During 2004–2008, a total of 59,519 diagnostic specimens were tested for influenza virus (Table 1). Over this 5-year period, there was an upward trend in the annual number of specimens tested, increasing from 13.1% of the total specimens in 2004 to 28.4% of the total specimens in 2008. A total of 3,131 (5.3%) specimens were positive for influenza virus. Among the influenza virus–positive specimens, a mean annual proportion of 72.2% (range 57.0%–86.4%) were positive for influenza A and 27.8% (range 13.6%–43.0%) for influenza B.

During 2010–2012, a total of 11,986 specimens from outpatients with ILI were tested for influenza virus. Over this 3-year period, there was a decreasing trend in the annual number of specimens tested; 58.2% of the total specimens were tested in 2010, compared with 17.6% in 2012. A total of 5,618 (46.9%) specimens were positive for influenza virus. Among the influenza virus–positive specimens, a mean annual proportion of 68.8% (range 52.0%–78.1%) were positive for influenza A virus, compared with 31.2% (range 21.9%–48.0%) for influenza B virus.

The higher percentage of influenza virus–positive specimens obtained during the second study period (46.9%), compared with percentage obtained during the first study period (5.3%), was due to the use of a more specific ILI definition for virologic surveillance and a more sensitive diagnostic method (i.e., reverse transcription PCR) during 2010–2012. For the first study period, the weekly number of hospitalizations for pneumonia and influenza and the proportion of influenza-positive diagnostic specimens were significantly correlated (p<0.01) (Figure 1). Likewise, for the second study period, the weekly number of hospitalizations for pneumonia and influenza and the proportion of influenza-positive specimens from outpatients with ILI were also significantly correlated (p<0.01) (Figure 1). The Spearman rank correlation coefficient was 0.587 during the first study period, compared with 0.314 during the second study period, when surveillance sampling was changed. These differences in correlation could be partly due to patient settings: during the first period, the specimens were obtained from patients in public acute-care hospitals, whereas during the second period, specimens were obtained from outpatients with ILI in community settings.

For all study years, except 2007, 2008, and 2011, the age-specific hospitalization rate per 100,000 person-years for pneumonia and influenza was lowest in the 25- to 44-year-old age group; for 2007, 2008, and 2011, the lowest rate was in the 15- to 24-year-old age group (Table 2). The annual hospitalization rate was consistently highest for persons >75 years of age.

In the 2 study periods, the age-specific influenza-associated hospitalization rates per 100,000 person-years for pneumonia and influenza showed a J-shaped pattern (Figure 2). The influenza-associated proportion was highest in children <6 months of age, and it was second highest in adolescents and young adults in the 15- to 24-year-old age group, followed by a decline in the older age groups in each of the 2 study periods.

During both study periods, the influenza-associated hospitalization rate per 100,000 person-years for pneumonia and influenza was highest among persons >75 years of age (452.2 and 458.9 during 2004–2008 and 2010–2012, respectively), second highest among children <6 months of age (213.9 and 363.6 during 2004–2008 and 2010–2012, respectively), and lowest among persons 25–44 years of age (8.1 and 9.7 during 2004–2008 and 2010–2012, respectively) (Table 3). The proportion of influenza-associated hospitalizations for pneumonia and influenza was highest among children <6 months of age (35.2% and 41.3% during 2004–2008 and 2010–2012, respectively). The proportion among persons 15–24 years of age was 25.1% during 2004–2008 and 23.6% during 2010–2012. The overall influenza-associated hospitalization rates per 100,000 person-years (≈30) and proportions (12%) were similar during the 2 study periods.

The weekly proportion of diagnostic specimens with positive test results for RSV was statistically significant in the regression models for 2 age groups (<6 months and 6–23 months of age) in the first study period and 1 age group (5–14 years of age) in the second study period (p<0.05 in these 3 models). Sensitivity analyses showed that the estimates of influenza-associated hospitalization rates and proportions with and without RSV in regression models for all the age groups varied within 2%.

Our analyses show that in Singapore, the age-specific influenza-associated hospitalization rates for pneumonia and influenza followed a J-shaped pattern that was also seen in analyses from other studies (1,7). The influenza-associated hospitalization rate was highest among persons >75 years of age, followed by children <6 months of age.

Our estimates of the number of influenza-associated hospitalizations in Singapore reflect the effect of influenza on the health care system in this tropical setting. The average length of hospital stay for pneumonia and influenza was ≈8 days. The mean annual estimate of influenza-associated hospitalizations for pneumonia and influenza during 2010–2012 was 1,535 (Table 3), which translates to 12,280 patient-days and a daily average of 34 occupied beds. We estimated that an additional 20 beds were occupied each day when influenza activity was at its peak. In the United States, studies estimating the national number of cases and hospitalizations averted by influenza vaccination support the use of influenza vaccination as a central tool for preventing influenza, and they highlight the need for increasing vaccination coverage and the need for more effective vaccines (15,16).

In Singapore, influenza vaccination is a key strategy for reducing the use of influenza-associated hospital services and influenza-associated illness and death. The Singapore MOH Expert Committee on Immunization recommended the use of influenza vaccine to protect vulnerable populations at higher risk for influenza-related complications; these populations include persons >65 years of age, adults and children with chronic medical conditions, pregnant women, and children 6 months to <5 years of age. Influenza vaccination has been shown to be cost-effective among the elderly and those with chronic medical conditions (17–21). Because the risk for influenza-associated hospitalization and death is high among the elderly and there may be lower effectiveness of influenza vaccine in the elderly (22–25), vaccination of household members of persons >65 years of age is encouraged to reduce transmission of influenza virus within the household (26–30).

In our study, children <6 months of age had the second highest age-specific influenza-associated hospitalization rate for pneumonia and influenza. Studies in different geographic regions (e.g., Finland, Hong Kong, and the United States) also found high hospitalization rates associated with influenza in children <1 year of age (6,31–36). Influenza vaccine is licensed only for use in persons >6 months of age. One way to reduce the risk of influenza in those <6 months of age is vaccination of their household contacts and caregivers, as recommended by the Advisory Committee on Immunization Practices in the United States (30).

Our findings underscore the importance of vaccination against influenza virus, in particular for the elderly, who are at higher risk for hospitalization and death. The Health Behavior Surveillance of Singapore, conducted by the Health Promotion Board by telephone and face-to-face with selected Singapore residents 18–69 years of age, showed that the proportion of persons who reported having been vaccinated against influenza in the preceding 12 months decreased significantly from 15.8% in 2010 to 11.2% in 2012 (p<0.001) (Health Promotion Board, unpub. data). In the 2012 Health Behavior Surveillance of Singapore, only 8.7% of adults 50–69 years of age reported having been vaccinated against influenza; this vaccine uptake percentage was half that for young adults 18–29 years of age (16.9%). In Hong Kong during the 2012–13 influenza season, vaccination coverage in the general population was 14.0%; the proportion of vaccinated persons was highest among persons >65 years of age (39.1%) and second highest among children 6 months to 5 years of age (28.4%) (37). In the United States during the 2011–12 influenza season 44.3% of children 6–23 months of age received full vaccination, and 38.3% of persons >18 years of age were vaccinated (38). The National Health Interview Survey in the United States showed an increase in influenza vaccination coverage by age: 26.1% coverage among persons 18–49 years of age, 44.0% in persons 50–64 years of age, and 69.4% in persons >65 years of age.

Various measures have been implemented to raise awareness and increase influenza vaccine uptake in Singapore. For example, since 2014, the use of Medisave, a compulsory national health care savings scheme, has been allowed for payment of seasonal influenza vaccination in populations at high risk for influenza-associated hospitalization or death, including children 6 months to <5 years of age and persons >65 years of age. In addition, the Health Promotion Board and health care providers have made consistent efforts to educate the public on the importance of influenza vaccinations.

In Singapore, the proportion of influenza-associated hospitalizations for pneumonia and influenza was highest in children <6 months of age and second highest in the 15- to 24-year-old age group. There are many factors that may have affected this pattern; for example, hospitalizations for pneumonia and influenza attributable to viruses other than influenza, and the rate of testing by age group may have had an effect (Figure 2). In addition, there may be a higher propensity for testing and hospitalizing children <6 months of age compared with persons in older age groups.

This study in Singapore provides population-based estimates of influenza-associated hospitalizations for pneumonia and influenza, which enables comparison with estimates for other countries. In Singapore, the overall influenza-associated hospitalization rate per 100,000 person-years for pneumonia and influenza was 28.3 during 2004–2008 and 29.6 during 2010–2012. The proportion of influenza-associated hospitalizations was 11.9% and 11.2%, respectively, for the same years. A similar study in Hong Kong (a subtropical location), using Poisson regression based on data from 1996–2000, estimated that the influenza-associated hospitalization rate per 100,000 person-years for all ages was 29.3 and that 11.6% of all hospitalizations for pneumonia and influenza were attributable to influenza (7). In a study in the United States during 1979–2001, the overall rate of influenza-associated hospitalizations per 100,000 person-years for pneumonia and influenza was 36.8, and the proportion of influenza-associated hospitalizations was 8.6% (1).

In addition, during the 2 study periods (2004–2008 and 2010–2012) in our study, the rates and proportions of influenza-associated hospitalizations for pneumonia and influenza among the elderly (Table 3) were higher than those in Hong Kong during 1996–2000. In Hong Kong, the influenza-associated hospitalization rate per 100,000 person-years was 58.7 (95% CI 43.3–73.7) for persons 65–74 years of age and 176.3 (95% CI 119.2–231.0) for persons >75 years of age; the influenza-associated proportion was 11.0 (95% CI 8.1–13.8) and 7.1 (95% CI 4.8–9.3) in these 2 age groups, respectively (7). In Singapore, the influenza-associated hospitalization rate per 100,000 person-years among children <5 years of age was 92.3 (95% CI 74.1–118.3) during 2004–2008 and 65.8 (95% CI 36.5–114.6) during 2010–2012 (results not shown); these rates far exceeded the rate of 18.5 in the same age group in the United States during 1979–2001 (1). Various factors must be considered when comparing influenza-associated hospitalization rates and proportions by country. For example, comparisons should consider the age structure of the populations; influenza vaccination coverage, the level of herd immunity, and access to and utilization of health care services within the communities; the socioeconomic profile of the communities; and the climatic variables and severity of influenza epidemics in the communities. There are also variations in the statistical models, data aggregation, study periods, sampling protocols, and the coverage of laboratory virus surveillance systems.

Our study has several limitations. First, influenza-associated hospital admissions for pneumonia and influenza could be affected by many factors, such as uptake of influenza vaccines in the population, dominant influenza virus types/subtypes and their antigenic drifts and shifts, seasonal variations in vaccine match to circulating influenza strains, changes in admission criteria and diagnostic practices in hospitals, and variations in health care–seeking behavior. Second, although we controlled for potential confounding factors in our modeling approach, there may still be several unmeasured factors that could affect the estimates of influenza-associated hospitalizations for pneumonia and influenza. Third, in 2012 in Singapore, the diagnosis coding system for hospitals changed from ICD-9 to ICD-10; this change could have complicated the practice and assessment of cause-specific diagnoses. Last, our analysis was confined to hospitalizations for which the principal discharge diagnoses were pneumonia and influenza. However, during influenza seasons hospitalization rates increased for conditions other than pneumonia (e.g., acute bronchitis, chronic respiratory disease, and congestive heart failure) (39). In our study, we did not estimate the effect of hospitalizations for these other influenza-attributable conditions.

Our findings allude to the importance of surveillance data for monitoring the effect of influenza and for assessing changes in influenza dynamics, as determined on the basis of a well-integrated virologic and epidemiologic surveillance system. For 2 reasons, we assumed that variation in influenza virologic surveillance by the MOH accurately reflected the patterns of influenza circulation in the general Singaporean population of Singapore: 1) during 2004–2008, the diagnostic respiratory specimens we used were from all public acute-care hospitals, and 2) the specimens during 2010–2012 were from outpatients with ILI at government primary care clinics and private general practitioner clinics that were geographically spread out across Singapore. In Singapore, 80% of the primary health care services are provided by private general practitioners; government clinics provide the remaining 20% of the primary health care services. The opposite is true for hospitalization care: public sector hospitalizations constitute 80% of all hospital admissions, and private sector hospitalizations constitute the remaining 20% (40). However, over time, there could be variation in laboratory testing patterns and sentinel sites for specimen collection in the community. Although we were unable to assess the effect of this variation, which may be somewhat reflected in the annual number of specimens tested, our modeling study was supported by sufficient year-round data for virologic surveillance.

Our findings have obvious policy implications. The J-shaped pattern observed for influenza-associated hospitalization rates during the 2 study periods was also seen in all of the individual years, both before and after the influenza A(H1N1)pdm09 pandemic. Our findings also underscore the importance of continuous surveillance in Singapore to identify populations at high risk for influenza-associated hospitalization or death and to guide public health policy priorities. The excess hospitalization estimates for pneumonia and influenza in our study reflect the considerable effect of influenza in Singapore, particularly among the rapidly aging population.