An intercountry comparison of the impact of the paediatric live attenuated influenza vaccine (LAIV) programme across the UK and the Republic of Ireland (ROI), 2010 to 2017

Abstract Background The universal paediatric live attenuated influenza vaccine (LAIV) programme commenced in the United Kingdom (UK) in 2013/2014. Since 2014/2015, all pre‐school and primary school children in Scotland and Northern Ireland have been offered the vaccine. England and Wales incrementally introduced the programme with additional school age cohorts being vaccinated each season. The Republic of Ireland (ROI) had no universal paediatric programme before 2017. We evaluated the potential population impact of vaccinating primary school‐aged children across the five countries up to the 2016/2017 influenza season. Methods We compared rates of primary care influenza‐like illness (ILI) consultations, confirmed influenza intensive care unit (ICU) admissions, and all‐cause excess mortality using standardised methods. To further quantify the impact, a scoring system was developed where each weekly rate/z‐score was scored and summed across each influenza season according to the weekly respective threshold experienced in each country. Results Results highlight ILI consultation rates in the four seasons' post‐programme, breached baseline thresholds once or not at all in Scotland and Northern Ireland; in three out of the four seasons in England and Wales; and in all four seasons in ROI. No differences were observed in the seasons' post‐programme introduction between countries in rates of ICU and excess mortality, although reductions in influenza‐related mortality were seen. The scoring system also reflected similar results overall. Conclusions Findings of this study suggest that LAIV vaccination of primary school age children is associated with population‐level benefits, particularly in reducing infection incidence in primary care.


| INTRODUCTION
Traditional seasonal influenza vaccination campaigns in the United Kingdom (UK) and the Republic of Ireland (ROI) routinely targeted those at higher risk of severe disease; those aged ≥65 year olds, individuals aged between 6 months and 64 years in a clinical risk group; and those pregnant women plus frontline healthcare workers to protect both themselves and their vulnerable patients. 1,2 In 2012, following recommendations by the UK Joint Committee on Vaccination and Immunisation (JCVI), the introduction of a universal paediatric influenza vaccination programme commenced with the ultimate vision to vaccinate all children aged 2 to 16 years of age in the UK through the incremental phased roll-out of a newly licensed live attenuated influenza vaccine (LAIV). 3 The programme was introduced based on the projected direct and indirect beneficial impact of vaccinating school age children on the wider population on reducing influenza transmission. 4 Observational data on such indirect reductions are however limited to individual country assessments in the UK and elsewhere. [5][6][7][8][9] The This differential roll-out of the paediatric programme across these five countries provided a unique opportunity to compare the indirect and overall effects of vaccinating children on the epidemiology of influenza across the UK and ROI across several seasons.
A number of approaches have been developed over the years to standardise the monitoring of influenza surveillance across countries, in particular the World Health Organization Pandemic Influenza Severity Assessment (WHO PISA) initiative, which looks to describe influenza activity and impact and to use this to inform national and global risk assessments more uniformly. 11 The present study aims to use established methods to determine potential differences in the impact of the paediatric influenza vaccination programmes on influenza between countries of the UK and the ROI.
This study assesses the overall and indirect impact of vaccinating primary school age children with influenza vaccine by comparing the epidemiology of influenza in the UK and ROI where varying vaccine strategies were implemented using a range of primary care, secondary care and mortality indicators.

| METHODS
Countries of the UK and the ROI were categorised according to the delivery method of their respective paediatric vaccination programme ( Figure 1).
The study periods were defined as the pre-programme (pre-vac- 1. The activity indicator used in primary care was general practice (GP) influenza-like illness (ILI) consultation rates.
2. The impact indicator used in secondary care was intensive care unit/high-dependency unit (ICU/HDU) admission rates.
3. The mortality indicators used were all-cause and influenzaattributable excess mortality.

| EuroMOMO and FluMOMO models
Mortality indicators were assessed using empirical thresholds based on the analysis of the EuroMOMO algorithm outputs; where the baseline threshold was defined as <2 z-score, the low threshold as 2 to <6 z-score, the moderate thresholds as 6 to <10 z-score, the high threshold as 10 to 16 z-score and the very high threshold as >16 z-score. The same threshold values were applied to all age groups.
The EuroMOMO model aims to provide weekly excess all-cause mortality estimates using a time series Poisson regression model whilst taking into account trends, seasonal variation and corrections for delays. 14 Z-scores from the model outputs were used to determine thresholds. Thresholds were applied to the weekly excess estimates for comparisons across countries.
Influenza-attributable mortality rates were also calculated using the FluMOMO algorithm, a multiplicative Poisson regression time series model with overdispersion. 15,16 The FluMOMO algorithm was run over six seasons (2011/2012 to 2016/2017).

All mortality models (EuroMOMO and FluMOMO version 4.2)
were run in STATA 13.

| Scoring system
To further quantify the impact, a scoring system was set up where each weekly rate/z-score was scored and summed across each

| Mortality indicator
The   No overall reductions in percentage change were observed through the scoring system between pre-and post-programme scores in all countries (

| EuroMOMO
Differences were not observed through the EuroMOMO model in pre-and post-programme seasons between countries.
During the pre-programme period, all countries but Wales brea- Through the scoring system, overall reduction in percentage change between average pre-and post-programme scores was only noted for Northern Ireland and England (10% and 6% decrease, respectively); however, this was not reflected in age-specific scores for these countries, where no reductions in percentage changes were noted (Table 5). Wales experienced no overall reduction in percentage however saw a reduction in percentage change between the average pre-and post-programme scores of 50% in the <15 year of age (Table 5). Similarly, the ROI saw a reduction of 88% between
Shaded grey cells represent age group data, which could not be calculated.
the average pre-and post-programme scores in the 15-64 years of age (Table 5).

| FluMOMO
Differences between countries who were vaccinating all primary school-aged children and those who are incrementally vaccinating pri- Our findings highlight that excess mortality may however also be related to a range of factors besides influenza, including systematic differences in the recording of these data, the contribution of other respiratory infections, winter pressures on health services and cold weather.
The strengths of this study include the use of standardisation methods to allow comparisons between countries. The MEM method has been analysed and adopted in the UK to be a better approach in reporting and assessing the impact on healthcare services during seasonal influenza periods. 17 This method has also been adopted by several European countries and has become part of a wider WHO initiative. 11 The EuroMOMO model is used by a network of approximately 26 European countries analysing and reporting all-cause excess mortality weekly. 14 Additionally, we introduce a new method of quantifying thresholds to compute scores to allow comparisons across different time periods (pre-vs. post-programme scores), which can be adapted at an individual country basis. Another strength is the use of national population-level data, which has allowed us to better analyse the impact of vaccine programmes in all countries. For example, information on all influenza ICU admissions across each respective countries was available for the duration of the study, and population denominators were used. Similarly, national death registration data from each country have been used, and excess mortality rates were calculated based on population denominators.
There are some limitations that need to be considered, as although we have tried to address differences between surveillance schemes in the countries by using standardised methods, underlying differences in factors such as case definitions, access to care and health-seeking behaviours between countries remain. For example, Conceptualization; writingreview and editing. Richard G. Pebody: Conceptualization; supervision; writingreview and editing.