Estimating Vaccine-Preventable COVID-19 Deaths Among Adults Under Counterfactual Vaccination Scenarios in The United States: A Modeling Study Using Observational Data
Article Information
Ming Zhong1, Tamara Glazer1, Meghana Kshirsagar1, Richard Johnston1, Rahul Dodhia1, Allen Kim1, Divya Michael1, Santiago Salcido1, Sameer Nair-Desai2, Thomas C. Tsai3, 4, Stefanie Friedhoff2, William B Weeks*, 1, Juan M. Lavista1
1Microsoft AI for Good Lab, Redmond, WA
2School of Public Health, Brown University, Providence, RI
3Department of Health Policy and Management, Harvard T.H. Chan School of Public Health, Boston, MA
4Department of Surgery, Brigham and Women’s Hospital, Boston, MA
*Corresponding author: William B Weeks. Microsoft AI for Good Lab, Redmond, WA, USA
Received: 30 August 2023; Accepted: 06 September 2023; Published: 13 September 2023
Citation: Ming Zhong, Tamara Glazer, Meghana Kshirsagar, Richard Johnston, Rahul Dodhia, Allen Kim, Divya Michael, Santiago Salcido, Sameer Nair-Desai, Thomas C. Tsai, Stefanie Friedhoff, William B Weeks, Juan M. Lavista. Estimating Vaccine-Preventable COVID-19 Deaths Among Adults Under Counterfactual Vaccination Scenarios in The United States: A Modeling Study Using Observational Data. Journal of Pharmacy and Pharmacology Research. 7 (2023): 163-167.
Share at FacebookAbstract
Introduction: In early 2021, effective SARS-CoV-2 (COVID-19) vaccines became available in the United States; by mid-April 2021, vaccine availability outstripped demand, daily vaccination rates peaked, and COVID-19 vaccines were found highly effective in adult populations. Accurate estimates of the number of vaccine-preventable deaths had higher vaccination rates been attained could have helped local policymakers and possibly persuaded more to get vaccinated.
Methods: Because existing estimation methodologies are limited, for the period 1/1/21 – 4/30/22, we simulated the number of vaccine-preventable deaths associated with two-dose COVID-19 vaccination that incorporated state-level population, death, and vaccination numbers and three scenarios of vaccination rate achievement.
Results: Nationally, we found that had 100% of the population became fully vaccinated during the period examined, 318,979 deaths, or approximately 50% of reported COVID-19 deaths, might have been prevented; had 85% been so, 28% might have been prevented. Across states, we found substantial variation in the proportion of avoidable COVID-19 deaths that might have been avoided at the state level, from 25% in Massachusetts to 74% in Alaska.
Conclusion: Our findings are sobering when considering the number of deaths and diversion of scarce and expensive healthcare resources that might have been averted had peak vaccination administration efforts been maintained.
Keywords
COVID-19; vaccine-preventable deaths; modeling
Article Details
1. Introduction
In early 2021, effective SARS-CoV-2 (COVID-19) vaccines became available in the United States; by mid-April 2021, vaccine availability outstripped demand and daily vaccination rates peaked [1,2]. Early on, COVID-19 vaccines were found highly effective in adult populations: analysis of 50 real-world studies showed full vaccination was associated with ≥82%, ≥81%, and ≥94% reductions in hospitalization, severe disease or ICU admission, and death, respectively [3]. Availability of accurate estimates of the number of vaccine-preventable deaths if higher vaccination rates been attained might have helped local policymakers persuade more to get vaccinated. However, existing estimation methodologies rely on data from a limited number of states [4] or use a single vaccine maximization rate scenario [5]. To address these limitations, for the adult US population, we simulated the number of vaccine-preventable deaths associated with two-dose COVID-19 vaccination that incorporated state-level population, death, and vaccination numbers and three scenarios of vaccination rate achievement. We anticipated that models of increasingly higher vaccination uptake would generate estimates of increasing potentially avoidable deaths from COVID-19.
2. Methods
2.1 Data sources
We conducted a modeling study that relied on United States state-level observational data on adult COVID-19 deaths and vaccination rates that were publicly reported between 1/1/21-4/30/22. From the New York Times, [1] we obtained numbers of adult (aged 18+) COVID-19 deaths; from the Centers for Disease Control (CDC), [2] we obtained numbers of adults (aged 18+) who had received at least two COVID-19 vaccination injections.
2.2 Statistical methods
We developed a COVID-19 vaccine-preventable death estimation model based on the assumptions that a) once a state reached its peak vaccination administration capacity, it could maintain that level indefinitely, and b) the effective reproductive number of COVID-19 cases did not change. These are both conservative assumptions: as there were no local supply constraints, weekly vaccination administration might have increased; as more of the population becomes vaccinated, the effective reproductive number decreases. These assumptions allowed us to create a model that generates a lower-bound estimate of COVID-19 deaths had different proportions of the adult population been vaccinated following state-specific peak vaccination administration rates, as follows.
For a given state, the week when vaccine administration reached its peak marks the starting point of the counterfactual simulation. Until then, the estimated number equals the reported number of weekly deaths. After that, we assumed the number of vaccines administered remained at the state- and week-specific peak level until 85%, 90% or 100% of the state’s adult population was vaccinated. For every week after peak vaccine administration, we applied state-specific death rates of vaccinated and unvaccinated populations to the modeled size of those populations to estimate a counterfactual COVID-19 death count. Vaccine-preventable COVID-19 deaths were obtained by subtracting estimated counterfactual COVID-19 death count estimates from reported ones (Figure). We repeated this computation for each state and totaled the results to obtain state and national estimates of avoidable COVID-19 deaths between 1/1/21-4/30/22. We followed STROBE guidelines. No administrative permissions are required to access the data used in this study. Because we used aggregated, publicly available data, the study did not require human subjects approval. The authors did not have access to information that could identify individual participants during or after the study.
3. Results
Nationally, assuming 100% of the population became fully vaccinated during the period examined, we estimated that 318,979 deaths, or approximately 50% reported COVID-19 deaths, might have been prevented (Table); had 85% been so, we estimated that 28% might have been prevented. Across states, we found substantial variation in the proportion of avoidable COVID-19 deaths that might have been avoided at the state level, from 25% in Massachusetts to 74% in Alaska.
Table 1: State-level estimates of the avoidable number and percentage of COVID-19 deaths had 100%, 90%, or 85% of the adult population been vaccinated, at peak state-level vaccination administration rates, between 1/1/21-4/30/22.
State or district |
Population aged 18 and older |
Reported deaths |
Avoidable number and percentage of reported deaths had this proportion of the adult population been vaccinated: |
|||||
100% |
90% |
85% |
||||||
N |
% |
N |
% |
N |
% |
|||
Alabama |
38,11,122 |
14,695 |
6,527 |
44% |
5,358 |
36% |
4,671 |
32% |
Alaska |
5,51,685 |
1,005 |
748 |
74% |
550 |
55% |
451 |
45% |
Arizona |
56,36,931 |
20,890 |
9,427 |
45% |
6,883 |
33% |
5,579 |
27% |
Arkansas |
23,20,067 |
7,660 |
4,497 |
59% |
3,562 |
47% |
3,066 |
40% |
California |
3,06,18,852 |
63,877 |
21,730 |
34% |
13,944 |
22% |
10,034 |
16% |
Colorado |
44,97,682 |
7,390 |
4,575 |
62% |
2,994 |
41% |
2,199 |
30% |
Connecticut |
28,39,879 |
4,751 |
1,962 |
41% |
971 |
20% |
476 |
10% |
Delaware |
7,70,098 |
1,977 |
1,012 |
51% |
686 |
35% |
523 |
26% |
District of Columbia |
5,77,314 |
548 |
164 |
30% |
107 |
20% |
78 |
14% |
Florida |
1,72,34,469 |
52,059 |
29,200 |
56% |
20,405 |
39% |
15,918 |
31% |
Georgia |
81,07,219 |
25,737 |
13,598 |
53% |
10,549 |
41% |
9,021 |
35% |
Hawaii |
11,15,979 |
1,128 |
734 |
65% |
453 |
40% |
312 |
28% |
Idaho |
13,39,965 |
3,479 |
2,318 |
67% |
1,793 |
52% |
1,523 |
44% |
Illinois |
98,59,679 |
19,681 |
10,173 |
52% |
6,973 |
35% |
5,374 |
27% |
Indiana |
51,64,353 |
15,177 |
7,467 |
49% |
5,792 |
38% |
4,842 |
32% |
Iowa |
24,28,640 |
5,583 |
2,879 |
52% |
2,050 |
37% |
1,635 |
29% |
Kansas |
22,13,106 |
5,756 |
2,977 |
52% |
2,187 |
38% |
1,792 |
31% |
Kentucky |
34,64,554 |
12,419 |
7,154 |
58% |
5,373 |
43% |
4,480 |
36% |
Louisiana |
35,60,226 |
9,760 |
5,182 |
53% |
4,208 |
43% |
3,641 |
37% |
Maine |
10,95,969 |
1,928 |
1,092 |
57% |
529 |
27% |
247 |
13% |
Maryland |
47,13,082 |
8,489 |
3,876 |
46% |
2,331 |
27% |
1,560 |
18% |
Massachusetts |
55,39,447 |
7,761 |
1,957 |
25% |
1,180 |
15% |
793 |
10% |
Michigan |
78,46,375 |
22,706 |
12,950 |
57% |
9,566 |
42% |
7,873 |
35% |
Minnesota |
43,36,514 |
7,342 |
4,258 |
58% |
2,778 |
38% |
2,034 |
28% |
Mississippi |
22,78,354 |
7,604 |
3,302 |
43% |
2,795 |
37% |
2,509 |
33% |
Missouri |
47,66,387 |
14,289 |
8,585 |
60% |
6,595 |
46% |
5,571 |
39% |
Montana |
8,40,029 |
2,392 |
1,464 |
61% |
1,111 |
46% |
934 |
39% |
Nebraska |
14,59,044 |
2,489 |
1,456 |
58% |
1,030 |
41% |
817 |
33% |
Nevada |
23,85,240 |
7,617 |
4,223 |
55% |
3,120 |
41% |
2,556 |
34% |
New Hampshire |
11,04,426 |
1,711 |
926 |
54% |
607 |
35% |
447 |
26% |
New Jersey |
69,43,663 |
14,246 |
5,540 |
39% |
3,139 |
22% |
1,938 |
14% |
New Mexico |
16,19,953 |
4,965 |
2,467 |
50% |
1,493 |
30% |
1,005 |
20% |
New York |
1,54,17,677 |
29,830 |
11,195 |
38% |
5,983 |
20% |
3,392 |
11% |
North Carolina |
81,95,134 |
16,482 |
8,604 |
52% |
6,462 |
39% |
5,389 |
33% |
North Dakota |
5,81,523 |
998 |
650 |
65% |
496 |
50% |
418 |
42% |
Ohio |
91,09,592 |
29,411 |
15,875 |
54% |
11,956 |
41% |
9,996 |
34% |
Oklahoma |
30,06,685 |
11,744 |
5,833 |
50% |
4,384 |
37% |
3,654 |
31% |
Oregon |
33,52,774 |
5,999 |
3,798 |
63% |
2,449 |
41% |
1,773 |
30% |
Pennsylvania |
1,01,66,029 |
28,407 |
14,146 |
50% |
9,561 |
34% |
7,276 |
26% |
Puerto Rico |
26,18,532 |
2,681 |
1,232 |
46% |
650 |
24% |
359 |
13% |
Rhode Island |
8,54,658 |
1,763 |
611 |
35% |
279 |
16% |
113 |
6% |
South Carolina |
40,33,799 |
12,382 |
6,784 |
55% |
5,187 |
42% |
4,388 |
35% |
South Dakota |
6,67,244 |
1,411 |
754 |
53% |
543 |
38% |
438 |
31% |
Tennessee |
53,19,360 |
18,987 |
11,047 |
58% |
8,507 |
45% |
7,192 |
38% |
Texas |
2,16,13,563 |
59,795 |
29,773 |
50% |
21,741 |
36% |
17,692 |
30% |
Utah |
22,75,225 |
3,453 |
1,815 |
53% |
1,233 |
36% |
939 |
27% |
Vermont |
5,09,761 |
498 |
287 |
58% |
137 |
28% |
63 |
13% |
Virginia |
66,79,468 |
15,119 |
7,123 |
47% |
4,418 |
29% |
3,062 |
20% |
Washington |
59,52,622 |
9,223 |
5,299 |
57% |
3,207 |
35% |
2,156 |
23% |
West Virginia |
14,33,072 |
5,483 |
3,350 |
61% |
2,566 |
47% |
2,166 |
40% |
Wisconsin |
45,57,330 |
9,154 |
5,445 |
59% |
3,809 |
42% |
2,986 |
33% |
Wyoming |
4,44,716 |
1,374 |
938 |
68% |
744 |
54% |
647 |
47% |
Total |
25,78,29,067 |
6,41,305 |
3,18,979 |
50% |
2,25,424 |
35% |
1,77,998 |
28% |
4. Discussion
We found that sustaining peak vaccination administration rates might have prevented half of COVID-19 deaths during the period examined. Complemented by recent estimates that 235,000 deaths were avoided because of successful vaccination between 12/1/20-9/31/21, [6] our work helps articulate the public health costs associated with suboptimal vaccination coverage. While our findings likely reflect local decision-making regarding the balance of social, public health, and economic priorities, such decisions might have differed had studies like ours informed them. Our study was limited because we did not have access to age- or race-specific data, causing us to assume that vaccine effectiveness was the same for all, regardless of age or race. While vaccine effectiveness varies according to age and race, the state-specific unvaccinated- and vaccinated-specific death rates that we used were the reported rates at the time reflecting the experienced death rates of the vaccinated and unvaccinated populations. In addition, we did not have information on variants, waning immunity, or single dose vaccines. Finally, we only had access to reported deaths – it is possible that COVID-19 deaths were underreported during the period examined, possibly making the proportion of deaths potentially saved even higher than what we estimated. Despite these limitations, our findings are sobering when considering the deaths and diversion of scarce and expensive healthcare resources that might have been averted had vaccination administration efforts been maintained at peak levels.
Acknowledgements: All authors have contributed to writing, designing, compiling and editing the final manuscript
Data availability: The datasets analyzed during the current study are publicly available at https://github.com/nytimes/covid-19-data and https://data.cdc.gov/Public-Health-Surveillance/Rates-of-COVID-19-Casesor-Deaths-by-Age-Group-and/3rge-nu2a
Conflict of Interest: The authors have no conflicts of interest to report. As the study used publicly available data, IRB review was not required.
Disclosures: Views expressed in this manuscript are those of the authors and do not reflect the official views of the US government; the initial draft of this manuscript was prepared prior to the government service of Dr. Tsai and Ms. Friedhoff.
Competing Interests Statement: The authors have no conflicts of interest to report.
Author Contributions
All authors warrant authorship.
An authorship contribution form is below.
Criteria 1 |
Criteria 2 |
Criteria 3 |
Criteria 4 |
|||
Author initials |
Contributed to conception or design |
Contributed to acquisition, analysis, or interpretation |
Drafted the manuscript |
Critically revised the manuscript |
Gave final approval |
Agrees to be accountable for all aspects of work ensuring integrity and accuracy |
MZ |
X |
X |
X |
X |
||
TG |
X |
X |
X |
X |
X |
|
MK |
X |
X |
X |
X |
||
RJ |
X |
X |
X |
X |
||
RD |
X |
X |
X |
X |
||
AK |
X |
X |
X |
X |
||
DM |
X |
X |
X |
X |
||
SS |
X |
X |
X |
X |
||
SN |
X |
X |
X |
X |
||
TCT |
X |
X |
X |
X |
X |
|
SF |
X |
X |
X |
X |
||
WBW |
X |
X |
X |
X |
X |
|
JMLF |
X |
X |
X |
X |
X |
References
- New York Times. An ongoing repository of data on Coronavirus cases and deaths in the US.
- Centers for Disease Control and Prevention. Rates of Covid-19 cases or deaths by age group and vaccination status.
- Hatcher SM, Endres-Dighe SM, Angulo FJ, et al. COVID-19 Vaccine Effectiveness: A Review of the First 6 Months of COVID-19 Vaccine Availability (1 January-30 June 2021) Vaccines (Basel) 10 (2022).
- Scobie HM, Johnson AG, Suthar AB, et al. Monitoring Incidence of COVID-19 Cases, Hospitalizations, and Deaths, by Vaccination Status - 13 U S Jurisdictions, April 4-July 17, 2021. MMWR Morb Mortal Wkly Rep 70 (2021): 1284-90.
- Pierson E, Gerardin J, Lash N. The lives lost to undervaccination, in charts. The New York Times 2021 September 14 (2021).
- Steele MK, Couture A, Reed C, et al. Estimated Number of COVID-19 Infections, Hospitalizations, and Deaths Prevented Among Vaccinated Persons in the US, December 2020 to September 2021. JAMA Netw Open 5 (2022): e2220385.