The national burden of influenza‐associated severe acute respiratory illness hospitalization in Rwanda, 2012‐2014

Background Estimates of influenza‐associated hospitalization are severely limited in low‐ and middle‐income countries, especially in Africa. Objectives To estimate the national number of influenza‐associated severe acute respiratory illness (SARI) hospitalization in Rwanda. Methods We multiplied the influenza virus detection rate from influenza surveillance conducted at 6 sentinel hospitals by the national number of respiratory hospitalization obtained from passive surveillance after adjusting for underreporting and reclassification of any respiratory hospitalizations as SARI during 2012‐2014. The population at risk was obtained from projections of the 2012 census. Bootstrapping was used for the calculation of confidence intervals (CI) to account for the uncertainty associated with all levels of adjustment. Rates were expressed per 100 000 population. A sensitivity analysis using a different estimation approach was also conducted. Results SARI cases accounted for 70.6% (9759/13 813) of respiratory admissions at selected hospitals: 77.2% (6783/8786) and 59.2% (2976/5028) among individuals aged <5 and ≥5 years, respectively. Overall, among SARI cases tested, the influenza virus detection rate was 6.3% (190/3022): 5.7% (127/2220) and 7.8% (63/802) among individuals aged <5 and ≥5 years, respectively. The estimated mean annual national number of influenza‐associated SARI hospitalizations was 3663 (95% CI: 2930‐4395—rate: 34.7; 95% CI: 25.4‐47.7): 2637 (95% CI: 2110‐3164—rate: 168.7; 95% CI: 135.0‐202.4) among children aged <5 years and 1026 (95% CI: 821‐1231—rate: 11.3; 95% CI: 9.0‐13.6) among individuals aged ≥5 years. The estimates obtained from both approaches were not statistically different (overlapping CIs). Conclusions The burden of influenza‐associated SARI hospitalizations was substantial and was highest among children aged <5 years.


| INTRODUCTION
Influenza virus infections cause substantial morbidity and mortality globally, in particular among individuals aged <5 and ≥65 years and persons with underlying medical conditions. [1][2][3][4][5][6] In addition, a higher burden of influenza-associated hospitalization has been reported among African children compared to other regions. 2,3 Influenza sentinel surveillance among patients hospitalized with severe acute respiratory illness (SARI) has been established in several African countries in the past decade, 7 and such data have contributed to global studies on influenza-associated hospitalization. 2,3 Nonetheless, such studies were limited to a pediatric population.
Estimates of the national burden of influenza-associated hospitalization across age groups are severely limited in Africa, having been described only in 3 countries on the continent. [8][9][10] Although the burden of influenza-associated severe illness may be higher in Africa compared to other regions, 2,3 the use of influenza vaccine and antivirals on the continent, including Rwanda, remains limited. 11 The World Health Organization (WHO) under the Global Influenza Program highlighted that there is a need for influenza disease burden estimates especially from low-and middle-income countries. 12 These estimates would enable governments to make informed evidence-based decisions when allocating scarce resources and planning intervention strategies to limit the impact and spread of the disease. In addition, national estimates would contribute to the global understanding of the burden of influenza-associated severe illness and guide policies at national and global levels.
We aimed to estimate the national and provincial number and rates of SARI and influenza-associated SARI hospitalizations among different age groups in Rwanda from January 2012 through December 2014.

| Data source 2: Retrospective record review of respiratory admissions in selected hospitals
To assess the completeness of data reported through the HMIS and estimate the proportion of SARI cases among any respiratory admission, we implemented an anonymized retrospective record review (using hospital admission books) of any respiratory admission in 6 (13%) of 45 public hospitals conducting influenza sentinel surveillance situated in each of the 5 provinces of the country over the same study period. Consistent with HMIS reporting practices, respiratory admissions were considered patients admitted with a clinician diagnosis of lower respiratory tract infection, pneumonia, bronchopneumonia, bronchitis, or any other pneumopathy. For each identified respiratory admission, gender, age, presence of fever or history of fever and cough, duration of symptoms, and date and place of admission were recorded.

| Data source 3: Influenza virus surveillance among patients hospitalized with severe acute respiratory illness
We conducted active prospective hospital-based surveillance for SARI at 6 public hospitals located in each of the 5 provinces of Rwanda A case of SARI was defined as a hospitalized person of any age presenting with either temperature ≥38°C or history of fever and cough of duration of ≤10 days. 15 The procedures of this surveillance program have been previously described. 16 Briefly, surveillance nurses completed case report forms that included demographic, clinical and epidemiological information for all enrolled SARI cases. In addition, respiratory specimens (nasopharyngeal and oropharyngeal swabs) were collected from all enrolled patients, placed in universal transport medium, stored at Conclusions: The burden of influenza-associated SARI hospitalizations was substantial and was highest among children aged <5 years.

K E Y W O R D S
burden, hospitalization, influenza, Rwanda, severe acute respiratory illness

Funding information
This work was supported by the Rwanda Ministry of Health and the US Centers for Disease Control and Prevention (co-operative agreement number; 1 U51IP 000523); APC for this article is funded by World Health Organization (WHO).
[Correction added on 25 January 2018, after first online publication: Funding information has been added.] 4-8°C and transported to the National Reference Laboratory in Kigali within 72 hours of collection for testing. Specimens were tested for influenza type A and B viruses using a real-time reverse transcriptase polymerase chain reaction assay. 16 Influenza A-positive samples were further subtyped. 17

| Data source 4: Population denominators
National and provincial age-and year-specific population denominators were obtained from projections of the 2012 Rwanda census data. 18 Rwanda had a population of 10 955 840 individuals in 2014 of which 1 604 444 (14.6%) were children aged <5 years. 18

| Estimation of the number and rate of SARI and influenza-associated SARI hospitalizations
To estimate the national number of SARI and influenza-associated SARI hospitalization, we used 2 different methods.

| Method 1
In Method 1, we used a four-step approach.
In Step 1.1, we estimated the national number of any respiratory hospitalization by multiplying the national number of respiratory hospitalization reported through the HMIS (data source 1) by the estimated proportion of underreporting. The proportion of underreporting was obtained by comparing the number of respiratory admissions reported through the HMIS (data source 1) from the hospitals where the retrospective record review was implemented and those recorded from the record review (data source 2).
In Step 1.2, we estimated the national number of SARI cases by multiplying the adjusted number of respiratory hospitalizations obtained in Step 1.1 by the estimated proportion of SARI cases among any respiratory hospitalization. To obtain the proportion of SARI cases among any respiratory hospitalization, we first reclassified any respiratory admission obtained from the record review as SARI if they met the SARI case definition using the clinical information collected during the review (data source 2). Subsequently, we calculated the estimated proportion of SARI cases among any respiratory hospitalization by dividing the number of reclassified SARI cases by the total number of recorded respiratory admission (SARI and non-SARI).
In Step 1.3, we estimated the national number of influenzaassociated SARI hospitalization by multiplying the number of SARI cases obtained in Step 2 by the influenza virus detected rate obtained from influenza sentinel surveillance implemented among inpatients with SARI (data source 3). For this estimation, we used annual influenza virus detection rate because the number of SARI cases enrolled did not display seasonal variations (data not shown).
In Step 1.4, we estimated the rates of SARI and influenza-associated SARI hospitalizations by dividing the estimated corresponding numbers obtained in Step 1.3 by the population at risk (data source 4).
The following equations were used to obtain the rates of SARI and influenza-associated SARI for Method 1: where Rate SARI is the rate of SARI hospitalization; HMIS is the national number of respiratory hospitalization reported through the HMIS (data source 1); Adj PropRep is the proportion of respiratory admissions reported through the HMIS (data source 1) from the hospitals where the retrospective record review was implemented and those recorded from the record review (data source 2) obtained in Step 1.1; Prop SARI is the proportion of any respiratory hospitalization obtained from the record review that met the SARI cases definition (data source 2) obtained in Step 1.2; Pop is the population at risk (data source 4); Rate Inf is the rate of influenza-associated SARI hospitalization; and Prop Inf is the detection rate of influenza virus among SARI cases tested (data source 3).

| Method 2
In Method 2, we used a three-step approach.
In Step 2.1, we estimated the SARI hospitalization rates at the hospitals where influenza sentinel surveillance was conducted using the WHO guidelines for estimating the disease burden associated with seasonal influenza. 12 First, we estimated the service population of the 6 sentinel sites by dividing the number of HMIS respiratory hospitalizations at each of the sentinel sites by the national number of respiratory hospitalizations reported through the HMIS (data source 1). This proportion was then applied to the population at risk over the study period (data source 4). Thereafter, we obtained the SARI hospitalization rates for the sentinel sites by dividing the total number of SARI hospitalizations at the sentinel sites (data source 2) by the estimated service population.
In Step 2.2, we estimated the rates of influenza-associated SARI hospitalization at the sentinel sites by multiplying the rates of SARI hospitalizations obtained in Step 2.1 by the influenza virus detected rate obtained from influenza sentinel surveillance implemented among inpatients with SARI (data source 3). We used the SARI and influenzaassociated SARI hospitalization rates at the sentinel sites as a proxy for the corresponding province as previously described. 8,9 In Step 2.3, we estimated the number of SARI and influenzaassociated SARI hospitalization by multiplying the corresponding rates (obtained in Step 2.2) by the population at risk.
The following equations were used to obtain the rates of SARI and influenza-associated SARI for Method 2: where Rate SARI is the rate of SARI hospitalization; SARI is the number of SARI cases obtained from the record review (data source 2); Pop is the population at risk obtained in Step  All analyses for Methods 1 and 2 were implemented overall and within the following age categories: <1, 1-4, 5-24, 25-44, 45-64, ≥65, <5, and ≥5 years of age. Provincial estimates were also calculated. Rates were expressed per 100 000 population. All estimates were reported as mean annual estimates over the study period.
We obtained the 95% confidence intervals (CI) using bootstrap resampling over 1000 replications for all relevant parameters included in the calculations for each method. This included (i) the age-, year-, and hospital-specific proportion of underreporting (including interhospital variability), (ii) the age-and year-specific proportion of SARI cases over total respiratory admissions, and (iii) the age-and year-specific influenza virus detection rate among SARI cases tested. The lower and upper limits of the 95% CI were the 2.5th and 97.5th percentiles of the estimated values obtained from the 1000 resampled datasets, respectively.
The statistical analysis was implemented using Stata 14.1 (StataCorp, College Station, TX, USA).

| Ethics
The influenza sentinel surveillance and the retrospective record review were deemed non-research by the Rwanda Ministry of Health and the US Centers for Disease Control and Prevention. The HMIS and census data were publicly available.

| National number and rate of SARI and influenza-associated SARI hospitalizations
The estimated number and rates of SARI and influenza-associated SARI hospitalizations were similar (overlapping CIs) between the 2 estimation approaches (Tables 1 and 2).  Figure 2). No statistically significant difference (overlapping CIs) of the SARI and influenza-associated SARI hospitalization rates was observed across Provinces (Tables 1   and 2).
T A B L E 1 Estimated mean annual numbers and rates of severe acute respiratory illness and influenza-associated severe acute respiratory illness hospitalizations, Rwanda, 2012-2014 (Method 1)

| DISCUSSION
We reported national and provincial estimates of influenza-associated SARI hospitalization in Rwanda over a 3-year period. Influenzaassociated SARI hospitalizations were substantial and were observed across age groups. However, the influenza-associated SARI hospitalization rates were highest among individuals aged <5 and ≥65 years.
Children aged <5 years accounted for 72% of the total influenzaassociated SARI hospitalizations.
Higher rates of influenza-associated respiratory hospitaliza-   24 In addition, ecological studies have suggested that influenza virus is responsible for hospitalizations and deaths also among patients presenting with circulatory illnesses or even non-respiratory and non-circulatory syndromes. 5,6,19 Third, we did not collect data on the total number of medical hospitalizations, hindering our ability to estimate the proportional contribution of SARI-associated admissions among any medical admission. Lastly, individuals that may have developed influenza-associated severe illness, but did not seek care would F I G U R E 2 Mean annual estimates of severe acute respiratory illness (SARI) and influenza-associated severe acute respiratory illness hospitalization rates (per 100 000 population) by age group, Rwanda, 2012-2014. (A) SARI hospitalization rates; (B) influenza-associated SARI hospitalization rates have been missed in our study; hence, our estimates should be considered minimum estimates.
In conclusion, we reported a substantial hospitalization burden associated with influenza virus infection especially in individuals aged <5 and ≥65 years. Our estimates were similar to those from other countries from Africa or global estimates for Africa but higher than those estimated from other regions, highlighting the heavier burden of influenza-associated severe illness on the continent. The RMoH is yet to implement a national influenza vaccination program. Should an influenza vaccination program be introduced in Rwanda, young children and the elderly may benefit most from annual influenza immunization.
No influenza vaccine is licensed for children aged <6 months, but this group may be protected through the vaccination of their mothers during pregnancy. 25,26