cases during the prediction period), although the relative ranking of risk groups remained consistent overall. There are limited data on vaccine effectiveness of bivalent COVID-19 vaccines [6, 16], although we made the conservative assumption that these vaccines have at least comparable benefits to booster doses of monovalent vaccines. Benefits of vaccination may wane and new variants and complex immune landscapes may affect vaccine effectiveness or clinical severity [37–39]. We also conservatively assumed no additional benefit of 3 or more booster doses due to limited data on relative vaccine effectiveness, although future data can better inform this assumption. Some benefit is likely, which suggests that our results may be underestimating averted outcomes. We applied the best available evidence for nirmatrelvir-ritonavir treatment effectiveness, although current estimates are limited for persons with varying degrees of prior infection, vaccination, and new variants. Our models do not account for prior infection, immunocompromised status (which is a group widely heterogeneous in vaccine response) [40, 41], or age-related waning [42], which suggests that our results may overestimate averted outcomes in some cases. We applied the same vaccine-induced protection estimates in the immunocompromised population. We assumed fixed fractions of reporting for COVID-19 outcomes, although these may be variable with significant underascertainment of COVID-19 cases due to subclinical infection and at-home rapid antigen tests; our study was most designed to inform strategies to avert severe COVID-19. Our analysis focused on mRNA vaccination and did not include Novavax COVID-19 vaccine. The attribution of SARS-CoV-2 infection to reported COVID-19 hospitalizations and deaths remains controversial, although we followed current standard public health classification of these outcomes; future work can improve attribution of SARS-CoV-2 to these clinical outcomes. We did not include costs or a formal cost-effectiveness analysis. Finally, this modeling study did not account for vaccine-induced reductions to transmission (ie, indirect effects of vaccination) [43, 44] or reductions in long COVID and subsequent postinfection complications [45] and therefore is an underestimate of the total public health impact of each strategy.

In this study, our findings suggest that prioritizing uptake of bivalent vaccines and nirmatrelvir-ritonavir treatment among the oldest age groups would significantly and effectively reduce the number of severe COVID-19 infections in California but will not reduce the entire burden of severe COVID-19. This study provides evidence on the public health benefit of utilizing both interventions in the US and highlights potential opportunities for policymakers to improve the promotion and accessibility of these life-saving interventions.

Supplementary Data

Supplementary materials are available at Open Forum Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author contributions. H. J. P. and N. C. L. take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: R. S. and N. C. L. Statistical analysis: H. J. P. and N. C. L. Acquisition, analysis, or interpretation of data: All authors. First draft of the manuscript: H. J. P. and N. C. L. Critical revision of the manuscript: All authors. Contributed intellectual material and approved final draft: All authors.

Acknowledgments. We thank the California Department of Public Health for sharing the data used in this article and appreciate all the individuals involved in data collection and curation. We specifically appreciate assistance from the CDPH COVID-19 Data Processing and Informatics Section and COVID-19 Modeling Team.

Disclaimer. T. M. L., S. J., and R. S. are employees of CDPH and were involved in the analysis and interpretation of the data and the review and approval of the manuscript. The funder otherwise had no role in the design and conduct of the study; analysis and interpretation of the data; and decision to submit the manuscript for publication. This work represents the viewpoints of the authors alone and not necessarily those of the CDPH, California Health and Human Services Agency, or National Institutes of Health (NIH).

Financial support. This study is supported by funding from the California Department of Public Health. N. C. L. is supported by the National Institutes of Health NIAID New Innovator Award (DP2AI170485).