Parental vaccination to reduce measles immunity gaps in Italy.

High-income countries are experiencing measles reemergence as the result of suboptimal vaccine uptake and marked immunity gaps among adults. In 2017, the Italian Government introduced mandatory vaccination at school entry for ten infectious diseases, including measles. However, sustainable and effective vaccination strategies targeting adults are still lacking. We use a data-driven model of household demography to estimate the potential impact on future measles epidemiology of a novel immunization strategy, to be implemented on top of the 2017 regulation, which consists of offering measles vaccine to the parents of children who get vaccinated. Model simulations suggest that the current vaccination efforts in Italy would not be sufficient to interrupt measles transmission before 2045 because of the frequency of susceptible individuals between 17 and 44 years of age. The integration of the current policy with parental vaccination has the potential to reduce susceptible adults by 17-35%, increasing the chance of measles elimination before 2045 up to 78.9-96.5%.

Introduction

The Global Measles and Rubella Strategic Plan 2012–2020 set the ambitious goal of eliminating measles in at least five World Health Organization (WHO) regions by 2020. Two years before the deadline, only the Americas have achieved measles elimination. Measles is endemic in 14 countries of the WHO European Region, including high-income countries such as Germany, Belgium, France, and Italy (World Health Organization Regional Office for Europe, 2016), and it still represents a major concern for public health.

In 2017, Italy experienced one of the largest measles outbreaks of the past decade in the European Region with four deaths and 5098 cases, 4042 of which were confirmed by positive laboratory results (Italian National Institute of Health, 2017; European Centre for Disease Prevention and Control, 2018). The highest incidence was observed in infants under one year of age. About 70% of the reported cases were older than 20 years, with a median age of 27 years (Italian National Institute of Health, 2017; Filia et al., 2017), suggesting that measles circulation in Italy is at least partially supported by transmission between adults. Significant immunity gaps in these age segments of the population have been highlighted by a serological screening of the population (Rota et al., 2008) and by recent modeling studies analyzing long-term processes that affect measles transmission dynamics in the Italian population (Merler and Ajelli, 2014; Trentini et al., 2017). The high fraction of measles-susceptible individuals of between 15 and 45 years of age is the result of past suboptimal routine vaccination coverage and the absence of major nationwide epidemics in recent decades, which allowed adolescents to escape both vaccination and natural infection (Filia et al., 2017; Trentini et al., 2017). In Italy, the first measles national immunization program was setup in 1983 with a single dose of measles vaccine being administered at 9 months of age. A second dose program was introduced in 1999. However, routine vaccination coverage remained below 80% until 2003, the year of approval of the Italian National Plan for the elimination of Measles and Congenital Rubella. Thereafter, vaccine uptake levels have progressively increased, even though a decrease in coverage has been detected in most recent years, possibly associated with vaccine hesitancy (Filia et al., 2017; Merler and Ajelli, 2014; Giambi et al., 2018). As a matter of fact, the national coverage reached a peak of 91% in 2010, which is well below the 95% threshold generally considered to be necessary for measles elimination (Anderson and May, 1991).

In July 2017, the Italian Government approved a regulation (119/2017) requiring parents to vaccinate their children before school entry against ten infectious diseases, including measles (Signorelli et al., 2018; D’Ancona et al., 2018; Italian Ministry of Health, 2017). Vaccination against measles is now free of charge and mandatory for all children under 16 years. Unvaccinated children are not allowed to attend kindergartens, and financial penalties are imposed on the parents of unvaccinated students attending higher school levels. This regulation has the potential to increase vaccine uptake in new birth cohorts and to immunize school-age children who have escaped routine vaccination (Trentini et al., 2019). However, the new policy will not impact the existing immunity gaps in older age groups. In particular, the achievement and maintenance of high vaccination coverage among children may not be enough to avoid the reemergence of measles in the future (Trentini et al., 2017; Trentini et al., 2019; Durrheim, 2017). In order to progress towards measles elimination, it is thus crucial for Italy to identify feasible, sustainable, and effective strategies to reduce the number of susceptible individuals among those who have already left the school system (Filia et al., 2017; Trentini et al., 2017; Durrheim, 2017; Thompson, 2017).

The aim of this work is to propose and investigate the effectiveness of a vaccination strategy to be introduced on top of the current policy. The proposed strategy consists of offering vaccination to the parents of all of the children who receive any measles vaccine dose.

Materials and methods

We simulated the socio-demographic structure of the Italian population over the 2017–2045 period, using an individual-based model of household generation and taking advantage of projections on the future evolution of the age distribution of the Italian population, as provided by the Italian National Institute of Statistics (ISTAT) (see Supplementary file 1) (Italian National Institute of Statistics, 2018; Billari et al., 2012). In the model, individuals are grouped into households following a heuristic approach similar to those previously introduced in the literature (see Appendix 1) (Fumanelli et al., 2012; Marziano et al., 2017).

The epidemiological status of the population is initialized at the beginning of 2017 using 100 stochastic realizations of the age-specific measles immunity profile estimated for Italy (Figure 1A) (Trentini et al., 2017). Measles vaccination between 2017 and 2045 is simulated by mimicking vaccination activities carried out during each year, taking into account the age and immunological status of each individual and keeping track of the vaccination history of the individual themselves and of her/his household members.

Figure 1.

Measles epidemiology under the current program (2017–2045).

(A) Mean measles age-specific epidemiological status as estimated by Trentini et al. (2017) at the beginning of the year 2017. Shown for each age is the percentage of individuals who are susceptible to infection or protected against infection by immunity provided by maternal antibodies and by immunity acquired through natural infection or routine (first or second dose) vaccination. (B) The age distribution of susceptible individuals at the beginning of 2017, as simulated in our model (orange), and the age distribution of suspected measles cases reported during 2017 to the National Measles and Rubella Integrated Surveillance System (green) (Italian National Institute of Health, 2017). (C) Mean yearly fraction of susceptible individuals in Italy as estimated by the model for the period 2017–2045 under the ‘current’ vaccination program. Different colors correspond to different age groups; vertical bars represent 95% confidence intervals (CI) of model simulations. (D) Mean measles age-specific epidemiological status as obtained by the model for 2045 under the ‘current’ vaccination program. Shown for each age is the percentage of individuals who are susceptible to infection or protected against infection by maternal antibodies and by immunity acquired through natural infection, routine (first or second dose) vaccination or vaccination at school entry.

Two vaccination programs are simulated. The first vaccination program, referred to as the ‘current’ program, corresponds to the vaccination policy currently in place in Italy, which consists of the routine vaccination of children at 15 months of age, the administration of a second booster dose at 5 years of age, and the check for compliance with this two-dose schedule at both pre-primary and primary school entry. Specifically, as a consequence of the 2017 regulation, children must have received one dose when entering pre-primary schools (at about 3 years of age) and two doses when entering primary schools (at about 6 years of age). The operating guidelines provided by the Ministry of Health also indicate the implementation during the transitional year 2017 of a catch-up campaign targeting all individuals below the age of 16 years who were not compliant with the two-dose schedule (D’Ancona et al., 2018; Italian Ministry of Health, 2018). Accordingly, in the model, routine vaccination with two-doses is performed every year while a catch-up campaign is simulated in 2017. In addition, from 2018 onwards, measles vaccine is annually offered at pre-primary and primary school entry (i.e., at 3 and 6 years of age) to all children who are not compliant with the routine schedule. Coverage levels for the first and second doses of routine vaccination are assumed to be constant over time and set equal to the most recent estimates of measles vaccination coverage at the national level: 85% and 83%, respectively (World Health Organization, 2016). In our simulation, the first dose is administered to children who have never been vaccinated and the second dose is administered to those who have only received one dose. We assume the same vaccination coverage for the 2017 catch-up campaign and for vaccination at school entry. In particular, for these vaccination activities, we assume a baseline coverage level of 50%, which corresponds to current estimates of the impact of the new regulation on measles vaccine uptake in the country (Italian Ministry of Health, 2019). Specifically, the coverage level at pre-primary school entry represents the percentage of vaccine uptake among 3-year-old children who have never been vaccinated, whereas the coverage at primary school entry represents the percentage of vaccine uptake among children who have received fewer than two doses.

A second vaccination program, referred to as ‘parental vaccination’, consists of the implementation from 2018 onwards of a novel strategy targeting the parents of vaccinated children on top of the current program. This supplementary immunization consists of offering measles vaccination to the parents of children who are receiving any dose of measles vaccine. In our simulation, parental vaccination is offered only once to each household, the first time the parents bring one of their children to be vaccinated under the current policy. In particular, we evaluate the impact of parental vaccination under three different coverage scenarios: 50%, 75% and 99%. These percentages represent the proportion of parents who are vaccinated as part of this strategy among all eligible parents, whose exact number depends on the coverage achieved in childhood vaccination programs. We assume that a single vaccine dose is offered to each parent during parental vaccination. In all the considered scenarios, measles vaccine efficacy is set at 95% (De Serres et al., 1995).

The effectiveness of each vaccination program is evaluated in terms of its impact on the overall and age-specific susceptibility to infection of the Italian population, on the effective reproduction number Re over the period 2017–2045, and on the amount of time required to achieve measles elimination. The effective reproductive number Re represents the expected number of secondary cases generated by one typical infected individual in a partially immune population, where the immunity within the population may be due to either vaccination or natural infection. Re provides important indications of the transmission potential of the virus in the population. If Re > 1 the infection may spread in the population; otherwise, the infection will die out. The year of measles elimination is here defined as the first year between 2017 and 2045, in which Re falls below 1.

Estimates of Re between 2017 and 2045 are obtained as follows:

we estimate the exponential growth rate r of the 2017 measles epidemic, by fitting a linear model to the logarithm of the weekly cases reported to the Italian National Institute of Health (Italian National Institute of Health, 2017; Chowell et al., 2004; Wallinga and Lipsitch, 2007);

we assume that the measles transmission dynamics follow a susceptible-latent-infectious-removed (SLIR) model and we adopt the Wallinga and Lipsitch approach (Wallinga and Lipsitch, 2007) to estimate Re in 2017 as , where 1/ω = 6.5 days is the average latent period and 1/γ = 7.5 days is the average infectious period, therefore considering an average generation time of 14 days (Anderson and May, 1991);

for each vaccination scenario and each year y between 2018 and 2045, we estimate the effective reproduction number  as the spectral radius of the next generation matrix encompassing information on the age-specific immunity levels resulting from vaccination and time varying demography, and the age-specific mixing patterns estimated for Italy (see Appendix 1) (Mossong et al., 2008; Diekmann et al., 1990; Diekmann et al., 2010).

The results presented in this paper are based on 1000 different model realizations for each vaccination scenario and include uncertainty regarding: the demographic projections of the age structure of the Italian population over the 2018–2045 period (Italian National Institute of Statistics, 2018); the age-specific measles immunity profiles estimated for Italy for 2017 (Trentini et al., 2017); the estimated growth rate r of the 2017 measles epidemic; and the age-specific mixing patterns of the Italian population (Mossong et al., 2008). Details are reported in Appendix 1 and all data required by our simulations are provided as Supplementary Files.

Sensitivity analysis

We perform different sensitivity analyses to assess the robustness of the obtained estimates when considering:

higher coverage for measles vaccination at school entry (75% and 99%, instead of 50% as assumed in the baseline analysis);

shorter/longer generation time for measles (10 and 18 days, instead of 14 days as assumed in the baseline analysis);

different assumptions on population mixing, including an alternative contact matrix estimated for Italy through a modeling approach (Fumanelli et al., 2012) and an homogeneous mixing in the population;

an alternative measles transmission model accounting for two distinct phases of infectivity.

In Appendix 1, we also report the results obtained when measles epidemiology is simulated by considering the vaccination strategy adopted in Italy before the introduction of mandatory vaccination at school entry in July 2017, and present a sensitivity analysis to assess the robustness of the estimates of the exponential growth rate r associated with the 2017 epidemic when including possible underreporting of cases (Ciofi Degli Atti et al., 2002).

Results

From the analysis of measles cases reported during the 2017 outbreak, we estimated an effective reproduction number of 1.66 (95% CI 1.55–1.76). According to our simulations based on the estimates provided by Trentini et al. (2017) at the beginning of 2017, 8.1% (95% CI 7.3–8.9) of the Italian population was susceptible to measles. About one third of the susceptible individuals were younger than 16 years, whereas 60% of them were aged between 18 and 45 years (Figure 1A). We estimate that the number of measles cases reported during the 2017 Italian outbreak represented only 0.1% (95% CI 0.09–0.12) of the susceptible population in Italy. This implies that, in Italy, about 4.9 million (95% CI 4.4–5.4) people may still be susceptible to measles infection. The age distribution of measles susceptible individuals matches the fraction of cases by age group reported during the 2017 outbreak (Figure 1B), thereby confirming the reliability of simulated immunity gaps in the population (Durrheim, 2016). According to our results, the catch-up campaign implemented in 2017 under the current program has contributed the immunization of 445189 (95% CI: 394797–487621) susceptible children under 16 years of age, producing a 9% reduction in the overall number of susceptible individuals in 2018 (Figure 1C). However, the obtained results show that after this initial drop, the overall fraction of susceptibles would progressively increase in the next decades, reaching 8.8% (95% CI 8.1–9.5) in 2045 (Figure 1C). This increase is ascribable to the replacement of elderly individuals, who are predominantly immune because of natural infection, with new birth cohorts that have been only partially immunized as a consequence of suboptimal coverage (Figure 1A and D). In particular, we estimate that in 2045 only 14.3% (95% CI: 12.5–15.9) of the susceptible population would be younger than 17 years, while individuals aged more than 45 years, who currently contribute only marginally to the residual measles susceptibility, would represent 69.6% (95% CI: 67.8–71.6) of the total number of susceptibles. As expected, the estimated percentage of susceptibles among individuals currently aged between 17 and 44 years would not be affected by this policy, remaining 16.2% (95% CI 14.5–17.7) in 2045 (Figure 1D).

The introduction of parental vaccination on top of the current program has the potential to progressively reduce the immunity gaps in adults as well as the overall susceptibility of the Italian population (Figures 2 and 3). Remarkably, the estimated total fraction of susceptible individuals in 2045 under the parental vaccination program ranges between 6.3% (95% CI: 5.8–6.9) for 99% vaccination coverage of parents and 7.6% (95% CI: 6.9–8.2) for 50% vaccination coverage, instead of 8.8% expected under the current program (Figure 3). This strategy targets age groups that would otherwise never be reached by the current immunization program, that is cohorts of individuals older than 16 years in the year 2017. In particular, by 2045, parental vaccination at 50% of coverage would result in a 17.1% (95% CI: 16.8–17.5%) reduction in the number of susceptible individuals aged between 17 and 44 years in 2017, while a 35% (95% CI: 34.1–35.9%) reduction in this age group is expected if 99% coverage is assumed.

Figure 2.

Impact of parental vaccination on the future age-specific immunity profiles.

Mean measles age-specific epidemiological status as estimated by the model for the year 2045 under different scenarios for the ‘parental vaccination’ program. Shown for each age is the percentage of individuals who are susceptible to infection or protected against infection by maternal antibodies and by immunity acquired through natural infection, routine (first or second dose) vaccination, vaccination at school entry or parental vaccination.

Figure 3.

Impact of parental vaccination on the proportion of measles-susceptible individuals.

Mean yearly fraction of susceptible individuals in the Italian population as estimated by the model for the period 2017–2045 under different scenarios for the ‘parental vaccination’ program. Different colors correspond to different age groups; vertical bars represent the 95% CI of the model simulations.

According to our analysis, the current program would decrease the measles effective reproduction number to 1.08 (95% CI 0.95–1.23) in 2045 (Figure 4). In our simulations, measles elimination is achieved before 2045 only in 12.0% of model realizations.

Figure 4.

Progress towards measles elimination.

(A) Mean effective reproduction number over time, as estimated by the model under the ‘current’ vaccination program and under different coverage scenarios for the ‘parental vaccination’ program. Shaded areas represent the 95% CI associated with model estimates. The red line represents the measles elimination threshold; elimination is achieved when the effective reproductive number is smaller than 1. (B) Probability associated with different time at measles elimination, as obtained by 1000 model realizations under the ‘current’ vaccination program and under different coverage scenarios for the ‘parental vaccination’ program.

The introduction of parental vaccination could accelerate the progress towards measles elimination, although the effectiveness of this strategy would depend on parents’ response to the new policy. Our simulations show that the effective reproduction number in 2045 would be 0.95 (95% CI 0.82–1.11), 0.89 (95% CI 0.77–1.07), 0.84 (95% CI 0.71–1.04) when 50%, 75% and 99% of eligible families accept parental vaccination, and that with these levels of vaccination acceptance measles elimination would be achieved on average in 2042, 2037 and 2031, respectively. Our results clearly show that parental vaccination has the potential to reduce the risk of major measles epidemics dramatically in the coming decades, although it is difficult to forecast the probability that measles outbreaks will be experienced in the future (details can be found in Appendix 1).

The performed analysis shows that an improvement of vaccine uptake at school entry to achieve 99% coverage in the current program may anticipate the timing of measles elimination to 2039. If vaccine uptake at school entry were to be 75%, measles elimination could be achieved in 2042, which is comparable to what might be obtained by reaching 50% of eligible families with parental vaccination in the baseline analysis. By contrast, our simulations show that, under the most optimistic scenario of 99% of coverage both for parental vaccination and vaccination at school entry measles elimination could be achieved, on average, as early as 2023.

The assumption of a shorter or longer generation time would affect model estimates of the effective reproduction number over time. In particular, under parental vaccination at 50% of coverage, a shorter (longer) generation time would result in an anticipation (delay) of the timing of measles elimination, which is estimated to occur before 2045 in 99.2% (16.8%) of model realizations. When a generation time lasting 18 days is considered, the current policy at current coverage levels was insufficient to achieve measles elimination by 2045 in 99.9% of model realizations.

The obtained estimates are qualitatively robust when considering alternative age-specific mixing patterns for the Italian population, although the inclusion of contact matrices estimated through the modeling approach (Fumanelli et al., 2012) results in delayed measles elimination under all considered vaccination scenarios. On the other hand, under the (hardly realistic) scenario of a population that mixes fully at random (i.e., by assuming homogeneous mixing), neither the current policy nor its combination with parental vaccination would be sufficient to achieve measles elimination by 2045.

Qualitative temporal patterns in the evolution of the effective reproduction number estimated by exploring different levels of measles transmissibility during the prodromal and exanthema phase are generally robust. The largest quantitative difference can be detected when most secondary cases are generated in the prodromal phase. In this case, under the current policy, measles elimination is predicted to occur before 2045 in 73.9% of model realizations instead of the 12.0% of model realizations seen for the baseline analysis. Similarly, under parental vaccination at 50% of coverage, when most of secondary cases are generated in the prodromal phase, measles elimination is predicted to occur before 2045 in 98.7% of model realizations compared to the 78.9% of model realizations seen in the baseline analysis.

Finally, when considering an extreme scenario in which only 25% of measles cases were reported during the 2017 outbreak, we estimate the exponential growth rate to be 0.29 (95% CI: 0.21–0.37), similar to that obtained when only reported cases are used: 0.29 (95% CI: 0.25–0.33). As estimates of the effective reproduction number depend only on the growth rate and on measles natural history, these results suggest that our findings are robust with respect to the reporting rate (and size) of the 2017 measles outbreak.

Details on the performed sensitivity analyses are reported and discussed in Appendix 1.

Discussion

In July 2017, the Italian Government approved a regulation requiring parents to vaccinate their children before school entry against ten infections, including measles. Recent estimates suggest that the new regulation allowed the vaccination of 50% of individuals who escaped routine vaccination (Italian Ministry of Health, 2019; Italian National Institute of Health, 2019). Our modeling study shows that the current policy would reduce measles susceptibility in the age segments of the population characterized by higher contact rates, resulting in a remarkable decrease in the infection transmission potential and making measles elimination a realistic target. However, if only 50% of unvaccinated children are vaccinated at school entry, disease elimination would probably be achieved only after 2045.

Offering vaccination to the parents of children who receive a measles vaccine dose could progressively reduce by 17–35% the immunity gaps among individuals who are between 18 and 45 years of age in 2018. The implementation of this program would decrease the overall susceptibility of the population by 6.2–22.0%, and would increase the probability of measles elimination before 2045 from 12.0% (estimated in the absence of this additional policy) to 78.9–96.5%. The effectiveness of this strategy clearly depends on both the coverage achieved through childhood immunization (routine programs and vaccination at school entry) and on the willingness of parents to be vaccinated themselves. The obtained estimates are generally robust with respect to different assumptions on the duration of measles generation time and on the relative transmissibility of measles during the prodromal and exanthema phases. On the other hand, under the assumption of homogeneous mixing in the population, neither the current immunization program nor parental vaccination appear to be sufficient to eliminate measles before 2045.

The study presents a few limitations that should be carefully considered in order to achieve a better interpretation of the obtained results. In particular, our estimates of the effective reproduction number were obtained using measles cases reported during the 2017 outbreak. The current degree of measles underreporting in statutory notifications is unknown. However, our estimates of the effective reproduction number are stable with respect to the possible underreporting of cases during the outbreak (Ciofi Degli Atti et al., 2002). The proposed analysis did not take into account potential geographical heterogeneities in measles immunity levels at the sub-national scale. Although the new regulation is expected to harmonize the vaccine offer and its uptake in Italy, significant regional differences in both immunization schedule and coverage have been reported in the past (Bonanni et al., 2015). Regions characterized by a lower than national average vaccine uptake in the past may therefore experience a delay in measles elimination with respect to the results presented in this work. In our work, future measles susceptibility might have been overestimated, as we did not explicitly model measles transmission, thus disregarding the impact of future measles spread on the immunity profile of the Italian population. Although both the occurrence and magnitude of future measles epidemics are largely uncertain and difficult to predict (Earn et al., 2000), changing patterns of measles transmission may affect both the number of susceptible adults and the incidence of severe disease in the coming years. However, the population infected during the 2017 Italian outbreak—one of the largest occurred in Europe in the last years—represented only 0.1% of the estimated susceptible population in the country (Italian National Institute of Health, 2017; Trentini et al., 2017). This suggests that the explicit inclusion of measles transmission may have a limited impact on short- or medium-term estimates of the immunity profile and measles transmission potential. The proposed analysis relies on the simplifying assumption that parents decide whether to vaccinate their children regardless of past vaccination behavior, although it is likely that parents vaccinate either all or none of their children. All children receiving vaccination indirectly present their parents with the opportunity to vaccinate themselves. As the children receiving vaccination may be clustered in a smaller number of households than is the case in our model, we are probably overestimating the potential number of parents who are eligible for measles vaccination. In particular, in our simulations, 98.7% of families with children between 1 and 15 years of age are considered as eligible for parental vaccination, whereas in a perfectly clustered model, this percentage would be 88.1%. On the other hand, clustering of unvaccinated children may have a larger effect on measles transmission dynamics than on the number of parents eligible for vaccination. Finally, we assumed that routine vaccination coverage would not be affected by the implementation of the new national policy, and that the coverage of the catch-up campaign conducted in 2017–2018 was the same as that of vaccination at school entry (both for pre-primary and for primary schools). However, data released in December 2018 by the Italian Ministry of Health suggest that the new regulation on mandatory vaccination at school entry may have indirectly affected the first-dose vaccine uptake for children under 3 years of age. In particular, the available records show that the first-dose vaccination coverage in the 2015 age cohort has increased from 91.4% in 2017 to 94.2% in 2018, (Italian Ministry of Health, 2019) although the fraction of unvaccinated children who were vaccinated thanks to the new regulation may vary depending on the age cohort considered (D’Ancona et al., 2018; Italian Ministry of Health, 2019). According to the most recent estimates, measles vaccination coverage in the 2014 age cohort has increased from 87.3% in 2016 to 94.4% in 2018, suggesting that the new regulation resulted in the vaccination of about 56% of unvaccinated children in this cohort (Italian National Institute of Health, 2019).

In conclusion, our analysis shows that a marked increase in childhood immunization rates would not be sufficient to achieve measles elimination in the short- or medium-term in Italy. These results confirm the need for appropriate strategies to vaccinate individuals who have already left the school system in order to reduce critical immunity gaps in young adults (Trentini et al., 2019; Filia et al., 2017; Trentini et al., 2017; Durrheim, 2017; Thompson, 2017; Wise, 2018; Gidding et al., 2007). Attempts made to date in this direction either have only been partially effective or have required remarkable efforts in terms of the costs to and commitment of the public health authorities (Gidding et al., 2007; Morice et al., 2003; Kelly et al., 2007). In Costa Rica, a measles-rubella vaccination campaign targeting adults aged 15–39 years was successfully conducted in 2000, but it required huge efforts of communication, social mobilization, and the use of house-to-house vaccination teams (Morice et al., 2003). In 2001–2002, a vaccination campaign in Australia targeting young adults aged between 18 and 30 years who visited their general practitioner (GP) had little effect on the immunity gaps, probably because of a lack of promotion and central coordination (Gidding et al., 2007; Kelly et al., 2007). In Europe, beyond some local attempts to immunize adolescents and individuals before school leaving, which have only marginally affected the vaccine uptake (Lashkari and El Bashir, 2010; Vazzoler et al., 2014), little has been done to reduce residual susceptibility in adults. Interventions recently set up include attempts to raise awareness in people attending social events that may represent potential hotspots for measles transmission (Public Health England, 2018). In this work, a new strategy is proposed, consisting of offering vaccination to parents of children who are being vaccinated against measles. Although the proposed policy can reach only a fraction of susceptible adults, that is those with children in the measles-vaccination age group, the obtained results suggest that this strategy may be both feasible and effective. In particular, our results suggest that vaccinating 50% of parents who agreed to vaccinate their children, and may therefore be inclined towards accepting vaccination, would promote measles elimination as well as reaching 50% of children who still escape measles vaccination despite the fact that vaccination is now mandatory in Italy (i.e., increasing vaccination coverage at school entry from 50% to 75%). The sustainability of the proposed strategy should be carefully evaluated by public health decision makers. However, a key advantage of this policy is that it does not require targeted activities to recruit parents, thus resulting in a relatively simple implementation protocol.

Beyond parental vaccination, alternative immunization strategies aimed at reducing residual susceptibility in adults may also be considered. These may include the extension of mandatory vaccination at university entry – an intervention already implemented in different US states. Other immunization efforts may include the introduction of proof of immunity as a condition for the enrolment of health care workers (HCWs), for whom measles vaccination is only recommended in most European countries (Galanakis et al., 2014; Maltezou et al., 2019). The need to improve vaccination coverage among HCWs is due to their potential to amplify measles outbreaks and their higher risk of exposure to the virus, as observed in the 2017 Italian outbreak in which 7% of cases were HCWs (Maltezou et al., 2019).

The achievement of measles elimination remains a global health priority. Actions may be also required to raise awareness and consensus about the benefits coming from vaccination and to increase the overall vaccine uptake. Country-specific policies should be identified and carefully evaluated by decision makers in order to anticipate the time of measles elimination as much as possible.

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Parental and mandatory school entry vaccination to reduce measles immunity gaps in Italy" for consideration by eLife. Your article has been reviewed by Neil Ferguson as the Senior Editor, a Reviewing Editor, and three reviewers. The following individuals involved in review of your submission have agreed to reveal their identity: Felicity Cutts (Reviewer #1).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This paper presents a modelling study of the impact of the new mandatory vaccination policy implemented in Italy since 2017 at school entry. The authors also explore the potential supplementation of the program with a campaign targeting parents. The impact is assessed using an individual-level model of household composition and simulating different vaccination policies up to 2045. The age specific susceptibility levels over the years are used to derived associated effective reproduction numbers and probabilities of outbreak.

Essential revisions:

The reviewers raise a number of concerns that must be adequately addressed before the paper can be accepted. Some of the required revisions will likely require further simulations and restructuring manuscript.

1) Restructure the manuscript in order to focus more on the third strategy (R+S+P).

2) Discuss what could be different strategies to reach older age groups (i.e. cohort with low level of immunity due to natural infection or vaccination) and their likely impact.

3) Ensure that the coverages described in the methods and used for the simulations are clearly defined in particular specify the denominators.

4) Do a sensitivity analysis regarding under reporting and size of 2017 outbreak; discuss plausibility of estimate in the Discussion section.

5) Specify if the age profile of observed cases used for validation of the model is based on laboratory cases.

6) Provide the ISTAT data as supplementary material to ensure permanency and provide a description of the assumptions of the model used for forecasting the demographic profile until 2045, this could be added to the Appendix.

7) Please incorporate uncertainty regarding demographic projections, estimation of r, contact matrices and initial susceptibility.

8) Include in the results a section on the sensitivity analysis based on results presented in the Appendix.

9) Carry additional sensitivity analysis with regards to estimation of R0.

10) Discuss the impact spatial heterogeneity in term of the predictions.

11) Please discuss the assumption behind the equation of the probability of an outbreak as a function of Re. A description in the Appendix of how the result is derived would be good. Is it the probability for a single introduction? How does it translate in terms of probability of a outbreak over a season? What is the impact of multiple reintroduction of measles?

Summary:

This paper presents a modelling study of the impact of the new mandatory vaccination policy implemented in Italy since 2017 at school entry. The authors also explore the potential supplementation of the program with a campaign targeting parents. The impact is assessed using an individual-level model of household composition and simulating different vaccination policies up to 2045. The age specific susceptibility levels over the years are used to derived associated effective reproduction numbers and probabilities of outbreak.

Essential revisions:

The reviewers raise a number of concerns that must be adequately addressed before the paper can be accepted. Some of the required revisions will likely require further simulations and restructuring manuscript.

We thank the reviewers and the editors for taking the time to review our manuscript and for the constructive feedback provided. We enclose here a point-by-point response to each of the reviewers’ comments. We have done our best to address all comments and suggestions by the reviewers. In particular, the major changes can be summarized as follows:

- We have restructured the manuscript to better highlight the impact of introducing parental vaccination on top of the current strategy. In particular, we decided to focus the main text on strategy “R+S+P” by assuming a baseline vaccination coverage at school entry of 50% on the basis of the latest available estimates on the impact of the new regulation in Italy. Alternative scenarios with different coverage for vaccination at school entry (namely, 0%, 75%, and 99%) were moved to the Appendix and are now presented as a sensitivity analysis. The title has been amended to reflect this change. The figures of the manuscript have been updated as well.

- Model estimates now include the uncertainty regarding: (i) the demographic projections of the age structure of the Italian population over 2018-2045; (ii) the age-specific measles immunity profile; (iii) the growth rate r of the 2017 measles epidemic; iv) the Italian contact matrices.

-We have added: (i) a sensitivity analysis to assess the robustness of the estimates with respect to possible underreporting and uncertainty about the size of the 2017 Italian outbreak; and (ii) a sensitivity analysis on the effective reproduction number. A new subsections entitled “Sensitivity analysis” has also been added to the main text.

- We have specified in the text all missing details about the data and model assumptions.

-The Italian population forecasts over 2018–2045 used in our analysis are now provided as a Supplementary file, as long as a brief description of methods adopted by ISTAT to produce these forecasts.

- We moved to the Appendix results concerning the probability of experiencing a measles outbreak and added an analysis showing how it depends on the number of measles introductions over the course of a season.

1) Restructure the manuscript in order to focus more on the third strategy (R+S+P).

In light of the reviewers’ comment, we have restructured the manuscript in order to highlight the impact of introducing parental vaccination on top of the current regulation (R+S). In particular, according to the latest available estimates, measles vaccination coverage in the cohort 2014 has increased from 87.3% in 2016 to 94.4% as of June 2018, suggesting that the new regulation contributed to vaccinate about 56% of unvaccinated children in this cohort. In light of this, we decided to present in the main text results obtained under program R+S, including routine vaccination in a two-dose schedule and assuming a baseline vaccination coverage at school entry of 50%. We now refer to this program as “current program”. Regarding parental vaccination, we explore three different scenarios obtained by adding parental vaccination on top of the “current program” and assuming three different coverage levels (50%, 75%, and 99%), as in the original manuscript.

Alternative scenarios of the parental vaccination on top of the “current program” with different coverage for vaccination at school entry (75% and 99%) and the scenario including routine vaccination only (R) were moved to the Appendix and are now presented as sensitivity analyses. The figures of the manuscript have been updated accordingly.

We would like to thank the reviewers for providing this suggestion, as we do believe that the manuscript has now a much better flow and it clarifies the most important aspects of our analysis.

2) Discuss what could be different strategies to reach older age groups (i.e. cohort with low level of immunity due to natural infection or vaccination) and their likely impact.

This is an interesting point of discussion. In the new version of the manuscript, we discuss possible strategies alternative to parental vaccination that could be used to reduce measles immunity gaps in older age groups. An alternative possible intervention consists in extending mandatory vaccination to universities. Indeed, two doses of the measles vaccination are already required in several US states for attending colleges and universities, and all students have to document immunity to measles before registering for classes. In Italy, to date, no vaccinations are required to enrol at universities. A second possibility is represented by targeting health care workers (HCW), for which measles vaccination is recommended, but not mandatory in most European countries including Italy [Galanakis et al., 2014]. We do believe that specific interventions targeting this subpopulation should be a priority, given the high number of measles cases reported among them in the 2017 Italian outbreak (315 out of 5,098 cases) and their potential to amplify measles outbreaks. Possible interventions in this direction include the performance of catch-up campaigns for HCW or the introduction of a proof of immunity as a requirement for admissibility to employment as HCW. The latter policy is already in place in some Italian regions (e.g., Puglia and Emilia Romagna) [Maltezou et al., 2019].

To address this point, we added the following sentences to the Discussion:

“Beyond parental vaccination, alternative immunization strategies aimed at reducing residual susceptibility in adults may be considered as well. These may include the extension of mandatory vaccination at university entry – an intervention already implemented in different US states. Other immunization efforts may include the introduction of a proof of immunity as a condition for enrolment of health care workers (HCWs), for which measles vaccination is only recommended in most of European countries [Galanakis et al., 2014, Maltezou et al., 2019]. The need for improving vaccination coverage among HCWs is due to their potential in amplifying measles outbreaks and their higher risk of exposure to the virus as happened in the 2017 Italian outbreak where 7% of cases were HCWs [Maltezou et al., 2019].”

3) Ensure that the coverages described in the methods and used for the simulations are clearly defined in particular specify the denominators.

We apologize for being unclear on this point, that is now specified in the section Material and Methods section as follows:

“Coverage levels for the first and second dose of routine vaccination are assumed to be constant over time and set equal to the most recent estimates of measles vaccination coverage at the national level – 85% and 83%, respectively. In our simulation, the first dose is administered to children who have never been vaccinated and the second dose is administered to those who have only received one dose.”

[…]

“Specifically, the coverage level at pre-primary school entry represents the percentage of vaccine uptake among 3 years old children who have never been vaccinated, whereas the coverage at primary school entry represents the percentage of vaccine uptake among children who have received less than two doses.”

[…]

“In our simulation, parental vaccination is offered only once to each household, the first time they bring one of their children to get vaccinated under the current policy. In particular, we evaluate the impact of parental vaccination under three different coverage scenarios: 50%, 75% and 99%. These percentages represent the proportion of parents who get vaccinated with this strategy among all eligible parents, whose exact amount depends on the coverage for childhood vaccination programs. We assume that one single vaccine dose is offered to each parent during parental vaccination.”

4) Do a sensitivity analysis regarding under reporting and size of 2017 outbreak; discuss plausibility of estimate in the Discussion section.

Following the reviewers’ suggestion, we have added a sensitivity analysis on underreporting and size of the 2017 outbreak. The current degree of reporting of measles in statutory notifications is unknown. A previous study estimating Italian measles incidence rates in the year 2000 from cases reported to a network of voluntary primary care paediatricians and from statutory notifications found that the former was 3.9 times higher than the latter [Ciofi degli Atti et al., 2002]. This would suggest a degree of reporting in statutory notifications of about 25%. The reporting rate of measles may have likely improved in the last decades. However, in the absence of updated estimates, we take this value as a worst-case scenario and assess the robustness of our results when accounting for a reporting rate of 25%.

In the new version of the manuscript, we have included an additional analysis where we estimate the attack rate and the growth rate associated with 1,000 synthetic time-series of measles infection cases generated on the basis of a MCMC approach applied to the likelihood of observing the original time series of reported measles cases when assuming a reporting rate of 25%. Under this assumption on measles reporting, the estimated overall number of measles cases in 2017 is 19,224 (95%CI: 18,724-19,745 vs 5,098 reported).

Remarkably, even when considering this worst case scenario on measles reporting, the estimated exponential growth rate obtained is 0.29 (95% CI: 0.21-0.37), similar to the one obtained when only reported cases are used: 0.29 (95%CI: 0.25-0.33). Since estimates of the effective reproduction number depend only on the growth rate and on measles natural history, this suggests that our results are robust with respect to reporting rate (and size) of the 2017 measles outbreak.

Methodological details on this new sensitivity analysis are reported in the revised version of the Appendix.

The outcome of this sensitivity analysis is commented in the Results section as follows:

“Finally, when considering an extreme scenario where only 25% of measles cases were reported during the 2017 outbreak, we estimate the exponential growth rate to be 0.29 (95% CI: 0.21–0.37), similar to the one obtained when only reported cases are used: 0.29 (95% CI: 0.25–0.33). Since estimates of the effective reproduction number depend only on the growth rate and on measles natural history, this suggests that our results are robust with respect to reporting rate (and size) of the 2017 measles outbreak.”

5) Specify if the age profile of observed cases used for validation of the model is based on laboratory cases.

We apologize for the lack of detail on this. The data used for model validation consist in the age-specific number of suspected measles cases reported to the National Health Institute.

In order to clarify this point, we have modified the caption of Figure 1 as follows:

“Age distribution of susceptible individuals at the beginning of 2017 as simulated in our model (orange) and age distribution of suspected measles cases as reported during the year 2017 to the National Measles and Rubella Integrated Surveillance System (green).”

Moreover, we have modified the following sentence in Introduction to update the estimated number of measles cases reported in 2017 and specify that 4,042 out of 5,098 suspected cases were lab-confirmed measles cases.

“In 2017, Italy experienced one of the largest measles outbreaks occurred during the last decade in the European Region with four deaths and 5,098 cases, 4,042 of which were confirmed by positive laboratory results.”

6) Provide the ISTAT data as supplementary material to ensure permanency and provide a description of the assumptions of the model used for forecasting the demographic profile until 2045, this could be added to the Appendix.

As suggested, the ISTAT data on the projections of the Italian population by age over the period 2017–2045 used in our simulations is now included Supplementary file 2.

Moreover, the procedure adopted by the ISTAT (Billari et al., 2012) to provide demographic forecasts of the population between 2018 and 2045 is now summarized in the Appendix as follows:

“Projections of the Italian population used in this study (see Supplementary file 2) are based on different stochastic realizations of official forecasts as provided by ISTAT and obtained through a method introduced in the literature by Billari and colleagues in 2012. This method relies on the framework of the so-called “random-scenario approach”, which is based on a series of subsequent expert-based conditional evaluations on the future evolution of different demographic indicators, given the values of the indicators at previous time points. Component-specific forecasts are combined and applied to an initial population (2017) to obtain different projections of the age-structure and overall size of the Italian population between 2018 and 2060.”

7) Please incorporate uncertainty regarding demographic projections, estimation of r, contact matrices and initial susceptibility.

We would like to thank the reviewers for pointing out this important methodological aspect. In the new version of the main text, we now present and discuss epidemiological results associated with different trajectories of the effective reproduction number (Re) as obtained by taking into account the uncertainty on:

1) initial measles susceptibility in 2017 as estimated by Trentini et al., 2017;

2) demographic trajectories of the Italian population age-structure as provided by available projections for the period 2017–2045;

3) contact patterns as resulting from a bootstrap procedure applied to the Italian POLYMOD contact matrix;

4) exponential growth rate r associated with the 2017 measles epidemic as resulting from the linear regression analysis.

Please note that in the original version of the manuscript we were using the contact matrix estimated for Italy by Fumanelli et al., 2012, which were obtained through the construction of a virtual population parameterized with detailed socio-demographic data. However, in order to appropriately include in our estimates, the uncertainty surrounding age-specific contact patterns observed in Italy, we have now considered the POLYMOD contact matrix for Italy [Mossong et al., 2008]. In particular, this allowed us to generate 1,000 bootstrapped contact matrices for Italy based on the publicly available individual contact diaries of the POLYMOD study.

Please note, results obtained in the original manuscript by using the contact matrix estimated by Fumanelli et al., have now been included as a sensitivity analysis on model results.

In order to better clarify all the different sources of uncertainty included in our estimates we have added the following sentence in the Materials and methods section:

“Results presented in this paper are based on 1,000 different model realizations for each vaccination scenario and include uncertainty regarding: the demographic projections of the age structure of the Italian population over 2018–2045; the age-specific measles immunity profiles estimated for Italy for 2017; the estimated growth rate r of the 2017 measles epidemic; and the age-specific mixing patterns of the Italian population.”

A description on how the different sources of uncertainty have been incorporated is also provided in the new version of the Appendix.

8) Include in the results a section on the sensitivity analysis based on results presented in the Appendix.

A new subsection “Sensitivity analysis” discussing the results of the performed sensitivity analyses has been added at the end of the Results section. In particular, we assess the robustness of our results with respect to:

1) Estimates of the effective reproduction number by considering scenarios with higher coverage for measles vaccination at school entry (75% and 99%, instead of 50% as assumed in our baseline analysis), either combined with parental vaccination or not.

2) Estimates of the effective reproduction number by assuming longer or shorter generation time (10 and 18 days, as compared with the 14 days assumed in our baseline analysis).

3) Estimates of the effective reproduction number under different assumptions on population mixing patterns in the form of an alternative contact matrix estimated for Italy [Fumanelli et al., 2012] and under the homogeneous mixing assumption.

4) Estimates of the effective reproduction number by explicitly accounting for two distinct phases of infectivity, instead of a single one.

5) Estimated exponential growth r when including possible under-reporting of cases for the 2017 outbreak (see reviewers’ comment #4).

Please note that sensitivity analysis 1, 3, and 4 were already present in the original version of the manuscript, but that new simulations have been performed to incorporate the different sources of uncertainty that are now accounted for in the baseline analysis (as detailed above – reviewers comment #7).

Details regarding the performed sensitivity analyses have been added in the Appendix but also mentioned in the main.

9) Carry additional sensitivity analysis with regards to estimation of R0.

In our modelling analysis, the estimate of the (effective) reproduction number depends on the estimated growth rate of the 2017 outbreak and on measles natural history. As such, we performed a new sensitivity analysis aimed at evaluating the estimates’ robustness to different assumptions on the duration of the generation time. Specifically, we considered a shorter and longer generation time (10 and 18 days) with respect to the value assumed in the baseline analysis (14 days).

We have now added a section at the end of the Results section in which we present estimates obtained through this sensitivity analysis:

“The assumption of a shorter or longer generation time would affect model estimates of the effective reproduction number over time. In particular, under parental vaccination at 50% of coverage, a shorter (longer) generation time would result in an anticipation (delay) of the timing at measles elimination, which is estimated to occur before 2045 in 99.2% (16.8%) of model realizations. When a generation time lasting 18 days is considered, the current policy, at current coverage levels, results instead insufficient to achieve measles elimination by 2045 in 99.9% of model realizations.”

Additional details and figures on the sensitivity analysis performed have also been added in the Appendix.

10) Discuss the impact spatial heterogeneity in term of the predictions.

We thank the reviewers for this useful comment, which allowed us to clarify a potential limitation of our analysis. Indeed, available epidemiological data and evidence show that Italian regions have experienced different immunization schedules and vaccination coverage in the last decades [Bonanni, 2015]. The new regulation on mandatory vaccination will hopefully promote a harmonization in the vaccine offer and uptake among children. However, different measles immunity levels have been detected across regions (e.g., Rota et al., 2008), so that regions that have been characterized by lower coverage in the past could experience a delay in measles elimination with respect to the results presented in this paper. On the other hand, we believe that all Italian regions (and possibly other countries as well) would benefit from the introduction of parental vaccination aimed at reducing measles immunity gaps in adults. The following sentence has therefore been added in the Discussion section:

“Although the new regulation is expected to harmonize vaccine offer and uptake in Italy, significant regional differences in both immunization schedule and coverage were reported in the past. Regions characterized by a lower than national average vaccine uptake in the past may therefore experience a delay in measles elimination with respect to results presented in this work.”

11) Please discuss the assumption behind the equation of the probability of an outbreak as a function of Re. A description in the Appendix ion of how the result is derived would be good. Is it the probability for a single introduction? How does it translate in terms of probability of an outbreak over a season? What is the impact of multiple reintroduction of measles?

We apologize for the lack of detail on this point. The equation used in the original version of the manuscript represents the outbreak probability after a single reintroduction and holds for stochastic SIR epidemics models (see for instance discrete time Markov Chains [Allen, 2008]). The equation can be generalized as p=1-(1/Re)n to obtain the probability of outbreak after n importations – either as separated events or not.

However, the estimated probability of an outbreak over a season can dramatically change under different assumptions on the numberof cases imported in the population over the considered period (e.g. in one year), which is a quantity that is difficult to properly estimate and forecast. As such, we acknowledge that the outbreak probability after a single reintroduction could represents a misleading measure of measles transmission potential. Thus, we decided to move results on this from the main text to the Appendix, by adding an appropriate discussion on how this probability would change when a different number of imported cases over a season is considered.

Please note that, as suggested, we added in the Appendix a description of the assumptions behind the equation p=1-1/Re, providing appropriate references and discussion on this point.