In August 2011, we began an observational, prospective cohort study of children at the Queen Sirikit National Institute of Child Health (QSNICH), the largest public tertiary care hospital for children, in Bangkok, Thailand. Children were eligible for enrollment if they were residents of metropolitan Bangkok, 0–36 months of age, routinely sought care at QSNICH, and not acutely ill at enrollment. We aimed to enroll an equal number of healthy children and children with underlying comorbidities (prematurity/low birth weight [<37 weeks gestation or <2500 gram], congenital heart disease, chronic respiratory disease, immunosuppression, neurologic/neuromuscular conditions, and metabolic abnormalities). High-risk children were identified and approached when seeking care at clinics. An age- and time-matched child without any underlying comorbidities was then subsequently enrolled from the neonatal ward, well-baby clinic, or other outpatient clinics. The sequence of enrollment could also be reversed if a healthy child was first identified. Children who were already enrolled were followed through in the study even if their matches could not be found. After the consent process for which study nurses explained in detail about the study but did not discuss about specific respiratory disease prevention including vaccination, written informed consent was obtained from a parent or guardian of each child. This included consent to allow researchers to review children’s medical records to verify influenza vaccination status. Children were followed for 2 years. This study was approved by the ethics committee of the QSNICH.

At enrollment, a trained study nurse administered a standardized questionnaire to each child’s caretaker to obtain baseline demographic and clinical data. Within 2 months of enrollment, the nurse visited the child’s home and conducted a survey to record family size, structure, and socioeconomic data. Each child’s vaccination card was reviewed at this home visit and again every 6 months. Study nurses called households weekly to assess whether the enrolled children had been ill with any new episode of ARI within the preceding 7 days. An ARI episode was defined as ≥2 of the following: fever, cough, sore throat, and runny nose. A new episode required ≥14 days of no symptoms between illnesses. For new episodes, the caretakers were asked to bring the children to the clinic to be evaluated, complete a brief clinical questionnaire, and have combined nasal and throat swabs collected. The caretakers were asked to contact the research team if the children developed symptoms requiring a clinic visit prior to weekly calls. Specimens were tested for influenza viruses by real-time reverse transcription–polymerase chain reaction (rRT– PCR).18 All children were treated by an attending physician according to Thailand’s standard of care. Study nurses called the caretakers 1–2 weeks later for a follow-up questionnaire of disease outcome.

Because influenza viruses circulate year-round with a peak during June–October,19 we defined influenza season as June through May of the following year (e.g., June 2012–May 2013). Full vaccination was defined as having received two vaccine doses administered ≥28 days apart in the current season or two doses administered ≥28 days apart in any previous season and one dose in the current season.20 Partial vaccination was defined as having received one dose in the current season and never being fully vaccinated in any previous season. A child was classified as being unvaccinated if they were not vaccinated in the given season or if they received the first of two recommended influenza vaccine doses within 14 days before ARI onset during the given season. While we were able to collect data on vaccination status up to a year before study enrollment, ARI onset data, used to calculate vaccine effectiveness (VE), were only available from the start of the study (August 2011) onward. During the two seasons of this study, six children received two doses of influenza vaccine <28 days apart (range, 25– 27 days); they were classified as partially vaccinated for the purpose of analysis.

All seasonal influenza vaccines licensed and used in Thailand were trivalent, inactivated vaccines. Although both Northern Hemisphere and Southern Hemisphere vaccines were available for use in Thailand, all government purchased vaccine (about half the supply) is Southern Hemisphere vaccine and the majority of the privately purchased vaccine is Southern Hemisphere vaccine.16 The Southern Hemisphere vaccines for the 2011–2012 and 2012–2013 seasons included the same influenza strains: an A/California/7/2009 (H1N1) pdm09-like virus, an A/Perth/16/2009 (H3N2)-like virus, and a B/Brisbane/60/2008-like virus.21 The 2011–2012 Northern Hemisphere vaccine was also the same; however, the 2012–2013 Northern Hemisphere vaccine changed the H3 and B components (A/Victoria/361/2011 (H3N2)-like virus and B/Wisconsin/1/2010-like virus).

Descriptive analysis was conducted to examine baseline demographic, socioeconomic, and clinical characteristics. For vaccination coverage and VE estimation, only children aged ≥6 months at the beginning of each influenza season were included in the analysis. Vaccination coverage was calculated by season. Only children with an onset date for the ARI were included in the VE calculation. VE was estimated as 100% x (1-hazard ratio) using the Cox proportional hazards approach. We used the Schoenfeld global goodness-of-fit test to test proportional hazards assumption. Other covariates included age at ARI, underlying medical condition (present/absent), and daycare attendance. Fully, partially, and any vaccinated children were compared with unvaccinated children. An influenza case was defined as an ARI episode for which the specimens tested positive for an influenza virus by rRT–PCR.

We conducted several sensitivity analyses in which we (i) shortened the influenza season from 12 months to 6 months,(ii) excluded ARI episodes in which the child took oseltamivir prior to swab collection, (iii) excluded influenza A (H3N2) from the 2012–2013 dataset because a drifted H3N2 strain emerged in 2012 affecting vaccine match to the circulating strains, (iv) included children with the 1st two doses administered <28 days apart as fully vaccinated, (v) excluded ARI episodes with specimens collected >7 days after illness onset, and (vi) shortened the time between last vaccination and onset of illness from 14 to 7 days. In addition, we examined the effect of prior vaccination on subsequent VE and if the VE differed by age and time since vaccination.

We calculated an ad hoc sample size. To detect a VE of 60% with significant 95% confidence limits, a minimum of 25 influenza cases among vaccinated children and five cases among unvaccinated children would be needed given the current follow-up time for each group. We defined a two-sided P-value of ≤0[C1]05 as statistical significance. All analyses were performed using Stata software version 11.0 (StataCorp LP, College Station, TX, USA).

From August 2011–May 2013, we enrolled 968 children aged 0–36 months (median age at enrollment, 10[C1]3 months; interquartile range [IQR], 4[C1]7–19[C1]7 months). Of those, 948 (97[C1]9%) had available demographic data and known influenza vaccination status (for those aged ≥6 months) and were included in the analytic dataset. The median duration of follow-up was 50 weeks (IQR, 22–75 weeks). From August 2011–May 2013, there were 2134 respiratory illness episodes and 1660 (77[C1]8%) met the ARI definition. Of the 1660 ARI episodes, 154 (9[C1]2%) episodes, occurring in 135 children, were in children treated with oseltamivir prior to swab collection (30 [19[C1]5%] of these had laboratory-confirmed influenza). Among the 948 children included in the analysis, 512 (54[C1]0%) were male and 394 (41[C1]6%) had ≥1 underlying medical conditions (Table 1). The most common medical condition was prematurity/low birth weight, accounting for 25[C1]7% of all enrolled children, followed by heart disease (11[C1]6%) and chronic respiratory disease (10[C1]0%). The mothers of 23 children (2[C1]4%) reported receiving influenza vaccine during pregnancy. Most of the caretakers (39[C1]3%) finished secondary school. Thirty percent of the families earned between 10 000 and 19 999 Thai Baht/month (average monthly household income of a Thai family in 2011 was 23 236 Baht22; 30 Thai Baht = 1 US dollar). Most families (49[C1]5%) had 2–4 members (average household size for a Thai family in 2010 was 3[C1]2).22

At the beginning of the 2011–2012 season, 317 children aged ≥6 months were eligible for influenza vaccination and were included in the analysis. In the subsequent season, 339 children were newly enrolled and eligible for vaccination, resulting in the analytic dataset having 656 children in the 2012–2013 season. Baseline characteristics of children by vaccination status are shown in Table S1. Characteristics statistically associated with vaccination status were further explored in the analysis examining association between influenza virus infection and vaccination status.

Influenza vaccination coverage was fairly low in this cohort of young children in Bangkok, possibly due to a relatively limited vaccine supply compared to the size of the population in vaccination target groups. Our estimate of influenza VE in the 2011–2012 season did not reach statistical significance due to the small number of influenza virus infections in the cohort. However, the point estimate of influenza VE suggests that the 2012 Southern Hemisphere vaccine was moderately effective (64%) at protecting young children against influenza and this increased to 79% among the target age-group; these data are consistent with influenza VE estimates among children in this age-group from other countries.23–32

In absolute terms, full vaccination coverage in our cohort of young children was low at 30%. However, our coverage estimates are much higher than preliminary estimates from the Universal Coverage Scheme during the same period (2[C1]3%).17 They are also higher than estimates from Canada (4–6% fully vaccinated during the 2006–2009 seasons) and are comparable to estimates after the first few years of influenza vaccination recommendations for children in the United States (4[C1]4% in 2002–2003, 8[C1]4% in 2003–2004, 17[C1]8% in 2004–2005, 20[C1]6% in 2005–2006, 21[C1]3% in 2006–2007, and 23[C1]4% in 2008–2009).33–40 Compared to the national estimates, coverage may have been higher in our cohort for several reasons, including increased access to vaccine through the private sector due to residence in a metropolitan area, increased awareness among caretakers because of the 2009 influenza pandemic, increased attention to preventing influenza in children among staff at the study site which is a tertiary care academic pediatric facility, closer residence to the study facility which facilitated easy access to medical care, and increased proportion of children with comorbidities who are a recommended priority group for influenza vaccination. Fifty percent of eligible unvaccinated children received the first dose of influenza vaccine after enrolling in the study and receiving regular follow-up at the study facility. Although the vaccination coverage findings from this study may not be generalizable to all Thai children, they highlight that influenza vaccination coverage among young children remains suboptimal, leaving much room for improvement. Furthermore, influenza vaccination in Thailand is considered optional in children unless they belong to one of the risk groups recommended for vaccination by the Ministry of Public Health, and because of limited vaccine supply (approximately 25% of all recommended individuals), vaccination coverage remains low.

In general, we found that medically high-risk and healthy children had different influenza vaccination coverage, and we observed some differences in their household and demographic characteristics. It is possible that these children were not drawn from the same population. Nonetheless, children in our cohort came from families comparable to other Thai families in terms of household income and number of household members.

The VE estimates for both years were consistent with the limited data from children <36 months of age reported elsewhere from earlier years. In other countries, the influenza VE in this population has ranged from 47 to 89%.23–32

Although globally there was concern that the match between circulating influenza A (H3N2) strains and that component of the vaccine in 2012 was poor41,42 (26% antigenic match in Thailand, M. Chittaganpitch, personal communication), exclusion of those episodes did not increase VE in our study. However, it may be that we only had a few ARI episodes associated with influenza A (H3N2) in our dataset.

Our study has several strengths and limitations. Ascertainment bias of influenza status was minimized using rRT– PCR, an assay that is maximally sensitive and specific for influenza viruses. Further, this study was conducted in a defined cohort with nearly 100% of ARI episodes managed at our study hospital, hence reducing potential for missed influenza cases. The bias from overreporting of children’s vaccination uptake by parents was minimized in our study by the review of vaccination card, a preferable method for determining vaccination status in young children.43 Nonetheless, children who got vaccinated may have been different from those who did not in ways that our study did not measure, and children in our cohort were not necessarily representative of those in greater Bangkok. Additionally, we calculated VE for Southern Hemisphere vaccines yet each season theoretically spanned both Southern Hemisphere and Northern Hemisphere vaccine distribution which could have attenuated our VE estimates. However, during this time, most vaccine in Thailand was Southern Hemisphere vaccine and our sensitivity analysis of the first 6 months of data did not change the results. Nevertheless, future VE studies in Thailand should ascertain which vaccine formulation was received to improve precision of the VE estimate. Finally, the study was not designed to estimate VE so was underpowered in the 2011–2014 season.

To our knowledge, this is the first study in Thailand to assess influenza vaccination coverage and effectiveness among young children. Influenza viruses frequently undergo antigenic drift necessitating annual reformulation of seasonal influenza vaccines. Hence, continued evaluation of influenza vaccination coverage and effectiveness is merited to evaluate national influenza vaccination program impact among young children who are at high risk of severe influenza.