Modelling the household-level impact of a maternal respiratory syncytial virus (RSV) vaccine in a high-income setting

Background Respiratory syncytial virus (RSV) infects almost all children by the age of 2 years, with the risk of hospitalisation highest in the first 6 months of life. Development and licensure of a vaccine to prevent severe RSV illness in infants is a public health priority. A recent phase 3 clinical trial estimated the efficacy of maternal vaccination at 39% over the first 90 days of life. Households play a key role in RSV transmission; however, few estimates of population-level RSV vaccine impact account for household structure. Methods We simulated RSV transmission within a stochastic, individual-based model framework, using an existing demographic model, structured by age and household and parameterised with Australian data, as an exemplar of a high-income country. We modelled vaccination by immunising pregnant women and explicitly linked the immune status of each mother-infant pair. We quantified the impact on children for a range of vaccine properties and uptake levels. Results We found that a maternal immunisation strategy would have the most substantial impact in infants younger than 3 months, reducing RSV infection incidence in this age group by 16.6% at 70% vaccination coverage. In children aged 3–6 months, RSV infection was reduced by 5.3%. Over the first 6 months of life, the incidence rate for infants born to unvaccinated mothers was 1.26 times that of infants born to vaccinated mothers. The impact in older age groups was more modest, with evidence of infections being delayed to the second year of life. Conclusions Our findings show that while individual benefit from maternal RSV vaccination could be substantial, population-level reductions may be more modest. Vaccination impact was sensitive to the extent that vaccination prevented infection, highlighting the need for more vaccine trial data.

1. The age of each individual is incremented by one day.
2. For each individual i, one of the following events may occur: (a) Death: with a probability based on i's age and sex, i dies and is removed from the population. An individual j is chosen to be the mother of a replacement individual as follows: i. The target age of the mother is determined on the basis of age-specific fertility rates.
ii. A set of candidate mothers is determined on the basis of age, eligibility to give birth and household status (for simplicity, individuals are not eligible to give birth while living with their own parents).
iii. j is selected at random from the pool of candidate mothers.
If the death of i results in a household containing only children, these individuals are reallocated as follows: i. Any children aged 18 or older form new single-person households.
ii. Any children aged less than 18 are randomly allocated (fostered) to other households containing at least one child.
(b) Couple formation: if i is currently single, with a probability based on i's age, i forms a couple with an individual j, chosen as follows: i. A set of candidate partners is determined on the basis of age, sex, and not currently being a member of a couple.
ii. j is selected at random from the pool of candidate partners.
The households of i and j are merged (along with any children currently residing with them) or, if both previously lived with their parents, a new household of size two is created.
(c) Leaving home: if i is currently living with their parents, with a probability based on i's age, i leaves their parents' household and forms a new household of size one.
(d) Couple separation: if i is currently in a couple, with a probability based on i's age, i separates from that couple and forms a new household; for simplicity, we assume that any children residing with the couple when they separate join the mother's household.

Population model parameters and data sources
Mortality: Age-specific mortality rates for Australia were sourced from the Australian Bureau of Statistics [28]. For convenience, we assume that no individual survives beyond 100 years, and the probability of death at 100 years was fixed at 1.0.

Fertility:
Age-specific fertility rates for Australia were sourced from the Australian Bureau of Statistics [24]. These rates were not used directly to generate births in our model, but rather used to estimate relative probabilities of births being attributable to mothers of a particular age. When a birth event was triggered, these relative probabilities were used to ascertain the age of the mother, and hence the subset of the female population eligible to be randomly chosen as the mother.
Women without children: A subset of women in Australia never have children. We assigned a flag to 13% of females when they were born, so that they were never selected as a candidate mother. We based this percentage on the proportion of women in Australia who have no children . This estimation combined data reported both on rates of marriage and divorce with data on rates of de-facto relationships, as the primary focus of our model was the dynamics of household units, rather than the status of relationships. We assume that individuals become eligible to leave their parents' household, either independently or as a member of a couple, at 18 years. As a consequence, individuals also become eligible to separate from a couple at 18 years. We assume that individuals cease being eligible to form or separate from couples at 60 years.
The primary aims of the demographic model were to capture a reasonable approximation of the size and composition of households in the Australian population in 2017 and to execute in a computationally efficient fashion. It is not feasible that the model accurately capture all the demographic complexity of a real population and, as described above, several simplifying assumptions have been made in the name of model parsimony. For example, our model currently simulates dynamics of households containing one or two adults/parents (of opposite sex) and zero or more children (as defined by their familial relationship to the parents in the household; they may themselves be adults who are yet to leave home). Clearly, this does not exhaust the potential range of household types observed in real populations. Furthermore, our model does not include immigration, which may introduce individuals of a range of ages.

Generating the starting population
We generated a population of 100,000 people with an age distribution corresponding to that of Australia in 2017, obtained from the Australian Bureau of Statistics [25]. To do this, we first fit a Siler survival probability function to Australian survival probabilities obtained from the World Health Organization Life Tables [31]. We then used this function to calculate the number of births required in each of the preceding 100 years to produce the target age distribution, scaled to a population size of 100,000.
Once the population corresponding to 2017 had been generated, demographic rates were assumed to remain stable over the period covered by the scenarios compared in the paper, on the basis that fertility rates -a key demographic driver of epidemiological dynamics -have remained relatively stable (at just under 2 births per woman) over recent decades in Australia.   Values used for the parameter sweeps were q ∈ 0.006, 0.008, 0.01, 0.015, 0.02, 0.03 (plotted by colour), q h ∈ 0.5, 1.0, 1.2, 1.4, 1.6, 1.8, 2.0, 2.2, 2.4, 2.5, 5.0, 7.5 (plotted by size) and ω ∈ 0.1, 0.2, 0.3, 0.4 (plotted by shape). The desired ranges for the model outputs, the proportion of infections caused by siblings (0.35-0.50) and the infant incidence in the first year of life (60,000-70,000), are framed by the red box. The selected parameter combination used for model simulations was q = 0.015, q h = 2.4 and ω = 0.2, circled in red. We observed a negative correlation between infant incidence and the proportion of infections caused by siblings. Intuitively this makes sense: higher levels of incidence mean more infection is circulating in the community and thus it is more likely that an individual will be infected by a source outside of the household.