Terence Stephenson DM FRCP FRCPCH is a consultant pediatric doctor in the UK and chair of the General Medical Council (GMC). He was formerly Dean of the Medical School and Professor of Child Health at the University of Nottingham from 2003-2009. And the Nuffield Professor of Child Health. He was President of the Royal College of Paediatrics and Child Health from April 2009 until May 2012. He then took up the role of chair of the Academy of Medical Royal Colleges in July 2012. He became a member of the GMC council in 2009. In September 2014 it was announced that he would become the chair of the GMC, succeeding Peter Rubin on 1 January 2015. In October 2014 it was announced that he had been appointed as a panel member for the Independent Panel Inquiry into Child Sexual Abuse. He has co-authored textbooks, written invited chapters and editorials, and published more than 150 peer-reviewed papers in academic journals. He has been described as leading by example.
Introduction: Influenza is a common cause of outpatient medical visits and hospitalizations among young children. According to data from the National Centre for Disease Control and Prevention, influenza morbidity and mortality during the period 2015-16 (predomimant virus type A / H1N1) was more severe than that of the previous period (predomimant virus type B). However, influenza type A caused the majority of deaths during both periods.
Purpose: The purpose of this study is to compare the influenza morbidity between the periods 2014-15 and 2015-16 among the paediatric population based on data from the Penteli General Children’s Hospital in Athens, Greece.
Materials and Methods: Retrospective cohort study during two viral seasons (2014-15, 2015-16). Inclusion criteria: Children <18 years of age, who were hospitalized with laboratory confirmed influenza infection (Rapid Infuenza Antigen Detecting Test, Polymerase Chain Reaction). The immunization status for influenza and the underlying health conditions were documented.
Results: A total of 58 children were hospitalized due to influenza over 2 viral seasons. 2014-15: Total of 16 admissions, 1 (6.25%) PICU admission (previously healthy child with encephalitis due to H3N2 virus). Influenza type: A: 87.5%, B: 12.5%. None of the children were vaccinated. Children in PICU with underlying conditions: 0%. Period 2015-16: Total of 42 admissions, 3 (7.14%) PICU admissions (two adolescents with Diabetes Melitus type I and Dravet syndrome respectively and a previously healthy child with influenza and Respiratory Syncytial Virus co-infection. All presented with respiratory failure due to H1N1 infection that required invasive ventilation). Influenza type: A: 78.57% (H1N1), B: 21.42%. None of the children were vaccinated. Children in PICU with underlying conditions: 66%. Children in PICU with underlying conditions during both virus seasons: 50%.
Conclusion: Influenza morbidity was higher during the period 2015-16 among children, which is in agreement with national data. However, in spite of influenza type B prevalence on the general population during the period 2014-15, higher morbidity due to type A was documented among children in our hospital. Interestingly, the percentage of children with underlying conditions that exhibited severe disease that led to PICU admission was equal to those with no co-morbidities. The fact that influenza morbidity each season cannot be predicted as well as the fact that all children admitted in PICU were unimmunized, highlight the importance of vaccination against influenza among the whole of the paediatric population.
1Paediatric intensive Care Unit, Penteli Children’s Hospital, Athens, Greece; 2Paediatric Department, Penteli Children’s Hospital, Athens, Greece; email@example.com
At the end of the Session participants will: (1) understand the nature of live attenuated influenza vaccine (LAIV); (2) learn about the universal LAIV programme for children in the UK; and (3) learn what is and is not known about the potential of LAIV to interrupt transmission at the population level.
EiP's Online Excellence in Pediatrics Library allow you to update your skills and knowledge anywhere and at any time. Simply click on the videos that interests and then you will have access to free-to-view online sessions, led by one of EiP's expert speakers, covering the latest changes in pediatrics. The Online Excellence in Pediatrics Library is self-paced, offering hands-on practical advice that can be viewed as many times as you wish and all from the comfort of your own office or home.
EiP's Online Excellence in Pediatrics Library allow you to update your skills and knowledge anywhere and at any time.
Seasonal influenza represents an important cause of morbidity and mortality especially for the risk of secondary bacterial infections, which is higher in children and elderly than in the general population. The burden of influenza is highest in young children under 5 years of age likely due to immunological immaturity,  and .
Increasing attack rates during epidemics lead to higher outpatient visit and hospitalisation rates , and . Influenza-associated hospitalisation rates are well described in children with underlying chronic conditions; however accumulating evidence showed that the increased risk also affected otherwise healthy children . Observational data indicated that although children with underlying conditions are at higher risk of death, the majority of paediatric deaths occur among healthy children .
The vaccination against influenza is recognised as an effective preventive intervention and each country is responsible for national programs and for defining targeted risk groups. In the majority of European countries, the influenza vaccine is recommended for children with underlying medical conditions. UK authorities announced plans to extend influenza vaccination to all children aged 2–16 years from 2014 . At present, Finland is the only European country which has implemented the routine influenza vaccination of healthy children (6 months to <3 years) .
In Italy, the course of influenza epidemics generally extends between December and April, with a peak in February  and each year the Ministry of Health promotes a vaccination campaign between mid-October and December. The official recommendation identifies at risk children as a target group for influenza vaccination (provided free of charge); only sub-unit, split or virosomal seasonal vaccine formulations can be administered in children (6 months to 17 years of age)  and . During the seasons 2011–2012 and 2012–2013, the composition of the vaccines varied only for the B virus strain (B/Wisconsin in 2011–2012, and B/Brisbane in 2012–2013), whereas the A(H1N1) and A(H3N2) antigens were present in both seasons. The two vaccine strains B/Wisconsin and B/Brisbane belong to two different lineages, i.e. B-Yamagata and B-Victoria respectively.
Most of the available evidence on the efficacy and effectiveness of seasonal influenza vaccine in a paediatric setting is derived from clinical trials and concerns almost entirely healthy children , , and . Although these studies adopted heterogeneous outcome definitions (e.g. from clinically defined influenza like-illness (ILI) in the outpatient setting to laboratory confirmed hospitalisations for influenza), they found efficacy estimates of around 70%, higher than those on effectiveness (around 40%). Despite the fact that influenza vaccination is primarily recommended in children with underlying conditions, insufficient evidence is available in this population. Moreover, the World Health Organization considers as a target group for influenza immunisation, children from 6 to 23 months, even though effectiveness data are scanty .
The objective of this national study was to determine the effectiveness of seasonal influenza vaccination against laboratory-confirmed influenza cases visiting the Emergency Department (hospitalised or not) in a large paediatric population over two consecutive seasons (2011–2012 and 2012–2013) and to provide evidence for vaccination recommendations in Italy.
In Italy, since 1999 an active surveillance on drug and vaccine safety in children has been conducted in various paediatric hospitals/wards located throughout the country . Italian paediatric hospitals/wards can admit children from 0 to 17 years of age. Overall, the network includes 11 sites from seven regions representative of the whole Country, and around 400,000 children visited the EDs of the participating centres each year. The network organisation facilitated the prompt set up of the investigation on influenza vaccine effectiveness during the A/H1N1 pandemic (in 2009) and in two following influenza seasons (2011–2012 and 2012–2013). The results of the A/H1N1 pandemic vaccination campaign were reported elsewhere .
Written informed consent was acquired from parents. Data were collected by trained pharmacists/physicians by interviewing parents during the ED visit (or hospital admission) of their children. Demographic data, medical history of chronic conditions, date of vaccination and type of vaccine were collected using a structured questionnaire. For the assessment of influenza vaccine effectiveness, children were defined as vaccinated if they had received at least one dose more than 14 days before symptom onset.
An influenza-confirmatory laboratory test was carried out in all children. The virus was detected through nasopharyngeal sample collection; stable viral transport medium was added to swabs. Specimens were collected and analysed by using a real-time reverse transcriptase-polymerase chain reaction (RT-PCR). In six centres the tests were analysed in internal laboratories, whereas the others sent the specimens to certified external laboratories.
The first phase of the study was performed in the 2011–2012 influenza season and was used as a pilot study to refine the 2012–2013 investigation. In order to concentrate enrolment and laboratory tests in the epidemic period the coordinator centre gave the start-up on the basis of data on influenza epidemics in Italy provided from the National surveillance of ILI incidence . The inclusion of children took place between 1 February and 31 March 2012 (for the 2011–2012 season), and between 14 January and 15 March 2013 (for the 2012–2013 season). The inclusion periods were the same for all centres.
Data were analysed according to a test-negative case-control study design: all children with a positive confirmatory laboratory test (to one of the viruses contained in the seasonal vaccine) were included as cases, whereas controls were children with a negative test. For effectiveness evaluation, odds of influenza vaccination were compared in cases and controls.
2.1. Study sites
The following paediatric hospitals and departments were participating: Giannina Gaslini Paediatric Hospital (Genova); Regina Margherita Paediatric Hospital (Torino); Department of Paediatrics, University of Padova; Paediatric Department, Treviso Hospital (Treviso); Anna Meyer Children's University Hospital (Firenze); Department of Paediatrics, University of Perugia; Pharmacology and Paediatrics and Developmental Neuroscience, Università Cattolica S. Cuore (Roma); Bambino Gesù Paediatric Hospital (Roma); Santobono-Pausilipon Paediatric Hospital-Virologic Unit Cotugno (Napoli); Giovanni Di Cristina Paediatric Hospital (Palermo); University Hospital of Messina. A common study protocol was approved by the Ethics Committee of each clinical centre. The study was coordinated by the National Centre of Epidemiology of the National Institute of Health in Rome.
2.2. Statistical analyses
Data were analysed with SPSS (v. 21.0). T-test was used to compare means, Wilcoxon–Mann–Whitney non-parametric test was used to compare medians and Chi-square test was used to compare percentages. Adjusted odds ratios (ORs) and 95% confidence intervals (CI) were estimated through a logistic regression model. ORs were adjusted for age, which was included in the logistic model as a continuous variable (in months). We estimated the seasonal influenza vaccine effectiveness (VE) as 1 minus the OR, expressed as a percentage.
Among the 773 eligible children, 69 (9%) were excluded (Fig. 1). The main reason for exclusion was lack of informed consent either to collect the nasopharyngeal swab (n = 25) or to be included in the study (n = 10). The 704 remaining children were classified as cases (262 children tested positive for one of the influenza viruses) and controls (442 children who tested negative). The percentage of hospitalised children was 56% (n = 148) among cases and 75% (n = 332) among controls. Overall, the age of the enrolled children ranged from 6 months to 16 years.
The proportion of cases ranged from 12% to 56% in the 11 centres. In 69% of cases and 55% of controls the test was performed the same day of symptom onset. In 97% of cases and in 93% of controls the test was carried out within 2 days. Among cases, B virus was detected in 126 children (48%), A(H1N1) in 59 (23%), unspecified A virus in 33 (13%), A(H1N1)pdm09 in 22 (8%) and A(H3N2) in 22 (8%). In the 2012–2013 season the virology unit of one clinical centre was able to characterise 40 of the 126 cases positive for influenza B virus: they all resulted belonging to B/Yamagata/16/88 lineage.
Cases and controls were similar with regard to gender and prevalence of chronic diseases, whereas a statistically significant difference was observed for age (46 months in cases and 29 months in controls) (Table 1).
Cases Controls p Number 262 442 Median age, months (IR) 46 (26–71) 29 (15–54) <0.001 % females 45 47 0.78 Chronic diseases: N. (%) 47 (18) 67 (15) 0.34 Duration of symptoms before admission to ED, median days 3 2 0.01 Symptoms of ILI at admission to ED Fever, median °C 39 39 0.13 Cough, N (%) 224 (85) 365 (83) 0.31 Rhinorrhea, N (%) 122 (47) 217 (49) 0.52 Malaise, N (%) 111 (42) 166 (38) 0.21 Sore throat, N (%) 85 (32) 135 (31) 0.60 Asthenia, N (%) 62 (24) 88 (20) 0.24 Vomiting, N (%) 58 (22) 130 (29) 0.04 Diarrhoea, N (%) 27 (10) 76 (17) 0.01 Bronchitis, N (%) 22 (8) 73 (17) 0.002 Hospitalisation, N. (%) 148 (56) 332 (75) <0.001 Length of staya (mean, days) 3.6 4.3 0.20 Type of virus, N (%) B 126 (48) – A (H1N1) 58 (23) – A (unspecified) 33 (13) – A (H1N1) pdm09 22 (8) – A (H3N2) 22 (8) –
IR: interquartile range, ED: Emergency Department.
a Among hospitalised children.
The median duration of symptoms before the visit to the ED was similar in the two groups (3 days vs. 2), as it was the level of fever (median of 39 °C in both groups). According to the ILI definition all children presented fever ≥38 °C. Cough was the most frequently associated symptom in both cases and controls (85% vs. 83%), followed by rhinorrhea, malaise, sore throat and asthenia. Vomiting or diarrhoea were more frequently reported in younger children (40% in patients up to 5 years and 21% in older ones). Sixty-eight percent of children were hospitalised through the EDs and the mean duration of hospitalisation was not statistically different in cases and controls (3.6 and 4.3 days respectively).
Only 25 children (4%) were vaccinated against influenza: seven of the 262 cases and 18 of the 442 controls (they had been vaccinated between October and mid-January). The date of vaccination was not available for six children (one case and five controls). However, it is likely that these children were vaccinated at least 14 days before hospital admission, since they were hospitalised between the end of January and February. Twelve out of the 25 vaccinated children (46%) reported a chronic disease (asthma, allergy, cardiomyopathy, spinal muscular atrophy [SMA 1 or 2], immunodeficiency, aplastic anaemia, coeliac disease, West syndrome).
The overall age-adjusted VE was 38% (95% CI: −52% to 75%) (Table 2). A slightly lower VE was estimated in the 2012–2013 season (VE 26%; 95% CI: −153% to 78%). Three out of seven vaccinated children were positive to unspecified A virus (one child) or A/H3N2 virus (two children) in the 2011–2012 season, whereas the remaining four vaccinated cases in the 2012–2013 season were positive to B virus. Nine children (one case and eight controls) received two doses of the vaccine in the same season (VE 79%; 95% CI: −57% to 100%).
Influenza vaccine CasesN(%) ControlsN(%) Total Crude OR (95% CI) Adj VEa(95% CI) Yes 7 (3) 18 (4) 25 35% (−65% to 77%) 38% (−52% to 75%) No 255 (97) 424 (96) 679 Total 262 (100) 442 (100) 704 Season 2011–2012
Influenza vaccine CasesN(%) ControlsN(%) Total Crude OR (95% CI) Adj VEa(95% CI) Yes 3 (5) 10 (8) 13 38% (−152% to 89%) 41% (−126% to 84%) No 58 (95) 119 (92) 177 Total 61 (100) 129 (100) 190 Season 2012–2013
Influenza vaccine CasesN(%) ControlsN(%) Total Crude OR (95% CI) Adj VEa(95% CI) Yes 4 (2) 8 (3) 12 23% (−194% to 83%) 26% (−153 to 78%) No 197 (98) 305 (97) 502 Total 201 (100) 313 (100) 514
- a Vaccine effectiveness adjusted by age.
When the analysis was restricted to hospitalised children a higher estimate of VE, with respect to the overall, was obtained (53%; 95% CI −45% to 85%).
Our study estimated around 40% reduction in visits to EDs and hospitalisations for ILI in children, although not statistically significant and with wide confidence intervals.
Even though the confidence intervals of the estimates were largely overlapping, a slightly lower effectiveness was estimated in the second year. The four vaccinated cases in the 2012–2013 season were positive to the B virus. Data from our study and virological surveys performed in Italy  showed that the B/Yamagata lineage was circulating in the latter season (whereas B/Brisbane strain, belonging to a different lineage, was included in the seasonal vaccine), which may explain the lower VE of the 2012–2013 vaccine with respect to the 2011–2012, when the A(H3N2) and A(H1N1) were mostly present. The matching between the vaccine and circulating strains of influenza season is a recognised factor influencing the VE .
The main limitation of the study derives from the low vaccination coverage observed in the Italian paediatric population (4% in the control group). This proportion was similar to that observed in Italy during the 2009 pandemic . Due to the few vaccinated children it was not possible to perform stratified analyses by variables of interest, such as type of virus/vaccine, age groups, presence of chronic conditions and prior vaccination status. Assuming as true the estimate of efficacy in our study, to reach statistical significance we should have had (with alpha error of 5% and power 80%), either a 25% proportion of vaccinated children or a study population of ILI larger than 4000. However, the number of children enrolled in our study is large in comparison with other recently published articles. In the I-MOVE study, the paediatric population (1–14 years) amounted to 512 children who were included in five European countries .
The adopted study design allows to control for the confounding effect of baseline clinical status. The reason relies on the definition of the control group, consisting of children who tested negative for the influenza virus vaccine . It is well documented that several conditions increase the likelihood of developing an ILI and represent, at the same time, an indication for vaccination. In our study, case and control subjects were similar with reference to the prevalence of chronic conditions, but not for symptoms at onset. For instance, vomiting and diarrhoea were more frequent in controls. These symptoms are more often associated with ILI presentation in younger children. The age difference is in line with that observed in other European countries. In the I-MOVE study, the difference in the mean age between cases and controls in the paediatric population (1–14 years) was 1.5 years, similar to the difference observed in our study .
Almost all nasopharyngeal swabs were carried out within 2 days from symptoms onset to the ED, which is associated with a greater specificity. The fact that results were obtained several days after having conducted the test, excludes the possibility that the exposure information may have been biased by the knowledge of case/control status (and consequently no recall or ascertainment bias may have played a role).
In Italy, influenza vaccination remains an unmet priority, as only 4% of children were vaccinated in the recent seasons . Efforts should focus on paediatricians to discuss the importance of influenza vaccination for preventing major complications in both at-risk and healthy children. Systematic reviews and meta-analysis of existing studies may provide the basis for a new awareness on the positive benefit-risk profile of the influenza vaccination even among healthy children.
Our study provides additional data on the effectiveness of the seasonal influenza vaccination in preventing visits to the Emergency Departments and hospitalisations for ILI, and adds further evidence for vaccination recommendations especially in children.
The study was partially funded by the Italian Medicines Agency (AIFA).
Copyright © 2014 The Authors. Published by Elsevier Ltd.
Vaccine. 2014 Jul 31;32(35):4466-70. doi: 10.1016/j.vaccine.2014.06.048. Epub 2014 Jun 21.
To evaluate the effectiveness of seasonal influenza vaccine in preventing Emergency Department (ED) visits and hospitalisations for influenza like illness (ILI) in children.
We conducted a test negative case-control study during the 2011-2012 and 2012-2013 influenza seasons. Eleven paediatric hospital/wards in seven Italian regions participated in the study. Consecutive children visiting the ED with an ILI, as diagnosed by the doctor according to the European Centre for Disease Control case definition, were eligible for the study. Data were collected from trained pharmacists/physicians by interviewing parents during the ED visit (or hospital admission) of their children. An influenza microbiological test (RT-PCR) was carried out in all children.
Seven-hundred and four children, from 6 months to 16 years of age, were enrolled: 262 children tested positive for one of the influenza viruses (cases) and 442 tested negative (controls). Cases were older than controls (median age 46 vs. 29 months), though with a similar prevalence of chronic conditions. Only 25 children (4%) were vaccinated in the study period. The overall age-adjusted vaccine effectiveness (VE) was 38% (95% confidence interval -52% to 75%). A higher VE was estimated for hospitalised children (53%; 95% confidence interval -45% to 85%).
This study supports the effectiveness of the seasonal influenza vaccine in preventing visits to the EDs and hospitalisations for ILI in children, although the estimates were not statistically significant and with wide confidence intervals. Future systematic reviews of available data will provide more robust evidence for recommending influenza vaccination in children.
We present the 2014/15 mid-season estimates of influenza vaccine effectiveness (VE) for the United Kingdom of England, Wales, Scotland and Northern Ireland (UK). This season is dominated by early circulation of influenza A(H3N2) virus, and an overall VE in preventing medically attended laboratory-confirmed influenza in primary care of only 3.4% and against A(H3N2) of −2.3%. This report provides clear evidence of antigenic and genetic mismatch between circulating A(H3N2) viruses and the respective 2014/15 northern hemisphere vaccine strain.
The UK has a long-standing selective influenza immunisation programme targeted at individuals at higher risk of severe disease, in particular all those 65 years and above and under 65-year olds in a clinical risk group, using inactivated trivalent influenza vaccine. The 2014/15 season is the second year where intranasally administered live attenuated influenza vaccine (LAIV) has been offered to pre-school age children in the UK with certain areas also vaccinating children of school-age . This winter has been characterised by early influenza activity, with A(H3N2) virus the dominant circulating sub-type. In England, by week 4 2015 peak influenza activity levels had exceeded those seen in the past three seasons, but not approached the peak levels seen in 2010/11 and 2008/09 . The current season has led to large numbers of care home outbreaks, often in highly vaccinated populations, hospitalisations and significant excess all-cause mortality in the over 65 year-old population.
The UK has well established methods to produce mid- and end-of-season estimates of VE in preventing primary care consultation due to laboratory-confirmed influenza infection [3,4]. The key aims of the present study were to provide early estimates of influenza VE in the UK to inform influenza prevention and control measures both for the remainder of this season and the forthcoming World Health Organization (WHO) convened meeting at the end of February, where vaccine composition is decided for the forthcoming northern hemisphere 2015/16 season.
Study population and period
Data were derived from five primary care influenza sentinel swabbing surveillance schemes from England (two schemes), Scotland, Wales and Northern Ireland. Details of the Royal College of General Practitioners (RCGP), Public Health England (PHE) Specialist Microbiology Network (SMN), Public Health Wales, Public Health Agency (PHA) of Northern Ireland and Health Protection Scotland (HPS) swabbing schemes have been published previously .
The study period ran from 1 October 2014 to 16 January 2015. Patients were swabbed as part of clinical care, with verbal consent. Cases were defined as persons presenting during the study period in a participating general practitioner (GP) practice with an acute influenza-like-illness (ILI) who were swabbed and then tested positive for influenza A or B viruses. An ILI case was defined as an individual presenting in primary care with an acute respiratory illness with physician-diagnosed fever or complaint of feverishness. Controls were individuals presenting with ILI in the same period that were swabbed and tested negative for influenza.
A standardised questionnaire was completed by the GP responsible for the patient during the consultation. Demographic, clinical and epidemiological information was collected from cases and controls, including date of birth, sex, pre-defined underlying clinical risk group, date of onset of respiratory illness, date of specimen collection, and influenza vaccination status for the 2014/15 season, with vaccination dates and route of administration (injection/intranasal) and whether resident in an area where a primary school vaccination programme was in operation.
Laboratory confirmation was undertaken using comparable methods with real-time polymerase chain reaction (RT-PCR) assays capable of detecting circulating influenza A and influenza B viruses and other respiratory viruses [5,6]. Samples were sent to respective laboratories as previously described . Further strain characterisation was also performed; influenza viruses were isolated in MDCK or MDCK-SIAT1 cells from RT-PCR positive samples from England as previously described [7,8]. Influenza A(H3N2) virus isolates with a haemagglutination titre ≥ 40 were characterised antigenically using post-infection ferret antisera in haemagglutination inhibition (HI) assays, with guinea pig red blood cells . Nucleotide sequencing of the haemagglutinin (HA) gene of a subset of influenza A(H3N2) viruses selected to be representative of the range of patient’s age, date of sample collection, geographical location, and antigenic characterisation of the influenza A(H3N2) virus isolate, if performed, was undertaken (primer sequences available on request), and phylogenetic trees were constructed with a neighbour-joining algorithm available in the Mega 6 software (http://www.megasoftware.net) . HA sequences from reference strains used in the phylogenetic analysis were obtained from the EpiFlu database of the Global Initiative on Sharing Avian Influenza Data (GISAID) (Table 1).
Persons were defined as vaccinated if the date of vaccination with the 2014/15 seasonal influenza vaccine was 14 or more days before onset of illness. Those in whom the period between vaccination and onset of illness was less than 14 days were excluded, as were those where date of vaccination was missing. Those with a missing date of onset or an onset date more than seven days before the swab was taken were also excluded.
VE was estimated by the test negative case control (TNCC) design. In this design, VE is calculated as 1-(odds ratio) obtained using multivariable logistic regression models with influenza PCR results and seasonal vaccination status as the linear predictor. VE was also estimated for influenza A only and for A(H3N2); Influenza A(H1N1) and B numbers were too small to examine. In the analyses evaluating VE for a specific type or strain, those positive for other types were excluded. Age (coded into four standard age groups, < 18, 18–44, 45–64 and ≥ 65 years), sex, clinical risk group, surveillance scheme (RCGP, SMN, HPS, Wales, Northern Ireland), residence in an area where primary school age vaccination programme operated and date of onset (month) were investigated as potential confounding variables. All statistical analyses were carried out in Stata version 13 (StataCorp, College Station, Texas).
A total of 2,278 individuals were swabbed in primary care during the study period and had a laboratory result available. The reasons for study inclusion and exclusion are outlined in Figure 1. Five persons were excluded because the influenza virus detected in them was a LAIV vaccine strain based either on sequence analysis or inferred based on influenza co-infection.
The details of the 1,341 individuals remaining in the study were stratified according to the swab result (Table 1). Positivity rates differed significantly by month, scheme and primary school age vaccination programme area.
Influenza A(H3N2) strain characterisation
During the study period, a total of 127 A(H3N2) circulating viruses were isolated from all referred clinical samples and antigenically characterised by HI analysis. The majority of A(H3N2) viruses analysed (100/127; 79%) were antigenically similar to the A(H3N2) virus component of the 2014/15 northern hemisphere vaccine A/Texas/50/2012 (≤ 4-fold difference) with antiserum raised against A/Texas/50/2012 in antigenic characterisation by HI); however a proportion of A(H3N2) viruses (21%) showed reduced reactivity (> 4-fold difference) (Table 2).
These viruses were antigenically similar to A/Switzerland/9715293/2013, the recommended A(H3N2) component of the 2015 southern hemisphere vaccine. A > 4-fold difference in HI assay titres with reference antiserum is considered to be significant antigenic drift. Of the 44 UK influenza A(H3N2) viruses analysed to date by sequencing, the HA genes of these viruses belonged in the genetic clade 3C, as does the 2014/15 A(H3N2) vaccine strain A/Texas/50/2012 and A/Switzerland/9715293/2013, one of the three recommended strains for the southern hemisphere 2015 vaccine composition. However, the majority (35/44; 79.5%) of the HA sequences of 2014/15 UK viruses analysed were further characterised within this clade to belong in subgroup 3C.2a of group 3C.2, with fewer (9/44 17.3%) in group 3C.3 (Figure 2). These groups are considered genetically distinct from the 2014/15 A/Texas/50/2012(H3N2)-like clade 3C.1 vaccine reference strain.
Model fitting for vaccine effectiveness estimation
When estimating vaccine effectiveness, age group, sex, month of onset, surveillance scheme and primary school age programme area were adjusted for in a multivariable logistic regression model. Only surveillance scheme, time period and primary school age programme area were significantly associated with having a positive swab, and all were confounders for vaccine effectiveness (changing the estimate by more than 5%). Information on risk group was missing for 131 samples (9.8%) and was therefore not included in the final model. If risk group was included, it was found not to be associated with being positive and the VE estimates remained similar.
Table 3 shows vaccine effectiveness estimates against all influenza, influenza A and influenza A(H3N2). There were inadequate samples to enable estimation of effectiveness against influenza A(H1N1)pdm09 or influenza B. The adjusted VE of influenza vaccine against all influenza was 3.4% and was very similar for A(H3N2), reflecting the fact that A(H3N2) is the dominant circulating virus strain this season.
This paper presents the mid-season estimates of influenza vaccine effectiveness for the UK. In a season dominated by early circulation of influenza A(H3N2) virus, we found the overall VE in preventing medically attended laboratory-confirmed influenza in primary care was only 3.4% and −2.3% specifically against A(H3N2). We also found clear evidence of antigenic and genetic mismatch between circulating A(H3N2) viruses and the 2014/15 northern hemisphere vaccine strain.
The UK, together with other European countries, the United States, Canada and Australia has well established systems to generate interim estimates of seasonal influenza VE. These early results are used to optimise in-season control and prevention measures, to inform other countries where the influenza season may have just started or is about to start, and to contribute to forthcoming WHO vaccine composition deliberations.
The UK, as observed in North America and some other European countries [11,12] has experienced a season dominated by early circulation of influenza A(H3N2) virus – with clear evidence of emergence of a drifted A(H3N2) strain, first seen in North America in spring 2014 , and then in Australia in mid-2014 . This drifted strain has reduced antigenic reactivity with antiserum raised to the current A(H3N2) vaccine strain (A/Texas/50/2012), and is antigenically more closely related to A/Switzerland/9715293/2013, the A(H3N2) virus selected as one of the three recommended components for the 2015 southern hemisphere influenza vaccine .
Characterisation of circulating influenza viruses involves both genetic and antigenic characterisation. Genetic analysis focusses on detailed comparison of the HA genes of viruses, tracking changes over time and linking this to phylogenetic analysis of sequence clustering to determine emerging virus groups and changes in receptor binding and other important antigenic epitopes. Genetic variation does not always correlate with antigenic variation. The interpretation of both data sources is complex, as not all viruses can be cultivated in sufficient quantity for antigenic characterisation, so that viruses for which sequence information is available may not be antigenically characterised, leading to potential bias in interpretation. This is particularly relevant to the discussion of antigenic characterisation data for A(H3N2) viruses in the 2014/15 winter season. Some circulating A(H3N2) viruses are difficult to grow in tissue culture as a result of genetic drift affecting receptor binding properties , and viruses grown in eggs may have egg adaptive changes which make the analysis of antigenic drift complex. Most A(H3N2) strains seen since February 2014 fell into the HA genetic clade (3C) for which A/Switzerland/9715293/2013 was a prototype representative strain. Antigenic and genetic characterisation data indicate that A/Switzerland/9715293/2013-like strains have circulated in the UK during winter 2014/15. There is a clear antigenic mismatch between the northern hemisphere H3N2 vaccine strain and the circulating variant in winter of 2014/15. The full picture of virological variation requires further detailed analysis, not possible at this stage of the 2014/15 season.
Our observation of an absence of significant effectiveness in preventing medically-attended laboratory-confirmed influenza in primary care due to A(H3N2) are congruent with the findings recently reported from the US  who report low effectiveness of 22% (95% confidence interval (CI): 5–35) and from Canada who report a VE of −8% (95% CI: −50 to 23) against laboratory-confirmed, medically-attended influenza A(H3N2) virus infection in primary care . The observation of low or non-significant effectiveness in 2014/15 in the UK and in North America correlates with the direct mismatch seen between the vaccine virus and A(H3N2) strains circulating this winter. Vaccine mismatch due to circulation of drifted strains does occasionally occur, although this is the lowest estimate of influenza VE reported by the UK over the past decade using the TNCC approach to measure VE [3,4]. It is also important to highlight the uncertainty of our estimate. The upper 95% CI of 35% shows we can be confident that VE is low at this point although we cannot be clear that influenza vaccine has no effectiveness this season. Indeed the significantly lower influenza positivity in areas where children of school age were vaccinated compared to non-pilot areas (Table 1) is suggestive of a possible impact of the childhood influenza vaccination programme. Furthermore, this mid-season analysis does not preclude the likelihood that the vaccine should offer protection from different types of influenza, such as influenza B that may still circulate later in the season. All these elements will form part of the end-of-season analysis including stratification by age-group and scheme.
The WHO has made their recommendations for the composition of the influenza vaccine for the 2015 southern hemisphere winter in September 2014, including a switch to a A/Switzerland/9715293/2013 (H3N2)-like virus . The WHO influenza vaccine composition group will convene shortly, at the end of February 2015, to consider recommendations for the forthcoming northern hemisphere 2015/16 winter, and the findings in this paper will contribute to their deliberations. The observation of low vaccine effectiveness this season highlights the vital importance of implementing other prevention and control measures for the remainder of this season, in particular the early use of influenza antivirals for post-exposure prophylaxis and treatment of vulnerable populations, such as the elderly, together with appropriate infection control measures.
We are grateful to the many primary care physicians in England, Wales, Northern Ireland and Scotland who supplied the clinical information on their patients; to the staff of the PHE Respiratory Virus Unit, the PHE Specialist Microbiology Laboratories, Public Health Wales Specialist Virology Centre, the West of Scotland Specialist Virology Centre and the Regional Virus Laboratory, Belfast who undertook analysis of specimens. We thank the staff of PHE, RCGP, Public Health Wales, Public Health Agency Northern Ireland and Health Protection Scotland teams who coordinate the GP schemes, in particular Abigail Sunderland and Praveen SebastianPillai from PHE; Richard Lewis and Hannah Evans from PHW; Catherine Frew, Alasdair MacLean, Samantha Shepherd & Celia Aitken from WoSSVC and Arlene Reynolds, Diogo Marques, Louise Primrose-Shaw and Karen Voy from HPS for overseeing data collection, and Ivelina Yonova (Practice Liaison), Sameera Pathirannehelage (SQL Developer), and David Mullett (Information Systems Manager) from RCGP/University of Surrey. We acknowledge the originating and submitting laboratories of the sequences from GISAID’s EpiFlu Database on which some of the analyses are based (see Figure 2). All submitters of data may be contacted directly via the GISAID website www.gisaid.org
Conflicts of interest
RGP wrote the first draft; FW and NA led on the statistical analysis; all co-authors contributed epidemiological and/or virological data, contributed to the interpretation of the results, reviewed the early draft and approved the final version.
- Department of Health. Public Health England. The national flu immunisation programme 2014/15. London: Public Health England. 28 Apr 2014. Available from:https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/316007/FluImmunisationLetter2014_accessible.pdf
- Public Health England (PHE). PHE Weekly National Influenza Report – week 4 report. London: PHE. 22 Jan 2015. Available from:https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/397454/Weekly_report_current_week_4x.pdf
- Fleming DM, Andrews NJ, Ellis JS, Bermingham A, Sebastianpillai P, Elliot AJ, et al. Estimating influenza vaccine effectiveness using routinely collected laboratory data. J Epidemiol Community Health. 2010;64(12):1062-7.http://dx.doi.org/10.1136/jech.2009.093450 PMID:19910645
- Andrews N, McMenamin J, Durnall H, Ellis J, Lackenby A, Robertson C, et al. Effectiveness of trivalent seasonal influenza vaccine in preventing laboratory-confirmed influenza in primary care in the United Kingdom: 2012/13 end of season results. Euro Surveill. 2014;19(27): 20851. http://dx.doi.org/10.2807/1560-7917.ES2014.19.27.20851PMID:25033051
- Ellis J, Iturriza M, Allen R, Bermingham A, Brown K, Gray J, et al. Evaluation of four real-time PCR assays for detection of influenza A(H1N1)v viruses. Euro Surveill. 2009;14(22):19230. PMID:19497254
- Gunson R, Maclean A, Davies E, Bennett S, Miller R, Carman WF. Development of a multiplex real-time RT-PCR that allows universal detection of influenza A viruses and simultaneous typing of influenza A/H1N1/2009 virus. J Virol Methods. 2010;163(2):258-61. http://dx.doi.org/10.1016/j.jviromet.2009.10.006 PMID:19854220
- Galiano M, Agapow P-M, Thompson C, Platt S, Underwood A, Ellis J, et al. Evolutionary pathways of the pandemic influenza A (H1N1) 2009 in the UK. PLoS ONE. 2011;6(8):e23779. http://dx.doi.org/10.1371/journal.pone.0023779PMID:21887318
- Matrosovich M, Matrosovich T, Carr J, Roberts NA, Klenk HD. Overexpression of the alpha-2,6-sialyltransferase in MDCK cells increases influenza virus sensitivity to neuraminidase inhibitors. J Virol. 2003;77(15):8418-25.http://dx.doi.org/10.1128/JVI.77.15.8418-8425.2003 PMID:12857911
- Zambon M. Laboratory Diagnosis of Influenza. In: Nicholson K, Hay A, Webster RG, editors. Textbook of Influenza. Oxford: Blackwell Science; 1998. pp. 291-313].
- Tamura K, Stecher G, Peterson D, Filipski A, Kumar S. MEGA6: Molecular Evolutionary Genetics Analysis version 6.0. Mol Biol Evol. 2013;30(12):2725-9. http://dx.doi.org/10.1093/molbev/mst197 PMID:24132122
- Centers for Disease Prevention and Control (CDC). CDC Health Advisory Regarding the Potential for Circulation of Drifted Influenza A (H3N2) Viruses. Atlanta: CDC. 3 Dec 2014. Available from:http://emergency.cdc.gov/HAN/han00374.asp
- Broberg E, Snacken R, Adlhoch C, Beauté J, Galinska M, Pereyaslov D, et al. Start of the 2014/15 influenza season in Europe: drifted influenza A(H3N2) viruses circulate as dominant subtype. Euro Surveill. 2015;20(4):pii=21023. Available from: http://www.eurosurveillance.org/ViewArticle.aspx?ArticleId=21023
- World Health Organization (WHO). Review of the 2014 influenza season in the southern hemisphere. Wkly Epidemiol Rec. 2014;89(48):529-41.
- ECDC Circulation of drifted influenza A(H3N2) viruses in the EU/EEAhttp://ecdc.europa.eu/en/publications/Publications/RRA-InfluenzaA-H3N2-Dec-2014.pdf
- Wilson IA, Cox NJ. Structural basis of immune recognition of influenza virus hemagglutinin. Annu Rev Immunol. 1990;8(1):737-87. http://dx.doi.org/10.1146/annurev.iy.08.040190.003513 PMID:2188678
- Flannery B, Clippard J, Zimmerman RK, Nowalk MP, Jackson ML, Jackson LA, et al. Early estimates of seasonal influenza vaccine effectiveness - United States, january 2015. MMWR Morb Mortal Wkly Rep. 2015;64(1):10-5. PMID:25590680
- Skowronski DM, Chambers C, Sabaiduc S, De Serres G, Dickinson JA, Winter AL, et al. Interim estimates of 2014/15 vaccine effectiveness against influenza A(H3N2) from Canada’s Sentinel Physician Surveillance Network, January 2015. Euro Surveill. 2015;20(4):21022.
- World Health Organization (WHO). Recommended composition of influenza virus vaccines for use in the 2015 southern hemisphere influenza season. Geneva: WHO; 22-24 Sept 2014. Available from:http://www.who.int/influenza/vaccines/virus/recommendations/consultation201409/en/
Eurosurveillance, Volume 20, Issue 5, 05 February 2015
In 2014/15 the United Kingdom experienced circulation of influenza A(H3N2) with impact in the elderly. Mid-season vaccine effectiveness (VE) shows an adjusted VE of 3.4% (95% CI: −44.8 to 35.5) against primary care consultation with laboratory-confirmed influenza and −2.3% (95% CI: −56.2 to 33.0) for A(H3N2). The low VE reflects mismatch between circulating viruses and the 2014/15 northern hemisphere A(H3N2) vaccine strain. Early use of antivirals for prophylaxis and treatment of vulnerable populations remains important.
An extensive body of the literature exists on mathematical and computational models for studying the spatio-temporal dynamics of influenza outbreaks. A main purpose of some of these models is to inform public policy regarding the selection and allocation of public health interventions and resources during a pandemic. Reliable forecasts of measures such as peak time, peak height, and magnitude during an outbreak would inform public health practitioners and healthcare workers on when to expect a surge in demand for healthcare resources and infrastructure and the overall expected public health impact of an outbreak. Although timely forecasts of these measures would be beneficial, making reliable predictions during an outbreak remains a public health challenge.
Several of the major approaches applied to modeling influenza transmission and dynamics have been applied to the forecasting of influenza outbreaks (see Table 1 for brief descriptions).[2-5] These models have been reviewed in the context of pandemic preparedness, control, and mitigation.[1, 6-8] However, there are no reviews discussing the application of these models to the forecasting of influenza outbreaks. The goal of this paper is therefore to present a systematic review of studies that have discussed approaches for influenza forecasting at the local, regional, national, or global level. The main aims are to (i) summarize existing approaches to influenza forecasting, (ii) present differences in measures of accuracy and evaluate the degree to which various performance measures are met, (iii) discuss limitations in the data sources, and parameter estimation that impede forecasting during outbreaks. The motivation of this paper is to inform further research on influenza forecasting and provide researchers and public health practitioners with a summary of the accomplishments and limitations in influenza forecasting.
Article selection and evaluation
The scope of this review included studies designed to predict influenza dynamics at the local, regional, national, or global level. First, we searched PubMed and Google Scholar for articles on influenza forecasting. A search for (“influenza, human”[MeSH Terms] OR (“influenza”[All Fields] AND “human”[All Fields]) OR “human influenza”[All Fields] OR “influenza”[All Fields]) AND (“forecasting”[All Fields] OR “forecasting”[MeSH Terms]) on PubMed retrieved 239 articles. Replacing “forecasting” with “prediction” in the previous search criteria resulted in 370 articles. A Google Scholar search for “influenza forecasting” retrieved 12 000 articles. Next, we focused on articles with “influenza” and “forecasting” or “prediction” in the titles and/or abstracts. Third, we selected articles that mentioned influenza forecasting as one of the aims in the abstract. After eliminating non-English articles, 35 articles remained. Lastly, we excluded articles focusing on topics such as forecasting emergency department visits, which have already been covered in a previous review. The study is therefore based on the remaining 16 articles, which included both prospective and retrospective studies. We group and present studies based on measures predicted.
We acknowledge that there were numerous endeavors made by various research groups and organizations toward real-time forecasting of the 2009 H1N1 pandemic. However, for several of these endeavors, we were unable to find published descriptions of the methodology used in forecasting. A brief description of the modeling approaches in the sixteen selected articles, in addition to advantages and limitations to using these methods for influenza forecasting can be found in Table 1. In Table 2, we present a summary of study characteristics.
The articles in Table 2 aimed to either forecast a single measure or multiple measures. Typical measures predicted included epidemic trend, duration, peak timing, peak height, and magnitude. For simplicity, we grouped these measures into magnitude, peak timing and intensity, and duration. We discuss differences in measures of accuracy, which appeared to depend on the modeling approach and the measure predicted.
Eleven of the sixteen studies forecasted the expected magnitude, daily or weekly influenza activity based on data on confirmed laboratory cases, and/or influenza-like illness. As noted, measures of accuracy differed across studies. Aguirre and Gonzalez, Viboud et al., and Jiang et al. used correlation coefficients to evaluate accuracy in daily and weekly forecasts of influenza activity. The correlation coefficient between the predicted and observed values ranged from 58% to 93·5% depending on the length of the forecasts. Although useful in comparing data trends, correlation coefficients do not measure the closeness of the predicted to the observed values.
On the other hand, the closeness of the predicted to the observed data could be evaluated using different measures of error. For instance, Jiang et al. observed different percent errors depending on when prediction was made. Prediction of the epidemic curve made a few days from the peak had an estimated 10·8% percent error, which was much lower than the 91·6% percent error observed using nine fewer data points. Similarly, Soebiyanto et al. presented several ARIMA models and evaluated accuracy based on the root-mean-squared-error (RMSE) of one-step-ahead predictions. They also considered the effects of including environmental variables such as humidity and temperature. The preferred models had RMSE approximately in the range of 0·47–0·61. Alternatively, Polgreen et al. presented a prediction market for influenza forecasting and assessed accuracy based on the proportion of correct predictions of a particular color code representing a level of influenza activity. The prediction markets yielded correct predictions 71%, 50%, and 43% of the time by the end of the target week, 1 week in advance, and 2 weeks in advance, respectively.
Some of the studies evaluated accuracy using prediction and confidence intervals. For instance, the true incidences were included in the 95% prediction intervals for epidemic forecasts made at the peak and after the peak for the 2009 pandemic in Japan by Nishiura. Predictions made for the 1968–1969 pandemic, also known as the Hong Kong flu, were presented graphically and assessed to have overlapped with the observed data in 42 of 44 cities. Influenza case estimates made by Chao et al. also overlapped with the estimated ranges from the US CDC.
Most of the previous methods were evaluated retrospectively or published after the 2009 pandemic. Towers and Feng presented forecasts of the 2009 pandemic in the US as it unfolded. They predicted the proportion of the infected population at 63% without vaccination and 57% with the inclusion of the planned vaccination scheme in the model. The 57% estimate was much higher than estimates presented by the CDC. However, real-time predictions of outbreak dynamics are extremely difficult compared with retrospective evaluations due to limitations in data and difficulty in obtaining reliable parameter estimates as we later discuss.
Peak timing and intensity
Methods applied to forecasting peak time have been shown to perform reasonably well when reliable data and parameter estimates are used. For instance, during the 2009 pandemic, Towers and Feng predicted that the peak would be observed in the US toward the end of October in week 42 with 95% confidence intervals between weeks 39 and 43. According to CDC reports, H1N1 peaked in the US during the second week of October. Ong et al. also predicted a few weeks in advance that the 2009 pandemic in Singapore would peak at the start of August. However, the peak height was overestimated. Chao et al. also showed that simulated 2009 H1N1 epidemic for LA County peaked at about the same time (mid-November) as reported by the LA county Department of Public Health.
Using web-based estimates of influenza activity, Shaman and Karspeck and Nsoesie et al. retrospectively illustrated that peak time could be predicted as early as 7 and 6 weeks, respectively, before the actual peak for seasonal outbreaks of influenza in the US. Unfortunately, web-based estimates do not always capture trends in influenza activity and could therefore distort accuracy of predicted outcomes.
Studies published before the 2009 pandemic also had some success. For example, the model discussed by Longini et al. retrospectively estimated the peak time for the 1968–1969 Hong Kong influenza pandemic within the 4-day epidemic peak period for 32% of the cities for which morbidity data were available. Using the same model as that discussed in, Longini et al., Aguirre and Gonzalez predicted the 1988 influenza epidemic in Havana, Cuba to peak on March 15th. However, the true peak was observed on March 1st, implying a deviation of approximately 2 weeks. Additionally, Hall et al. showed that pandemic amplitude could be predicted to within 20% and peak timing within a week in retrospective evaluations using ILI and mortality data for three pre-2009 pandemics. Andersson et al. observed a median error of 0·9 weeks and a median deviation of approximately 28% for predictions of the peak time and peak height, respectively, for seven seasonal outbreaks (from 1999 to 2006) in Sweden.
Compared with the other metrics, the peak time appears to be the easiest to forecast. However, forecasting the peak height is more complex and is usually over- or underestimated.
Outbreak duration is typically defined in terms of baseline levels of infection. Compared with the other metrics, fewer papers have focused on predicting outbreak duration. Aguirre and Gonzalez correctly predicted the end of the 1988 epidemic in Havana, Cuba. Based on a retrospective study of three pandemic events, Hall et al. predicted pandemic durations within 2 weeks of the actual duration. In contrast, Hyder et al. retrospectively illustrated that duration could be underestimated by as little as 2 weeks and as much as 14 weeks for some influenza seasons.
The previously discussed results suggest that reliable forecast of influenza dynamics is possible. However, diversity in modeling approaches, and differences in measures of accuracy makes forecast comparison difficult.
The number of new infections at any time during an influenza outbreak depends on several biological, behavioral, and environmental factors that influence the transmission of influenza viruses. These factors include immunity, virulence factors, contact type and patterns, and climatic conditions that influence viral survival. The inclusion of these parameters in models for influenza forecasting could improve forecast accuracy. However, in addition to the difficulty of estimating true influenza incidence from laboratory confirmed cases and ILI, estimating transmission and severity parameters during pandemics remains a challenge. We discuss these challenges.
Unlike seasonal outbreaks of influenza, pandemics are rare and usually result from novel influenza viruses. A meager understanding of the natural history of the virus hinders the estimation of transmission and severity parameters in real time. Estimating the transmission potential of an emerging infection early on is important as it would help determine whether control measures should be varied and whether more stringent measures are required to control or mitigate an outbreak.[24, 25] In several publications, the transmissibility and natural history of influenza have been estimated at the household, school, or community level using observational data.[26, 27] However, data are typically unavailable or incomplete during the early stages of an outbreak resulting from a novel influenza virus.
The disease severity, which is another important measure, is commonly estimated based on case fatality, hospitalization rates, and clinical attack rates. Approximations of case fatality and hospitalization rates could be underestimated due to subclinical and asymptomatic cases. Although clinical attack rates could be estimated at the community level, data on laboratory-diagnosed cases might be delayed. Nevertheless, studies conducted during the 2009 pandemic suggested that estimates of severity and transmissibility improved as the pandemic progressed.[27, 28]
Traditional systems for monitoring ILI and acute respiratory tract infections rely on reports from general practices, family doctor clinics, diagnostic test laboratories, and public health departments for influenza surveillance.[3, 4, 14] There is typically 1–2 week lag(s) in the publishing of reports, and reported cases are sometimes retrospectively adjusted. Additionally, the exact number of influenza cases is unobtainable due to unreported cases and asymptomatic infections.
In view of the challenge in obtaining timely influenza surveillance data from conventional methods, alternative sources of data such as Google Flu Trends have been considered. Google Flu Trends attempts to provide estimates of influenza activity based on Internet search data. Other data sources, such as flu prescription drug sales, nonprescription medication sales, school absenteeism, ILI symptom reports on social media, and emergency department chief complaints, have also been evaluated as proxies for capturing ongoing influenza trends.
Although these novel data sources provide information in near real time, which is useful for daily or weekly forecasts of influenza activity,[18, 19, 30] there are several limitations to using these data. Limitations include reduced application in low-resource countries and deviations from influenza patterns presented by traditional surveillance systems. For example, Cook et al. compared H1N1-related search queries on Google Insight to traditional surveillance data for the H1N1 pandemic in Singapore. The outbreak peaked in August 2009; however, search query data suggested an earlier peak and also decreased to about 20% of the search volume around the epidemic's peak time. Furthermore, during the 2012–2013 influenza season, estimates of influenza activity provided by Google Flu Trends did not match estimates provided by traditional influenza surveillance systems. The challenge therefore remains for timely estimates of influenza activity for weekly forecasts at different geographical levels.
Reliable forecasts of measures such as trend, peak height, and peak time during influenza outbreaks would inform healthcare practitioners on when to expect changes in demand for healthcare resources. Practitioners could therefore prepare for surges in influenza cases by acquiring the necessary resources (such as vaccines and antiviral treatments) and alerting essential personnel (such as nurses and doctors). However, forecasts must be interpretable to be useful. It is therefore important for studies to clearly define the predicted event, the temporal and spatial applicability of the approach, quantify the likelihood of the event either based on a probabilistic statement or relative to other similar events, and highlight the limitations (see Figure 1). In addition, defining a global measure of accuracy for evaluating the correctness of various forecasting methods would ease the process of forecast comparison. Lastly, several of the studies discussed in this review are retrospective. The challenge therefore remains in evaluating and quantifying the performance of these methods in real time.
This work is supported by the Intelligence Advanced Research Projects Activity (IARPA) via Department of Interior National Business Center (DoI/NBC) contract number D12PC000337, and the US Government is authorized to reproduce and distribute reprints for Governmental purposes notwithstanding any copyright annotation thereon.
The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of IARPA, DoI/NBC, or the US Government.
Influenza and Other Respiratory Viruses Volume 8, Issue 3, pages 309–316, May 2014
Forecasting the dynamics of influenza outbreaks could be useful for decision-making regarding the allocation of public health resources. Reliable forecasts could also aid in the selection and implementation of interventions to reduce morbidity and mortality due to influenza illness. This paper reviews methods for influenza forecasting proposed during previous influenza outbreaks and those evaluated in hindsight. We discuss the various approaches, in addition to the variability in measures of accuracy and precision of predicted measures. PubMed and Google Scholar searches for articles on influenza forecasting retrieved sixteen studies that matched the study criteria. We focused on studies that aimed at forecasting influenza outbreaks at the local, regional, national, or global level. The selected studies spanned a wide range of regions including USA, Sweden, Hong Kong, Japan, Singapore, United Kingdom, Canada, France, and Cuba. The methods were also applied to forecast a single measure or multiple measures. Typical measures predicted included peak timing, peak height, daily/weekly case counts, and outbreak magnitude. Due to differences in measures used to assess accuracy, a single estimate of predictive error for each of the measures was difficult to obtain. However, collectively, the results suggest that these diverse approaches to influenza forecasting are capable of capturing specific outbreak measures with some degree of accuracy given reliable data and correct disease assumptions. Nonetheless, several of these approaches need to be evaluated and their performance quantified in real-time predictions.