Modeling Reemergence of Vaccine-Eliminated Infectious Diseases Under Declining Vaccination in the US

Key Points

Question

How will declining childhood vaccination rates affect the risk of outbreaks and reemergence of previously eliminated infectious diseases in the US?

Findings

At current state-level vaccination rates, measles may become endemic again; increasing vaccine coverage would prevent this. Under a 50% decline in childhood vaccination in the US, the simulation model predicted 51.2 million measles cases over a 25-year period, 9.9 million rubella cases, 4.3 million poliomyelitis cases, 197 diphtheria cases, 10.3 million hospitalizations, and 159 200 deaths.

Meaning

Childhood vaccination at a high coverage level is needed to prevent resurgence of vaccine-preventable infectious diseases and their infection-related complications in the US.

Abstract

Importance

Widespread childhood vaccination has eliminated many infectious diseases in the US. However, vaccination rates are declining, and there are ongoing policy debates to reduce the childhood vaccine schedule, which may risk reemergence of previously eliminated infectious diseases.

Objective

To estimate the number of cases and complications in the US under scenarios of declining childhood vaccination for measles, rubella, poliomyelitis, and diphtheria.

Design, Setting, and Participants

A simulation model was used to assess the importation and dynamic spread of vaccine-preventable infectious diseases across 50 US states and the District of Columbia. The model was parameterized with data on area-specific estimates for demography, population immunity, and infectious disease importation risk. The model evaluated scenarios with different vaccination rates over a 25-year period. Inputs for current childhood vaccination rates were based on 2004-2023 data.

Main Outcomes and Measures

The primary outcomes were estimated cases of measles, rubella, poliomyelitis, and diphtheria in the US. The secondary outcomes were estimated rates of infection-related complications (postmeasles neurological sequelae, congenital rubella syndrome, paralytic poliomyelitis, hospitalization, and death) and the probability and timing for an infection to reestablish endemicity.

Results

At current state-level vaccination rates, the simulation model predicts measles may reestablish endemicity (83% of simulations; mean time of 20.9 years) with an estimated 851 300 cases (95% uncertainty interval [UI], 381 300 to 1.3 million cases) over 25 years. Under a scenario with a 10% decline in measles-mumps-rubella (MMR) vaccination, the model estimates 11.1 million (95% UI, 10.1-12.1 million) cases of measles over 25 years, whereas the model estimates only 5800 cases (95% UI, 3100-19 400 cases) with a 5% increase in MMR vaccination. Other vaccine-preventable diseases are unlikely to reestablish endemicity under current levels of vaccination. If routine childhood vaccination declined by 50%, the model predicts 51.2 million (95% UI, 49.7-52.5 million) cases of measles over a 25-year period, 9.9 million (95% UI, 6.4-13.0 million) cases of rubella, 4.3 million cases (95% UI, 4 cases to 21.5 million cases) of poliomyelitis, and 197 cases (95% UI, 1-1000 cases) of diphtheria. Under this scenario, the model predicts 51 200 cases (95% UI, 49 600-52 600 cases) with postmeasles neurological sequelae, 10 700 cases (95% UI, 6700-14 600 cases) of congenital rubella syndrome, 5400 cases (95% UI, 0-26 300 cases) of paralytic poliomyelitis, 10.3 million hospitalizations (95% UI, 9.9-10.5 million hospitalizations), and 159 200 deaths (95% UI, 151 200-164 700 deaths). In this scenario, measles became endemic at 4.9 years (95% UI, 4.3-5.6 years) and rubella became endemic at 18.1 years (95% UI, 17.0-19.6 years), whereas poliovirus returned to endemic levels in about half of simulations (56%) at an estimated 19.6 years (95% UI, 14.0-24.7 years). There was large variation across the US population.

Conclusions and Relevance

Based on estimates from this modeling study, declining childhood vaccination rates will increase the frequency and size of outbreaks of previously eliminated vaccine-preventable infections, eventually leading to their return to endemic levels. The timing and critical threshold for returning to endemicity will differ substantially by disease, with measles likely to be the first to return to endemic levels and may occur even under current vaccination levels without improved vaccine coverage and public health response. These findings support the need to continue routine childhood vaccination at high coverage to prevent resurgence of vaccine-preventable infectious diseases in the US.

This study estimates the number of cases and complications in the US under scenarios of declining childhood vaccination for measles, rubella, poliomyelitis, and diphtheria.

Introduction

Since the early 1900s, vaccines have been developed for numerous infectious diseases and widely implemented across the US through routine childhood vaccination.^1,2 High childhood vaccination coverage has led to the elimination of several infectious diseases (defined as cessation of sustained local transmission, including for measles, rubella, diphtheria, poliomyelitis, and others, and the global eradication of smallpox).^1,2 Continued vaccination for many of these vaccine-eliminated infectious diseases focuses on maintaining high population immunity above a critical threshold to prevent disease reemergence.

Although the infectious diseases eliminated from the US continue to have sporadic occurrences of cases, the majority are due to importation of infections acquired elsewhere, often from an underimmunized US traveler returning from an endemic country.^3,4,5,6 Therefore, the US remains vulnerable to reintroduction and reemergence of vaccine-eliminated infectious diseases.

Vaccination rates in the US have been declining and there are now ongoing policy debates aimed at reducing the childhood vaccine schedule.^7,8 The declining vaccination rates accelerated since the start of the COVID-19 pandemic⁷ because of many factors, which include policy (eg, increased use of personal belief exemptions to childhood vaccine schedules), misinformation, distrust, and other societal and person-level factors.^9,10,11,12 This increasing antivaccine sentiment has coincided with an increase in the number of outbreaks and cases of vaccine-preventable diseases in the US. Since 2024, there has been an increase in the number of measles outbreaks (including a large outbreak that emerged in West Texas), with a significant number of pediatric hospitalizations.¹³

Recently, a larger US policy debate has begun on considering revision of the long-standing childhood vaccination schedule, including cessation of routine vaccine recommendations and removal of school-based vaccine mandates for currently eliminated diseases, which would lead to substantially lower vaccine coverage.^8,14

The objective of this modeling study was to better understand the potential long-term effects of declining or ceasing routine childhood vaccination for measles, rubella, diphtheria, and poliomyelitis to inform policy. Specifically, we estimate the risk of reemergence of previously eliminated infectious diseases and their major infection-related complications.

Methods

Model Description

We developed a simulation model to estimate the number of measles, rubella, poliomyelitis, and diphtheria cases under different scenarios of declining vaccination in the US. The stochastic, discrete time, age-specific, individual-based model simulated infection importation and dynamic transmission for each pathogen stratified across the 50 US states and the District of Columbia (hereafter, “states”) (additional details on the model appear in the eAppendix in Supplement 1 [under the “model overview” heading]).

At the start of the simulation, each individual in the population was assigned to an age group and a state of residence, along with a pathogen-specific immunity status determined by their age and state, which dictated their initial compartment assignment. Susceptible individuals who were exposed to infection progressed through the model’s individual health states from susceptible to exposed, infectious, and then recovered for each pathogen.¹⁵ Each pathogen was modeled separately with unique features and considerations (additional details appear in the Table and in the eAppendix and eTable 1 in Supplement 1).

Table. Epidemiological Model Inputs, Vaccination, and Clinical Complications for Vaccine-Eliminated Infectious Diseases.

	Vaccine-eliminated infectious disease
	Measles	Rubella	Poliomyelitis	Diphtheria
Model inputsa
Basic reproduction number16,17,18,19,b	12	4	4	2.5
Latent period, d20,21,22,c	10	12	5	3
Duration of infectiousness, d20,21,22	8	14	21	14
Infection importation rate, No. of cases/yd	34	5	0.2	0.2
Vaccine efficacy, %16,20,23,24,e	97	97	90-100	90-100
Population immunity25,26,27,28,29,30,31,32,33	Age- and state-specific	Age- and state-specific	Age- and state-specific	Age- and state-specific
Risk of infection-related complicationsf	Postmeasles neurological sequelae34: 0.1% Hospitalization20,35,36: 20% Death37: 0.3%	CRS38,39: 65% of pregnancies at 0-16 wkg Death6,40,41: 30% of CRS cases	Paralytic poliomyelitis4: 0.5% of unvaccinated cases Hospitalization4,20: 100% of paralytic polio cases Death42: 10% of paralytic polio cases	Hospitalization5,20: 100% of unvaccinated cases Death5,16: 10% of unvaccinated casesf

Abbreviation: CRS, congenital rubella syndrome.

^a Additional details on the model inputs and their ranges of values used in the probabilistic sensitivity analysis appear in the eAppendix in Supplement 1.

^b Refers to the mean number of secondary cases that an index case would infect in a fully susceptible population; the estimate shown is the assumed base case. As a guide, the basic reproduction number for seasonal influenza is generally estimated to be 1.1 to 2.0.

^c Mean time between exposure and onset of infectiousness.

^d Assumed for the entire US. A description of the data and references to support these assumed importation rates appear in the eAppendix in Supplement 1. Infection importation rate is further modeled specific to each state and its fraction of susceptible population over time.

^e Defined as protection against infection for measles and rubella. For diphtheria and poliovirus, the vaccines are assumed to provide complete protection against severe disease and reduce transmission by 90% but not generate protection against infection (eAppendix in Supplement 1).

^f Overall estimates are not age-specific, but are based on current clinical management, supportive care, and disease-directed therapy for hospitalized cases (eg, use of antibiotics and antitoxin therapy for diphtheria). Hospitalization and death from diphtheria and poliovirus are modeled only among unvaccinated and immunologically naive individuals, whereas for measles and rubella, individuals who are susceptible or lacking an immune response to the measles-mumps-rubella vaccine may be at risk of hospitalization or death if infected.

^g Assumes risk is 65% of rubella infections during pregnancy weeks 0 to 16, with pregnancy based on individuals aged 15 to 44 years and their probability of being pregnant.

Each state was modeled independently and had its own model parameterization; therefore, outbreaks from one state did not lead to cases or outbreaks in other states. At the start of the model creation (data inputted up to the end of 2024), those with immunity were assigned to the recovered compartment. The recovered compartment included individuals with vaccine-induced or infection-acquired immunity against infection, and assumed no waning of protection (eAppendix in Supplement 1).¹⁵ Vaccines for measles and rubella were assumed to generate all-or-nothing protection against infection with imperfect vaccine efficacy (the Table and eTables 2-3 in Supplement 1), whereas vaccination for poliovirus and diphtheria was modeled differently (eAppendix and eTables 4-5 in Supplement 1).^16,20,23,24

Poliovirus had a distinct model given vaccine-induced protection with the inactivated poliovirus vaccine (has been exclusively used since 2000 in the US) is mostly against clinical severity (ie, paralytic poliomyelitis) and transmission rather than against infection (eAppendix in Supplement 1).^21,23,42,43 We also simulated vaccine-derived protection for diphtheria to reduce clinical severity and transmission but not block infection or colonization (eAppendix in Supplement 1).^5,16,20 We used literature estimates for key natural history parameters (such as latent period and duration of infectiousness) for each disease (Table).

A key model input for each pathogen was the basic reproduction number, which guided the parameterization for the force of infection term for each pathogen, which was varied in the sensitivity analysis.^16,17,18,19 Transmission was modeled as a dynamic process, meaning that the overall risk of infection was related to the number of actively infectious persons in the population in a given state.¹⁵ The model accounted for age-specific heterogenous social mixing based on published contact matrices.⁴⁴ We accounted for protection generated by maternal immunity through passive transfer of antibodies for approximately 6 months (eAppendix in Supplement 1).^15,45 The model included infection-related complications for each pathogen (Table). We accounted for demography with state-specific birth rates and state- and age-specific death rates (eTable 6 in Supplement 1).⁴⁶

Given the infectious diseases have been eliminated from the US, the daily risk of infection importation for each infectious disease was modeled based on historical estimates of risk for infection importation (eAppendix in Supplement 1). Imported cases of infectious diseases occur most frequently from an underimmunized US traveler who acquires the infection while traveling in an endemic country, and then returns to the US with the potential to spread the infection.^3,4,5,42 We assumed the rate of infection importation for each infectious disease was Poisson distributed, and the infection importation rate was scaled proportionally as population susceptibility changed over the simulation period (eAppendix in Supplement 1). We modeled state-specific infection importation risk in relation to their total population, meaning that more populated states had a higher absolute risk of infection importation. We used total population as a broad surrogate for the amount of travel and infection importation risk (eAppendix in Supplement 1) but did not formally account for the effect of mobility or flight travel on infection importation risk.

Data, Model Parameterization, and Validation

We simulated the population for each state by using demography and unique population immunity profiles for each infectious disease. For demography, we included 2019 state-specific estimates from the US Census Bureau for birth rate, population size, and age distribution and used state- and age-specific death rates from the National Center for Health Statistics.⁴⁶ For generation of the age-specific profiles of population immunity, we used a combination of data from the National Immunization Surveys and data reported in published literature (additional information appears in the eAppendix in Supplement 1 and in the Table).^{25,26,27,28,29,30,31,32,33} We used state-specific estimates for population immunity (based on available vaccine coverage data from the National Immunization Surveys) for individuals 24 years of age or younger. For individuals older than 25 years of age, we assumed a general US-wide adult population immunity profile for each infectious disease informed by published literature.^{25,26,27,28,29,30,31,32,33} These data included seroprevalence estimates or vaccine data for the measles-mumps-rubella vaccine; the diphtheria, tetanus, and pertussis vaccine; and the inactivated poliovirus vaccine.

Each infection had unique pathogen characteristics, including a basic reproduction number, latent period, duration of infectiousness, and risk of complications (Table). Risks of infection-related complications were computed using literature-based risk estimates (ie, fraction of cases with the complication).

To evaluate model validity, we ran the model under current vaccination coverage levels over a 5-year period for each infectious disease. We then compared the estimated case numbers with publicly available observed case numbers reported by the US Centers for Disease Control and Prevention to assess how accurately the model reflects current transmission dynamics (eAppendix in Supplement 1).⁴⁷

Model Analyses and Scenarios

We simulated the models under various scenarios of changing childhood vaccination coverage in the US. Current state-specific baseline childhood vaccination coverage (defined as the mean childhood vaccination rate from 2004 to 2023) was compared with an estimated immediate decline by 5% to 100% (relative percentage reduction) in childhood vaccination coverage (eTable 7 in Supplement 1). The 100% reduction scenario modeled cessation of routine childhood vaccination. We also modeled an immediate increase in childhood vaccine coverage by 5% to 10%. Each scenario was modeled over a 25-year period to assess long-term transmission dynamics.

Model Outcomes

The primary outcomes were estimated infections with measles, rubella, poliomyelitis, and diphtheria. The secondary outcomes were estimated infection-related complications (postmeasles neurological sequelae, paralytic poliomyelitis, congenital rubella syndrome, hospitalization, and death) and the probability and timing for an infection to reestablish endemicity (eAppendix in Supplement 1).

For measles, the neurological sequelae included conditions ranging from primary measles encephalitis with persistent complications, acute disseminated encephalomyelitis, inclusion body encephalitis, and subacute sclerosing panencephalitis (universally fatal and often delayed 1-10 years after the primary infection).³⁴ Hospitalization for measles is often due to complications of pneumonia or dehydration, and more rarely, encephalitis. All clinical cases of paralytic poliomyelitis and respiratory diphtheria are recommended for hospital admission for inpatient management.

We estimated the absolute cumulative cases for each outcome under different vaccination scenarios. Given the different timing of outbreaks and return to endemicity due to stochasticity with some periodic behavior, we reported cumulative counts. Given the model was stochastic, we summarized the results over 2000 simulation runs. We report the mean (after winsorizing the top and bottom 2% of observations) and the 95% uncertainty interval (UI) based on the middle 95% that accounted for stochastic uncertainty.

We also identified the probability and timing for when an infectious disease returned to endemic levels under a given scenario (eAppendix in Supplement 1). This was based on when the infectious disease demonstrated sustained local transmission, conservatively defined as when an approximated national effective reproduction number of 1 or greater was sustained over a 12-month period.¹⁵ Alternative definitions for endemicity were also evaluated (eAppendix in Supplement 1).

Sensitivity Analysis and Uncertainty

We conducted a probabilistic sensitivity analysis using multiple model inputs that were simultaneously varied across a range of plausible values to determine their effect on the results (eTable 1 in Supplement 1). We varied the basic reproduction number, the population immunity profile, and the infection importation rate for each pathogen using Latin hypercube sampling. In addition, for poliovirus and diphtheria, we varied the transmission reduction factor due to vaccination. This generated a range of model outcomes around the primary study estimate, which accounted for parameter uncertainty in addition to stochastic uncertainty (eAppendix in Supplement 1). We subsequently estimated the partial-rank correlation coefficients for each model parameter to assess their relative effects on the outcome.

We also generated a sensitivity analysis in which the estimates for infection-related complications were varied (eAppendix in Supplement 1). We simulated several alternative model specifications for the base-case analysis, including a static risk of infection importation (ie, not proportional to the susceptible population), and assuming diphtheria vaccination protected against infection but not transmission.

This study was not human research given its use of publicly available aggregated data and computer simulation. The analytic codes and data are publicly available.⁴⁸

Results

At the start of the simulation (in 2024), we observed moderate state-level variation in childhood vaccine coverage (from 2004 to 2023) ranging from 88% to 96% for measles-mumps-rubella vaccination, ranging from 78% to 91% for diphtheria (the diphtheria, tetanus, and pertussis vaccine series), and ranging from 90% to 97% for the inactivated poliovirus vaccine series (see the eAppendix for vaccine definitions and eTable 7 for current vaccination rates by state in Supplement 1). Baseline immunity in children (3-4 years of age) ranged across the states from 85% to 93% for measles, from 85% to 93% for rubella, from 78% to 91% for diphtheria (against severe diphtheria), and from 90% to 97% for poliovirus (against paralytic poliomyelitis).

To evaluate model validity, we ran the model under a base-case scenario with recent state-level vaccine coverage levels over a 5-year period. The number of model-predicted cases broadly aligned with recent historical estimates accounting for natural stochastic variation in the estimates, supporting the validity of the current model (eFigure 1 and eAppendix in Supplement 1).

Over a 25-year period and under a scenario with current state-level vaccination rates, the simulation model predicted there would be 851 300 cases (95% UI, 381 300-1.3 million cases) of measles, 190 cases (95% UI, 154-230 cases) of rubella, 18 cases (95% UI, 3-61 cases) of poliomyelitis, and 8 cases (95% UI, 1-22 cases) of diphtheria. Under this scenario, we projected 851 cases (95% UI, 372-1270 cases) with postmeasles neurological sequelae, 170 200 hospitalizations (95% UI, 76 200-250 000 hospitalizations), and 2550 deaths (95% UI, 1130-3760 deaths) (Figure 1 and eTable 8 in Supplement 1). Under a scenario with state-level measles-mumps-rubella vaccination rates that were 5% higher over 25 years, there would be 5800 cases (95% UI, 3100-19 400 cases) of measles. With a vaccination rate that was 10% higher, there would be 2700 cases (95% UI, 2200-3400 cases) of measles. In contrast, under a scenario with a 10% decline in measles-mumps-rubella vaccination, the model estimates 11.1 million (95% UI, 10.1-12.1 million) cases of measles.

Figure 1. Cumulative Cases of Measles, Rubella, Diphtheria, and Poliomyelitis Over a 25-Year Period Under Different Scenarios of Routine Childhood Vaccination in the US.

The model estimated the number of cumulative cases for each infectious disease under different routine childhood vaccination scenarios (x-axis), including vaccine decline (5%-50%) and vaccine increase (5%-10%). Each vaccination scenario is a relative percentage of the current state-specific level of childhood vaccination (shaded area). The ranges for the current state-specific vaccination levels appear under the x-axes. The point estimates represent mean cumulative cases over the 25-year simulation period and the bars represent the 95th percentile of the stochastic uncertainty interval (UI) around the estimate. The y-axis is on a log scale. The median vaccination rate was 91.6% for measles and rubella, 93.0% for poliomyelitis, and 84.2% for diphtheria.

If routine childhood vaccination declined by 25%, we estimated 26.9 million (95% UI, 25.5 million-28.1 million) cases of measles would occur within the 25-year period, 790 cases (95% UI, 429-1700 cases) of rubella, 87 600 cases (95% UI, 3 cases to 1.7 million cases) of poliomyelitis, and 11 cases (95% UI, 1-52 cases) of diphtheria. Under this scenario, we projected there would be 26 900 cases (95% UI, 25 500-28 100 cases) of postmeasles neurological sequelae, 100 cases (95% UI, 0-1300 cases) of paralytic poliomyelitis, 1 case (95% UI, 0-3 cases) of congenital rubella syndrome, 5.4 million hospitalizations (95% UI, 5.1-5.6 million hospitalizations) due to all infections, and 80 600 deaths (95% UI, 76 500-84 300 deaths) due to all infections.

If routine childhood vaccination declined by 50%, the simulation model predicted there would be 51.2 million cases (95% UI, 49.7-52.5 million cases) of measles over a 25-year period, 9.9 million cases (95% UI, 6.4-13.0 million cases) of rubella, 4.3 million cases (95% UI, 4 cases to 21.5 million cases) of poliomyelitis, and 197 cases (95% UI, 1-1000 cases) of diphtheria. Under this scenario, we projected there would be 51 200 cases (95% UI, 49 600-52 600 cases) of postmeasles neurological sequelae, 5400 cases (95% UI, 0-26 500 cases) of paralytic poliomyelitis, 10 700 cases (95% UI, 6700-14 600 cases) of congenital rubella syndrome, 10.3 million hospitalizations (95% UI, 9.9-10.5 million hospitalizations) due to all infections, and 159 200 deaths (95% UI, 151 200-164 700 deaths) due to all infections (Figure 1, Figure 2, and eTables 8-9 in Supplement 1).

Figure 2. Cumulative Cases of Infection-Related Complications for Measles, Rubella, Poliomyelitis, and Diphtheria Over a 25-Year Period Under Different Scenarios of Routine Childhood Vaccination in the US.

The model estimated the cumulative cases of infection-related complications (y-axis) over 25 years under each scenario of change in childhood vaccination (see Table for model inputs) relative to the current state-specific levels of childhood vaccination (shaded area). The ranges for the current state-specific levels appear under the x-axes. The point estimates represent mean cumulative cases over the 25-year simulation period and the bars represent the 95th percentile of the stochastic uncertainty interval (UI) around the estimate. The y-axis is on a log scale. The median vaccination rate was 91.6% for measles and rubella, 93.0% for poliomyelitis, and 84.2% for diphtheria.

Further scenarios are provided in Figure 1 and in eFigure 2 in Supplement 1, including under complete cessation of routine childhood vaccinations. The cumulative counts of infection-related complications under each vaccination scenario are plotted in Figure 2. The timing and probability of the first infection-related complication varied by complication type and vaccination decline under the different scenarios (eFigure 3 in Supplement 1).

There were key differences in the timing of return to endemicity by disease and vaccine coverage scenario, with measles being the first to return to endemicity in most simulations (Figure 3). Under the status quo scenario (reflecting current childhood vaccination levels), measles was predicted to return to endemicity in 83% of simulations at an estimated 20.9 years (95% UI, 17.4-24.6 years), whereas the other diseases most likely would not return to endemicity. However, this finding changed under the scenario with 5% higher state-level vaccination coverage, in which measles did not return to endemic levels.

Figure 3. Predicted Probability and Timing for Return to Endemicity for Measles, Rubella, Diphtheria, and Poliovirus Over Time Under Different Scenarios of Declining Vaccination in the US.

The model estimated the probability of reestablishing endemicity for a given scenario over 2000 simulations; the location of the point estimate (in respect to the x-axis) represents the mean time to reestablish endemicity for the proportion of simulations that achieved endemicity over the 25-year simulation period (ie, this estimate for timing does not include scenarios when endemicity was not reestablished). Endemicity was defined as when the national approximated effective reproduction number reached 1 or greater over a 12-month period, suggesting sustained local transmission, which is a conservative definition. Alternative definitions for endemicity were also evaluated (eAppendix and eFigure 13 in Supplement 1). The size of the point estimate represents the probability of reestablishing endemicity for a given scenario over 2000 simulations. The horizontal bars represent the 95th percentile of the uncertainty interval around this estimate on time to endemicity (for the proportion of simulations that achieved elimination). Endemicity was reestablished for measles in a very high proportion of simulations, including under current levels of vaccination. Diphtheria was the least likely to reestablish endemicity. In the high vaccination scenario (≥5% change in coverage from current levels of vaccination), no disease reestablished endemicity. The median vaccination rate was 91.6% for measles and rubella, 93.0% for poliomyelitis, and 84.2% for diphtheria.

Under a 25% reduction in state-level vaccination coverage, poliovirus became endemic in 11% of simulations (mean time of 23.3 years), whereas diphtheria and rubella did not become endemic in any simulations. Under a 50% reduction in childhood vaccination, measles returned to endemicity in 99.8% of simulations at a mean time of 4.9 years (95% UI, 4.3-5.6 years), rubella returned to endemicity in 100% of simulations at 18.1 years (95% UI, 17.0-19.6 years), poliovirus returned to endemicity in 55.6% of simulations at a mean time of 19.6 years (95% UI, 14.0-24.7 years), and diphtheria returned to endemicity in less than 1% of simulations and took more than 2 decades (diphtheria did become endemic in a scenario with a larger vaccine decline). There was state-level variation in susceptibility to outbreaks and return to endemicity for each pathogen (Figure 4 and eFigures 4-10 in Supplement 1), and the population in Texas was at the highest risk for measles.

Figure 4. State-Level Heterogeneity in Cumulative Incidence of Measles, Rubella, Diphtheria, and Poliovirus Under a 25% Decline in Routine Childhood Vaccination Over 25 Years in the US.

The model generated estimates of cumulative incidence for each infectious disease under a scenario with a 25% decline in childhood vaccination by state over the simulation period. The rate is averaged over the entire 25-year simulation period. Other vaccination scenarios and pathogens appear in eFigures 4-6 in Supplement 1. Rubella and diphtheria cases were rare in this scenario; therefore, the y-axes for B and C have small increments.

In a sensitivity analysis, the most influential parameters on the study results were the pathogen basic reproduction number and the infection importation rate; other model parameters also contributed (eFigure 11 in Supplement 1). The overall key study conclusions were generally consistent across the sensitivity analyses (eFigure 12 in Supplement 1). With a higher basic reproduction number input, the critical population immunity threshold to prevent sustained transmission was higher, leading to a nonlinear increase in the number of cases and earlier return to endemicity under some vaccination scenarios.

A higher infection importation rate increased the chance of reestablishing endemicity after a state’s population immunity dropped below the critical threshold; therefore, a higher infection importation rate often led to higher cumulative cases and earlier return to endemicity. There was large stochastic variation due to the random timing of disease importation (modeled as a Poisson-distributed process) and subsequent spread of infection. There was much more uncertainty in the results for poliomyelitis and diphtheria, in part due to the rarity of importations for these infections.

We also found that assuming lower initial population immunity reduced time to return to endemicity in some scenarios. We compared alternative definitions of endemicity, which often estimated a modestly higher proportion of simulations leading to endemic transmission and earlier onset to endemicity (eFigures 13-14 in Supplement 1).

When infection importation was modeled as static (rather than dynamic), the results were similar (eFigure 15 in Supplement 1). When imported cases were distributed across age groups by population size (rather than by susceptibility), the results were similar (eFigure 16 in Supplement 1). When diphtheria vaccination was modeled as protective against infection (rather than against transmission), the results were similar (eFigure 17 in Supplement 1). When the measles basic reproduction number was varied, endemicity persisted at a basic reproduction number of 11 under current vaccine coverage levels; at a basic reproduction number of 10, a 5% decrease in vaccine coverage was needed for endemicity (eFigure 18 in Supplement 1).

Discussion

In this modeling study, we estimated how declining routine childhood vaccination would affect the risk of reemergence of previously eliminated vaccine-preventable infectious diseases in the US. This study focused on measles, rubella, poliomyelitis, and diphtheria, which had been eliminated from the US through widespread vaccination; these diseases were chosen because they are of high consequence to public health due to their infectiousness and risk of severe complications (including death).

We found that reductions in routine childhood vaccination will lead to reemergence and return to endemicity for all 4 infectious diseases under certain scenarios of vaccine decline; however, the timing and magnitude of case numbers and their critical population immunity threshold differed substantially by disease. Under widespread declining vaccination, measles is expected to be the first disease to have large outbreaks and return to endemicity. For rubella, poliovirus, and diphtheria, vaccine coverage levels would have to drop lower for a longer period before outbreaks and sustained transmission are observed.

A key finding is that, under current vaccination levels, measles may be likely to return to endemic levels within the next 20 years, driven by states with routine vaccination coverage below historical levels and below the threshold needed to maintain elimination of transmission. This study suggests that, over the long-term, the US may be approaching the threshold of losing sufficient population immunity to reliably prevent endemic transmission of measles, and higher childhood vaccination rates may be required to avert this. An important caveat is that our simulation study assumed constant levels of current childhood vaccination over time, without accounting for potential increases in childhood vaccination uptake in response to rising outbreak risk or significant public health response activities during outbreaks (such as reactive vaccination). One possible outcome may be that the US has a period of endemic transmission with measles followed by efforts to increase vaccine coverage and reachieving elimination status. Under current vaccination levels, the other infectious diseases (rubella, diphtheria, and poliovirus) are likely to maintain elimination status in the US; however, sporadic outbreaks of varying size and duration may occur after importation of these diseases. These findings underscore the importance of high levels of routine childhood vaccination to prevent reemergence of these eliminated infectious diseases.

The trends in decline and delay of vaccination in the US are complex; however, a key relationship has been increasing vaccine hesitancy and decline in routine childhood vaccination rates over time.^7,13 Recently, more abrupt changes to the childhood vaccine schedule have been debated, including removal of certain vaccines entirely.⁸ Even though a high level of population immunity remains due to decades of vaccination and, historically, infection-acquired immunity, declines in routine vaccinations will increase the number of susceptible persons over time and may lead to resurgence of many infectious diseases.

This study aimed to provide estimates for the infectious disease risks under different vaccine decline scenarios, accounting for state-level variation in demography, population immunity, and infection importation risk for each disease. Prior studies have assessed the relationship between declining vaccine coverage, including through nonmedical exemptions to childhood vaccine school entry requirements, and outbreak risk.^9,49,50 Other studies have documented some local transmission of other eliminated diseases, such as the 2022 case of poliomyelitis in New York City, which found evidence of subsequent transmission via poliovirus detection in New York State wastewater across many locations.⁵¹ Our study adds to this literature by estimating outbreaks, total cases, and chance of reestablishing endemicity for each infection across a broad range of scenarios projected over time in the US during a time of substantial vaccine decline and policy debate.

The risk and pattern of reemergence for each vaccine-eliminated infectious disease are unique. The differences are largely explained by a few key factors. First, the pathogen characteristics—primarily its infectiousness (summarized by its basic reproduction number)—broadly determines the threshold of population immunity required to prevent sustained transmission (colloquially often referred to as herd immunity). A higher basic reproduction number indicates a higher population immunity is required to prevent sustained transmission, and thus achieve elimination of the infectious disease. In our study, measles had the highest basic reproduction number of 12. Therefore, once population immunity drops below the corresponding critical threshold for herd immunity (approximately 92%), a nonlinear increase in cases may occur if endemic transmission is reestablished.

Second, the initial population immunity affects the risk of return to endemicity (related to historical vaccination). The higher population immunity against infection is above the critical threshold, the higher the buffer of protection exists before the consequences of reduced childhood vaccination, which decreases total population immunity, are realized.

Third, the risk of importation of an infectious case to the US differs dramatically between diseases. In the cases of these vaccine-eliminated infectious diseases, outbreaks only start when an individual is infected outside the US (index case) and then returns to the US and subsequently infects others. Most commonly, the index case is a US traveler who is underimmunized and is traveling outside the US in an endemic country and then returns to the US infected and has the potential to start an outbreak. Of the 4 infectious diseases assessed in the current study, measles is the disease most commonly imported, whereas poliovirus and diphtheria are the rarest.³ Additional factors related to heterogeneities in immunity, vaccine-derived protection (whether against infection or against disease severity and transmission), and other epidemiological considerations contribute to the current study findings.

These study estimates provide data to guide decisions on the importance of continued routine childhood vaccination to prevent infectious disease cases and their complications. Relatedly, these data can provide expectations on the risk of outbreaks and the timeline for a return to endemicity for different infectious diseases based on the degree of vaccination decline. Although many scenarios include wide 95% UIs inherent in rare chance events (eg, importation of poliovirus or diphtheria), the scenarios illustrate how different vaccine uptake can mitigate the mean and upper bound of the 95% UI (worst case scenarios) for cumulative cases and disease burden.

Once an infectious disease returns to endemic levels, reachieving elimination may be challenging without sufficient efforts to address the root cause of declining vaccination. As outbreaks become more common, case ascertainment is likely to be imperfect, meaning that more transmission is likely to be occurring than is being detected. Although rare, infection-related complications may become increasingly common under the most extreme scenarios of declining vaccination as endemic transmission is reestablished. In the case of rubella, once sufficient transmission is established, the risk of congenital rubella syndrome increases based on the degree of susceptibility due to underimmunization among individuals in the current population of childbearing age.

Limitations

This study has limitations. First, this modeling study used a simple modeling structure for these infectious diseases and made simplified assumptions in the simulation of disease transmission, immunity, social mixing, and demographic factors. We used the best available literature and data-informed estimates to parameterize the model; however, there was uncertainty and variation around these estimates. We generally selected conservative estimates for model inputs, meaning the risk of outbreaks and speed of reemergence under declining vaccination may be greater than predicted.

Second, we simulated state-level demography, immunity, vaccination, and transmission; however, there was likely substantial heterogeneity within the state level that was not captured (eg, heterogeneous vaccination coverage), and transmission is likely more focal. Within-state heterogeneity in these factors would generally lead to larger outbreaks and faster reemergence of endemic transmission. We modeled transmission within states as independent in the simulation, whereas outbreaks in one state would likely result in outbreaks in other states through travel of infectious persons; therefore, the base-case model is an underestimation of total cases given the lack of spillover from state to state.

Third, there was variation in the model estimate due to random chance, some of which was captured in the model’s 95% UIs. Pathogens with rare chance events (eg, importation of poliovirus or diphtheria) had larger stochastic uncertainty as a result. In these scenarios with large stochastic uncertainty, we focused attention on how each vaccine scenario affected the mean outcome and upper bound (worst case scenario), which is of most public health relevance.

Fourth, the infection importation rate is an important model parameter; however, there is likely imperfect measurement in surveillance data due to imperfect case ascertainment and reporting. We used estimates informed by the US and other similar countries. We assumed state-specific infection importation risk was related to the total population as a surrogate for travel, although inclusion of mobility and flight data would have improved the model.⁵²

Fifth, we modeled a homogenous relative decline in vaccination under different scenarios, but this may be more heterogenous by state. Sixth, vaccine-derived protection is complex for each pathogen and disease (especially poliomyelitis and diphtheria), including differences in protection against clinical severity, infection, and transmission. We used a simplified model with vaccine assumptions against each pathogen and performed sensitivity analyses on alternative modes of protection with similar findings (eAppendix in Supplement 1).

Seventh, we did not explicitly model isolation and quarantine, but our choice of basic reproduction number was based on relevant empirical data from the US and similar settings that implicitly account for these infection control measures. We conservatively chose values on the lower end of the literature estimates.

Eighth, we did not account for outbreak-related reactive vaccination from public health departments or behavior change on vaccine uptake during outbreaks. Ninth, we did account for age-specific social mixing patterns, but there was homogenous mixing within those groups. However, there may be assortative mixing by vaccine status, which could lead to earlier outbreaks and return to endemicity. We generated mean risk estimates for complications related to each disease, but did not account for age-specific differences in outcomes.

Tenth, we included both symptomatic and subclinical infections in our case estimates. In addition, the simulation was over a long-time horizon, which may make prediction more challenging, and we did not model vaccine-related adverse events, which are rare.⁵³

Conclusions

Based on estimates from this modeling study, declining childhood vaccination rates will increase the frequency and size of outbreaks of previously eliminated vaccine-preventable infections, eventually leading to their return to endemic levels. The timing and critical threshold for returning to endemicity will differ substantially by disease, with measles likely to be the first to return to endemic levels and may occur even under current vaccination levels without improved vaccine coverage and public health response. These findings support the need to continue routine childhood vaccination at high coverage to prevent resurgence of vaccine-preventable infectious diseases in the US.

Supplement 1.

eAppendix

eTable 1. Model inputs for probabilistic sensitivity analysis

eTable 2. Baseline immunity against measles by state and age group

eTable 3. Baseline immunity against rubella by state and age group

eTable 4. Baseline immunity against diphtheria by state and age group

eTable 5. Baseline immunity against poliovirus by state and age group

eTable 6. State-specific demography: Birth and age-specific annual death rates

eTable 7. Current routine childhood vaccination rate by state and vaccine for 1+ doses MMR, 4+ doses of DTaP, and 3+ doses of IPV at 35 months

eTable 8. Infection-related complications under different vaccine scenarios

eTable 9. Infection-related complications under lower and upper bound assumptions

eFigure 1. Evaluation of model validity

eFigure 2. Predicted cumulative cases of measles, rubella, diphtheria, and poliomyelitis over 25 years under different scenarios of childhood vaccination

eFigure 3. Probability and timing of at least one infection-related complication under different scenarios of childhood vaccination

eFigure 4. State-level heterogeneity in cumulative incidence of measles, rubella, diphtheria, and poliomyelitis under 50% lower levels of routine childhood vaccination in the United States over 25 years

eFigure 5. State-level heterogeneity in cumulative incidence of measles, rubella, diphtheria, and poliomyelitis under current levels of routine childhood vaccination in the United States over 25 years

eFigure 6. State-level heterogeneity in cumulative incidence of measles, rubella, diphtheria, and poliomyelitis under 25% lower levels of routine childhood vaccination in the United States over 25 years

eFigure 7. State-level cumulative incidence of measles under various levels of routine childhood vaccination in the United States over 25 years

eFigure 8. State-level cumulative incidence of rubella under various levels of routine childhood vaccination in the United States over 25 years

eFigure 9. State-level cumulative incidence of diphtheria under various levels of routine childhood vaccination in the United States over 25 years

eFigure 10. State-level cumulative incidence of poliomyelitis under various levels of routine childhood vaccination in the United States over 25 years

eFigure 11. Partial rank correlation coefficients of the probabilistic sensitivity analysis

eFigure 12. Results from probabilistic sensitivity analysis for cumulative cases compared to primary results

eFigure 13. Comparison of alternative definitions of endemicity for predicting probability and timing for return to endemicity for measles, rubella, diphtheria, and poliomyelitis

eFigure 14. Mean US population immunity and timing for return to endemicity in simulations with endemicity for measles, rubella, diphtheria, and poliomyelitis

eFigure 15. Sensitivity analysis with static infection importation rate

eFigure 16. Sensitivity analysis using population-weighted distribution of imported cases

eFigure 17. Sensitivity analysis on mode of vaccine-derived protection for diphtheria

eFigure 18. Sensitivity analysis on the basic reproduction number (R₀) for measles

eReferences

Supplement 2.

Data sharing statement