National burden of hospitalized and non‐hospitalized influenza‐associated severe acute respiratory illness in Kenya, 2012‐2014

Background Influenza‐associated respiratory illness was substantial during the emergence of the 2009 influenza pandemic. Estimates of influenza burden in the post‐pandemic period are unavailable to guide Kenyan vaccine policy. Objectives To update estimates of hospitalized and non‐hospitalized influenza‐associated severe acute respiratory illness (SARI) during a post‐pandemic period (2012‐2014) and describe the incidence of disease by narrow age categories. Methods We used data from Siaya County Referral Hospital to estimate age‐specific base rates of SARI. We extrapolated these base rates to other regions within the country by adjusting for regional risk factors for acute respiratory illness (ARI), regional healthcare utilization for acute respiratory illness, and the proportion of influenza‐positive SARI cases in each region, so as to obtain region‐specific rates. Results The mean annual rate of hospitalized influenza‐associated SARI among all ages was 21 (95% CI 19‐23) per 100 000 persons. Rates of non‐hospitalized influenza‐associated SARI were approximately 4 times higher at 82 (95% CI 74‐90) per 100 000 persons. Mean annual rates of influenza‐associated SARI were highest in children <2 years of age with annual hospitalization rates of 147 (95% CI of 134‐160) per 100 000 persons and non‐hospitalization rates of 469 (95% CI 426‐517) per 100 000 persons. For the period 2012‐2014, there were between 8153 and 9751 cases of hospitalized influenza‐associated SARI and 31 785‐38 546 cases of non‐hospitalized influenza‐associated SARI per year. Conclusions The highest burden of disease was observed among children <2 years of age. This highlights the need for strategies to prevent influenza infections in this age group.


| INTRODUCTION
Influenza is an important cause of respiratory illness in Kenya, accounting for up to 27% of all medically attended cases of respiratory illness. 1 However, most Kenyan studies have focussed on providing subnational estimates of influenza disease burden. [2][3][4][5][6][7][8][9] The only national estimates of the incidence of influenza-associated severe acute respiratory illness (SARI) were calculated for the period July 2009-August 2011, 10 when the influenza A(H1N1)pdm09 (H1N1pdm09) virus circulated at markedly high levels 11 and presented with unusual epidemiology, affecting younger adults and sparing those ≥65 years of age. 12 As H1N1pdm09 virus circulated in subsequent years, the severity and epidemiology of illness associated with the pandemic influenza strain could have changed, as has been observed for other pandemic viruses. 13 Previous national estimates of the incidence of influenzaassociated SARI described the burden of severe influenza in children <5 years of age, and persons aged ≥5 years. From studies conducted in temperate countries, the risk of influenza is known to vary by age, being highest at extremes of age. 14 Moreover, children <6 months of age are not eligible for vaccination and the World Health Organization (WHO) recommends vaccination for children 6-23 months of age. 14 Here, we adapted the methodologic approach used previously, 10 to update estimates of the burden of hospitalized and non-hospitalized influenza-associated severe acute respiratory illness (SARI) during a non-pandemic period (from 2012 through 2014), and describe the incidence of severe influenza disease by narrow age categories to include individuals <6 months of age, <2 years of age and ≥65 years of age which would be useful when considering interventions. The influenza vaccine is not included as part of the national immunization program in Kenya. Our results could inform Kenyan health authorities about the relative value of introducing influenza vaccination in specific age groups.

| METHODS
We adapted a previously described method used to estimate the burden of severe influenza disease in Kenya. 10 We calculated base rates of hospitalized SARI from Siaya County Referral Hospital (SCRH) in Nyanza region. These rates were extrapolated to the other regions in the country, by adjusting for regional risk factors for SARI and hospitalization. To calculate rates of non-hospitalized SARI, non-hospitalized SARI cases were indirectly estimated from healthcare utilization data on the proportion of acute respiratory illnesses that were hospitalized in Nyanza region. Region-specific data on the proportion of influenzaassociated SARI were used to estimate regional rates of hospitalized and non-hospitalized influenza-associated SARI cases.
There were a few differences between our methodology and the methodology originally proposed by Fuller et al. 10 A description of these differences and their possible impact on the results of the study, including the equations used in the analysis, have been described in detail in the Appendix S1.

| Base site
SCRH is located in Nyanza region which has the highest HIV prevalence at 15% 15 and the highest mortality among children aged <5 years within the country at 82 deaths per 1000 live births. 16 SCRH is embedded in a Health and Demographic Surveillance System (HDSS) in Karemo division. The Karemo division of the HDSS has a population of approximately 80 000 inhabitants. 5 We used SCRH and HDSS data from January 2012 through December 2014 as the basis for estimating the incidence of hospitalized SARI in the Nyanza region (henceforth referred to as base rates), which were then extrapolated to the other regions using the steps outlined further below.

| Calculation of base rates of hospitalized SARI
Hospitalized SARI was clinically defined as a patient hospitalized with acute onset illness (within 14 days), reported fever (or recorded temperature of ≥38°C), and a cough. We calculated the annual agespecific base rates of hospitalized SARI by dividing the age-specific number of hospitalized SARI patients at SCRH who were enrolled in the HDSS, by the age-specific population of Karemo residents within the HDSS. These base rates of hospitalized SARI, which were assumed to be representative of the Nyanza region, were calculated for each of the following age groups: 0-5 months, 6-11 months, 12-23 months, 2-4 years, 5-14 years, 15-49 years, 50-64 years, and ≥65 years (Appendix S1: Equation 1).

| Calculation of regional rates of hospitalized SARI
To obtain region-specific rates of hospitalized SARI, we first calculated an adjustment factor for each of the other 7 regions in Kenya.
For children <5 years old, 6 risk factors for SARI were used for adjustment: malnutrition (weight for age Z-score ≤−2), low birthweight (<2500 g), non-exclusive breastfeeding (during the first 4 months of life), household pollution (as indicated by use of solid fuels for cooking), crowding (≥5 persons in a household), and HIV prevalence. For individuals ≥5 years old, 3 SARI risk factors were used for adjustment: household air pollution, crowding, and HIV prevalence. The regional adjustment factor was multiplied by the base SARI rate to estimate the region-specific hospitalized SARI rates (Appendix S1: Equation 3).

| Calculation of regional rates of non-hospitalized SARI
The rates of non-hospitalized SARI were estimated using data on selfreported episodes of pneumonia from a healthcare utilization survey (HUS) that was conducted in the Nyanza region. 22 In the HUS, pneumonia was defined as reported episodes of acute respiratory illness (ARI) characterized by cough and difficulty breathing for more than 2 days, or a diagnosis of pneumonia by a healthcare worker within the last 12 months. For the purposes of this study, non-hospitalized pneumonia as described in the HUS was equated to non-hospitalized SARI. The definition of pneumonia in the HUS could have included cases of ARI that do not require hospitalization, and consequently, the definition of SARI in the current study differs from the WHO definition that all cases of SARI require hospitalization. 23 The HUS provided data for 2 broad age groups: individuals <5 and ≥5 years of age.
As the HUS data were only available for the base region, we extrapolated the percentage of all those who reported pneumonia that resulted in hospitalization to the other regions by multiplying this percent by the proportion of individuals seeking healthcare for ARI for each region, compared to the base region (Appendix S1: The total rates of regional SARI (combining hospitalized and nonhospitalized SARI) in the community were calculated by multiplying the region-specific rate of hospitalized SARI by the reciprocal of the estimated proportion of those with pneumonia who were hospitalized.
In the final step, the rate of non-hospitalized SARI was calculated as the difference of the overall SARI rates in the community and the rate of hospitalized SARI (Appendix S1: Equation 4).

| Calculation of regional rates of hospitalized and non-hospitalized influenza-associated SARI
The proportion of SARI due to influenza was collected from six influenza surveillance sites spread across the country: SCRH representing Nyanza region, Kakamega County Referral Hospital (CRH) representing the Western region, Mombasa CRH representing the Coast region, Nakuru CRH representing the Rift Valley region, Nyeri CRH representing the Central region, and Kenyatta National Hospital (KNH) representing the Nairobi region ( Figure 1). As surveillance data were not available in the Eastern and North Eastern regions during the study period, surveillance data from the sites in the neighboring regions were used to estimate the average proportion of hospitalized influenza-associated SARI that was assigned to these two regions.
The region-specific rates of hospitalized and non-hospitalized influenza-associated SARI were calculated by applying the proportion of SARI cases that tested positive for influenza virus in each site to the regional rates of both hospitalized and non-hospitalized SARI (Appendix S1: Equation 6 and 7).

| Calculation of the number of hospitalized and non-hospitalized SARI cases
To obtain the number of cases, the region-specific rates of hospitalized and non-hospitalized SARI were multiplied with regional population data from the national census 18 (Appendix S1: Equations 8 and 9).
The national census data of 2009 were projected to the study period using the annual population growth rate. 24 A similar approach was used to estimate the region-specific number of hospitalized and nonhospitalized influenza-associated SARI cases (Appendix S1: Equations 10 and 11).

| Data analysis
Data analysis was performed using Microsoft Excel, R version 3.3.1 25 and Stata version 13.0. 26 Confidence intervals (CIs) of the regional SARI rates were estimated by running 1000 iterations of each of the risk factors for hospitalized SARI (Table S1) by allowing the specific risk factor to vary within their 95% CI limits, while keeping all the other risk factors constant. The 2.5th and 97.5th values of these outputs were presented, respectively, as the lower and upper limits of the 95% CI of the regional hospitalized SARI rates. All rates were reported per 100 000 persons.

| Ethical review
The study protocol at SCRH was approved by both the ethical review
Rates of hospitalized and non-hospitalized SARI were highest in 2012 and lowest in 2014 ( Figure 2).   Table 2). Rates of hospitalized influenza-associated SARI were highest in 2012 and lowest in 2014 ( Figure 5).

| DISCUSSION
From 2012 through 2014, we estimated that the annual rate of hospitalized influenza-associated SARI was 21 per 100 000 population, while the annual rate of non-hospitalized influenza-associated SARI was 4 times higher at 82 per 100 000 population. Despite regional and year-to-year variation on rates of influenza-associated SARI, the group consistently most at risk for hospitalization was children <5 years of age, and especially those <2 years of age. Among older age groups, those 5-14 years of age and adults ≥65 years of age had the highest hospitalization rates, though not in the same magnitude as that reported for young children.
Our findings support the hypothesis that the burden of influenza in Kenya during the influenza A(H1N1)pdm09 pandemic period (i.e. 2009 to 2010) was higher than during non-pandemic years.  Our rates of influenza-associated hospitalization among those <2 years were twice as high as that among children 2-4 years.
Children <2 years of age are at high risk for severe influenza disease. 14 Among the <2 years age group, rates of influenza-associated hospitalization were estimated to be higher in the 6-23-month age group than in the 0-5-month age group. These findings are in contrast to the USA, where rates of influenza-associated hospitalization are higher in the 0-5-month age group than in the 6-23-month age group 30  if signs/symptoms were associated with pre-existing underlying conditions or resulted in atypical presentations of influenza. 33,34 We also found that children 5-14 years of age had high rates of influenza-associated hospitalization (9-10 per 100 000 persons) among those ≥5 years of age, similar to rates among the elderly (9-11 per 100 000 persons). Similarly in India, children 5-14 years of age had the second highest rates of influenza-associated hospitalizations among individuals ≥5 years of age. 35 In the India study, rates of hospitalization among school going children were 2 to 5 times higher than those observed among the elderly. While these findings may reflect differences in healthcare seeking between younger and older individuals, the results also suggest that school going children may be an important risk group to consider for influenza prevention strategies.
Our study had a few limitations. During our study period, patients could have sought care in facilities other than SCRH, leading to an underestimation of our SARI rates. The calculation of non-hospitalized SARI rates is based on a healthcare utilization survey conducted >10 years ago that described the proportion of individuals aged <5 years and ≥5 years of age who were hospitalized with pneumonia.
It is important to note that healthcare utilization may have changed over the past 10 years and healthcare-seeking behavior may vary with age, especially among the elderly, which could impact the precision of non-hospitalized case estimation. It may also be argued that hospitalized influenza-associated SARI is in fact a more severe form of respiratory illness than non-hospitalized influenza-associated SARI, because individuals with more severe disease are more likely to seek medical care. 22 The hospitalized and non-hospitalized SARI case definitions used in the study would therefore not refer to illnesses of equal severity. In spite of these limitations, we were able to provide reasonable estimates of rates of hospitalized and non-hospitalized influenza-associated SARI, enabling appreciation of influenza as an important public health matter in Kenya.

| CONCLUSION
Influenza virus is associated with a substantial amount of severe respiratory disease burden in Kenya, especially among very young children. Children <2 years of age could be considered as a priority group for influenza vaccination in Kenya. As influenza vaccine is not licensed for children <6 months of age, maternal vaccination of pregnant women could be considered to protect this age group.