1. Medicine
  2. Epidemiology and Global Health
Download icon

Reducing respiratory syncytial virus (RSV) hospitalization in a lower-income country by vaccinating mothers-to-be and their households

  1. Samuel PC Brand  Is a corresponding author
  2. Patrick Munywoki
  3. David Walumbe
  4. Matthew J Keeling
  5. David James Nokes
  1. Zeeman Institute of Systems Biology and Infectious Disease Research (SBIDER), University of Warwick, United Kingdom
  2. School of Life Sciences, University of Warwick, United Kingdom
  3. Epidemiology and Demography Department, KEMRI-Wellcome Trust Research Programme, Kenya
  4. Mathematics Institute, University of Warwick, United Kingdom
Research Article
Cite this article as: eLife 2020;9:e47003 doi: 10.7554/eLife.47003
17 figures, 6 tables and 1 additional file

Figures

Schematic plot for the RSV transmission model and vaccination programme.

Infectious individuals (red character figures) transmit to other individuals inhabiting the same house, and to other individuals in other households based on the ages of the both the infector and infectee. Red and blue arrows represent possible realised infections over a short period of time. Bottom right household demonstrates the vaccination strategy; the mother has received a maternal antibody boosting (MAB) vaccine which increased transfer of protective antibodies to newborns (green background shading), meanwhile other household members have received an immune response provoking (IRP) vaccine (blue background shading).

RSV hospitalisation at KCH: dynamics and age profile of hospitalised patients.

(A) Weekly RSV hospitalisations before implementation of vaccinations. Black curve gives mean prediction of RSV household transmission model after regression against weekly incidence data (red dots). Grey shaded area indicates the 99% prediction interval for the model. Also shown is the number of under ones in the population (dashed line). (B) Age profile of hospitalisations at KCH before vaccination. Error bars give 99% prediction intervals for model.

Figure 2—source data 1

Hospitalisation data, and model predictions, are given as MATLAB data files along with script for plotting figure.

https://cdn.elifesciences.org/articles/47003/elife-47003-fig2-data1-v2.zip
Mean force of infection (2002–2016) between households and mean infection rates within households.

(A) The mean force of infection (infectious contacts received per person per day) of RSV due to transmission from without the household on three age groups: under-ones, school age children and everyone else, including adults. (B) Mean force of infection due to transmission without the household on individuals inhabiting each household size. (C) The mean per-capita daily rate at which different age groups become infected with RSV from within their household. (D) The mean total daily rate of RSV infection within households of different sizes.

Figure 3—source data 1

The model predictions are given as MATLAB data files, along with the script for plotting figure.

https://cdn.elifesciences.org/articles/47003/elife-47003-fig3-data1-v2.zip
Median forecast effectiveness of RSV vaccination for different mixed strategies over a 10-year period for 100% maternal vaccine effective coverage (A and C) and 50% maternal vaccine effective coverage (B and D).

(A and B) Median percentage reduction in hospitalisations at KCH. (C and D) Percentage reduction in total RSV infections in the population.

Figure 4—source data 1

Reductions in hospitalisations and infections for each of the 500 forecasting simulations are given as MATLAB data files, along with script for plotting figure.

https://cdn.elifesciences.org/articles/47003/elife-47003-fig4-data1-v2.zip
10-year forecast of RSV vaccination effectiveness for a mixed strategy of an MAB vaccine provided 75 days of additional RSV protection for newborns and a 75% IRP vaccine household coverage.

(A) Forecast weekly hospitalisations for a baseline of no vaccination (blue) and the mixed vaccination strategy (red). Shown are median forecast (curves) and 95% prediction intervals (background shading). (B) Forecast age distribution of total RSV hospitalisations at KCH. Median forecast (bars) and 95% prediction intervals (error bars).

Figure 5—source data 1

Hospitalisation predictions for each of 500 forecasting simulations is given as a MATLAB data file, along with a MATLAB function for combining the forecasting and Poisson hospitalisation rate uncertainties into a prediction interval and plotting script.

https://cdn.elifesciences.org/articles/47003/elife-47003-fig5-data1-v2.zip
Forecast vaccination efficiency against hospitalisations and all infections, defined as number of cases averted per vaccine used (both MAB and IRP).

MAB vaccine coverage was 100% unless unavailable, however MAB protection duration varied (different coloured bars) and IRP household coverage was also varied. See Table 1 for a list of scenario. (A) Median avoided hospitalisations at KCH per vaccine over 500 simulations. (B) Median avoided RSV infections in population per vaccine over 500 simulations.

Figure 6—source data 1

A MATLAB script for converting 500 forecasting simulation outcomes into efficiency metrics, and plotting them.

https://cdn.elifesciences.org/articles/47003/elife-47003-fig6-data1-v2.zip
Appendix 1—figure 1
Distribution of peak month for RSV hospitalisations at KCH.
Appendix 2—figure 1
Growth in number of possible household configurations as complexity of the underlying age-and-disease state model grows.

Calculated for a maximum household size of 10.

Appendix 2—figure 2
Schematic diagram of the basic age-and-disease state compartmental model for the individuals inside the households.
Appendix 2—figure 3
Household occupancy characteristics calculated on each 1 st Jan 2000–2017.

Top: Percentage of U1s in households of a certain size or smaller. Middle: Percentage of U1s in households with only one U1 and households with one or two U1s. Bottom: Household size distribution.

Appendix 2—figure 4
Comparison of numbers of households of sizes 1–10 on each 1 st Jan 2000–2017 (dots) against simulated values (curve).

Simulation is from Sept 2001 - Sept 2016. Horizontal axis is days since 1 st Jan 2000.

Appendix 2—figure 5
Comparison of total numbers of U1s and O1s on each 1 st Jan 2000–2017 (dots) against simulated values (curve).
Appendix 3—figure 1
Ratio of KHDSS residents to non-residents weekly accessing KCH for confirmed RSV treatment.

Red curve is polynomial fit R(t).

Appendix 3—figure 2
Plots of fitted weekly hospitalisations and the age distribution of hospitalisations for four scenarios (differing values of the schools based baseline RS).

In each case, parameter inference was performed and the maximum likelihood estimators used.

Appendix 3—figure 3
Maximum likelihood parameters for the different school transmission rate scenarios.

bU1, bO1 are respectively the under-one and over-one mixing components of the community mixing rate matrix. τ is the rate at which a household member infectiously contacts each other household member. M¯=1/α is the mean period of maternal protection after birth. m=(mξ mϕ) is the mean vector of the random seasonality, and σξ, σϕ and ρξϕ are respectively the standard deviations of the seasonal amplitude, seasonal phase and the correlation between the two, derived from the estimated covariance matrix Σξϕ.

Appendix 4—figure 1
Vaccine effectiveness for the four school mixing scenarios at 100% MAB coverage.
Appendix 4—figure 2
Colorblind-friendly version of Figure 4 from main text.

Forecast effectiveness of RSV vaccination for different mixed strategies over a 10 year period for 100% maternal vaccine effective coverage (A and C) and 50% maternal vaccine effective coverage (B and D). (A and B) Percentage reduction in hospitalisations at KCH. (C and D) Percentage reduction in total RSV infections in the population.

Tables

Table 1
Modelled vaccination scenarios.

Each combination of MAB vaccine effectiveness and coverage, with IRP vaccine coverage below was one scenario. The baseline scenario being no effective MAB vaccine and 0% coverage of IRP vaccine.

DescriptionRange
Additional period of protection from RSV at birth due to maternal antibody boosting (MAB) vaccine (P).0 (no vaccine), 15, 30, 45, 60, 75, 90 days
Coverage of mothers with MAB vaccine50%, 100%
Coverage of households with newborns with immune response provoking (IRP) vaccination (Vcov)0%, 25%, 50%, 75%, 100%
Table 2
Parameters from literature and chosen for model.
ParameterDescriptionValueData source
σO1Susceptibility (O1s)0.75Henderson et al., 1979
ι2relative infectiousness (O1s)0.5Kinyanjui et al., 2015
νRate of waning of immunitytwo per yearAgoti et al., 2012
γ1Rate of recovery for under-ones1/9 per dayHall et al., 1976
γ2Rate of recovery for over-ones1/4 per dayHall et al., 1976
bSCommunity transmission rate at schools0,1/3,2/3,1 per dayrange
ηAgeing rate for U1sone per yearmodel choice
ϵBase external infection rate (whole population)10 per daymodel choice
Table 3
Inferred parameters.
ParameterDescriptionValue
bU1Community transmission rate for U1s0.22 [0.18,0.27] per day
bO1Community transmission rate for O1s0.20 [0.18,0.21] per day
τTransmission rate to each other member of household0.040 [0.032, 0.048] per day
M¯Mean duration of maternal protection at birth21.6 [17.2, 26.1] days
mξMean amplitude of log-seasonality0.61 [0.51, 0.72]
mϕMean timing of log-seasonality peak (phase)67.7 [40.2, 77.7] days
σξStd. amplitude of log-seasonality0.20 [0.098,0.31]
σϕStd. timing of log-seasonality peak (phase)38.7 [30.0, 48.5] days
ρξϕCorr. coefficient between log-seasonal amplitude and phase−0.035 [-0.12, 0.072]
Appendix 3—table 1
Parameters estimated from KHDSS data.
ParameterDescriptionValueData source
μ(n, t)Birth/turnover rate for households of size n on day tVaries, see aboveKHDSS
r(n, t)Rate of change of numbers of households of size n on day tVaries, see aboveKHDSS
PH→A,tConditional age distribution given household config. on day tVaries, see aboveKHDSS
PA→H,tConditional household config. distribution given age category on day tVaries, see aboveKHDSS
Appendix 3—table 2
Age-dependent hospitalisation probabilities per infection derived from Kinyanjui et al., 2015.
Age categoryProbability of hospitalisation per infection
0-1 month0.10
1-2 month0.10
2-3 month0.063
3-4 month0.059
4-5 month0.054
5-6 month0.025
6-7 month0.019
7-8 month0.022
8-9 month0.012
9-10 month0.016
10-11 month0.013
11-12 month5.1x10−3
1-2 years old2.6x10−3
2-3 years old7.5x10−4
3-4 years old2.2x10−4
4-5 years old3.8x10−5
Appendix 3—table 3
Model parameters inferred from hospitalisation data.
bU1Community transmission rate for
U1s
0.22 [0.18,0.27] per day
bO1Community transmission rate for
O1s
0.20 [0.18,0.21] per day
τTransmission rate to each other
member of household
0.040 [0.032, 0.048] per day
MMean duration of maternal protec-
tion at birth
21.6 [17.2, 26.1] days
mξMean amplitude of log-seasonality0.61 [0.51, 0.72]
mφMean timing of log-seasonality
peak (phase)
67.7 [40.2, 77.7] days
σξStd. amplitude of log-seasonality0.20 [0.098,0.31]
σφStd. timing of log-seasonality peak
(phase)
38.7 [30.0, 48.5] days
ρξφCorr. coefficient between log-
seasonal amplitude and phase
-0.035 [-0.12, 0.072]

Additional files

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)