Background
In the Netherlands, hepatitis A virus (HAV) infection is not considered a major public health problem because of the low incidence of infection; the majority of new infections are observed in defined risk groups such as men who have sex with men (MSM) and travellers to countries where HAV is endemic. Due to improved socio-economic and sanitation conditions since the end of the Second World War and an increase in vaccine-induced protection in recent years [
1], the force of infection (FOI; the risk per susceptible person per unit time of becoming infected) has steadily declined [
2‐
4]. However, the size of the susceptible population and the likelihood of outbreaks occurring are expected to increase in the coming decades because of the cohort effect: younger birth cohorts have less exposure to HAV, and over time the segment of the population with natural HAV immunity is being replaced. Seroprevalence surveys conducted in 1995/6 and 2006/7 showed that 77% of persons in the Netherlands born before 1945 were anti-HAV positive [
1,
5], compared with fewer than 10% born after 1960 [
5].
Dynamic demographic processes such as mortality, fertility, and migration that influence the age-distribution of the population also have an impact on the projected number of HAV cases, and consequently the estimated population-level disease burden. Complications from HAV are more frequent and more severe in adults than in children [
6]; hence, any public health benefit from a decreasing trend in the force of infection may be offset by the increased morbidity and mortality expected with an older average age at infection.
In the current study we developed a simple deterministic sex- and age-structured model of the transmission dynamics of HAV, integrated with a dynamic model of demographic change. The transmission model was needed because exposure is birth cohort-dependent, under the assumption of a decreasing FOI that is attributable to improvements in hygiene. We calculated disease burden based on estimated annual incidence generated through simulation of HAV transmission over the period 2000–2030, to assess the interaction between the diminishing of natural immunity in the population (i.e., the replacement of immune cohorts) and population ageing on the burden of HAV in the Netherlands. We also compared the predicted burden associated with each of these two factors alone by removing the other from the model. Recently, HAV burden has been computed based on current notification data and estimates of under-reporting/under-ascertainment [
7], but forecasts have not been made. Our aim was not to precisely estimate the current or future disease burden, but to qualitatively investigate the differences in estimated burden under various transmission scenarios.
Discussion
The modelled burden of HAV in the Netherlands is projected to drop in the coming decades, from 310 DALYs per year in 2000 to about one-fifth this number in 2030. This is attributable to the exponentially decreasing FOI trend assumed in the baseline simulation. However, population ageing and consequent replacement of immune birth cohorts by cohorts with low natural immunity, and the relative increase of the size of the elderly population (for whom the risk of developing symptomatic infection and associated morbidity and mortality is greatest), had a compensatory effect with respect to the estimated future burden. A greater proportion of new HAV infections, and a consequent relatively stable level of DALYs over time, was projected for the older (55+ years) age-groups.
Incorporation of a realistic model of demographic change was influential. The difference between assumptions of a static and dynamic demography with respect to incidence was visible in the 55+ years age-groups. The age distribution of new infections in 2030 was shifted from persons aged 55+ years to the <55 years age groups in the static model variant, because there is a larger proportion of the population represented by those under 55 years in the static compared with the dynamic demographic model. Consistent with this, in the dynamic demography (baseline) model the drop in DALYs forecast for 2030 was mostly localised to persons under 55 years of age.
The greater total burden forecast for the dynamic compared with the static demography is due to the increasing proportion of the population represented by the elderly over time, and to the loss of natural immunity in the older cohorts. Through comparison of the projected total disease burden in the baseline simulation with the results of the simulation with no cohort effect (in which the initial proportion immune was assumed constant across all age-groups), the cohort effect had a slightly larger impact on the estimated disease burden for infections occurring in 2030 than did the ageing and growth of the population.
This baseline scenario assumed a steadily decreasing FOI over the simulation period, in which the decline in FOI associated with improvements in socio-economic conditions following the end of the Second World War was assumed to continue to drop at the rate of 5% per year. If a constant FOI over time is assumed instead (Scenario 2), then an increasing estimated future burden between 2000 and 2030 is forecast (Figure
5).
The impact of the projected increase in population-wide susceptibility due to the loss of immune cohorts was explored in three large outbreak scenarios. Simulation of an outbreak among the elderly (80+ years) had the least impact in terms of burden in the total population (239 DALYs in 2015–16). An outbreak in men aged 25–44 years was predicted to result in a greater total disease burden (265 DALYs in 2015–16), but a simulated outbreak in primary school-aged children aged 5–9 years was associated with the largest estimated burden (279 DALYs in 2015–16). This was not merely due to the relative size of the population represented by this age-group, as indicated by the DALYs per 100,000 measure.
Previous studies of the transmission dynamics of HAV infection have accounted for the cohort/natural immunity replacement effect [
18,
28]; although these studies used realistic age-structured models fitted to seroprevalence data, they have assumed a steady-state demography with predictions for population growth achieved via rescaling [
28]; population ageing was not taken into account. Strengths of our study are the simulation of demographic change and the estimation of the separate contributions from ageing and the replacement of naturally immune cohorts to the projected burden of disease.
There are several limitations to the current study. The first concerns how realistically the model represents the current epidemiological situation for HAV in the Netherlands. We assumed that all immunity in the population is naturally acquired; the current evidence points to a moderate proportion of immunity due to vaccination (recommended to travellers since 1994); 12.6% of the Pienter 2005/2006 survey participants had been vaccinated against HAV [
1]. Vaccination is also recommended for high-risk target groups, namely patients with chronic liver disease and Turkish and Moroccan children before travelling to their country of origin. MSM are offered an HBV vaccination and often choose to be vaccinated for HAV as well. We also assumed a fixed transmission rate for travel-related infection across age-groups and time, which is a clear simplification; the oldest age-groups may have a lower likelihood of travelling to endemic countries, and travel-related transmission may change over time due to changes in the frequency of travel to, and vaccination coverage in, destination countries with endemic HAV. Thus, modelled incidence may be too low if in the future there is an increased rate of travel to endemic countries, and/or prevalence in these countries does not improve. In recent years there has been a marked rise in the proportion of imported cases (between 31 and 51% of all notified infections in the period 2007–2010 were acquired outwith the Netherlands [
29]).
A second limitation concerns the adequacy of the transmission and disease burden models. Age-specific contact patterns were not incorporated, meaning that homogenous mixing between age-groups was assumed, a strong simplification. However, this should not be an issue, as our main goal was to estimate the HAV disease burden over time. The proportion of babies born immune was fixed instead of being dependent on the number of women of child-bearing age; however, because of the relatively rapid loss of maternal immunity this simplification has a minimal effect. We used a weighted average of disability weights according to the estimated overall distribution of acute illness severity [
7], and the same distribution was assumed for all age groups due to a lack of relevant data. We have also not modelled interventions such as liver transplantation which would reduce the number of fatal cases and thus the YLL component of DALYs.
Third, we have not attempted to simulate the effects of control measures that would come into force should an outbreak be detected, such as the vaccination of contacts. Our simulations thus represent the extreme situation in which no intervention takes place.
The scenario in which a constant FOI over time was simulated is perhaps unrealistic in the context of the widespread availability of vaccination and effective public health education. This scenario was useful, nevertheless, for illustrating how, in the absence of reduction in the transmission rate, the combination of demographic change and the replacement of naturally immune cohorts can predict a rising disease burden. However, our baseline scenario in which we assumed a 5% annual decline in the FOI was realistic; there is evidence that HAV incidence in the Netherlands, although relatively stable from the mid-1970s to mid-1990s [
26] is now steadily decreasing. Data on acute HAV cases retrieved from the Dutch national notification system indicate a decrease in HAV notifications between 1995 and 2005, from 6.5 to 1.3 per 100,000 [
9]. This likely reflects a combination of a reduction in importation (affecting the proportion of travel-related cases, model parameter φ), possibly attributable to improved vaccination rates among travellers or to reduced endemicity in the destination countries (affecting the transmission rate β
trav) [
30].
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
SM conceptualised the study, carried out the simulations and burden computations, and drafted the manuscript. MK conceptualised the study, advised on model design and parameterisation, interpreted the results, and edited the manuscript. M-JM, AS and EC interpreted the results and edited the manuscript. All authors read and approved the final manuscript.