Burden and seasonality of medically attended influenza like illness (ILI) in Ethiopia, 2012 to 2017

Background The influenza virus spreads rapidly around the world in seasonal epidemics, resulting in significant morbidity and mortality. Influenza-related incidence data are limited in many countries in Africa despite established sentinel surveillance. This study aimed to address the information gap by estimating the burden and seasonality of medically attended influenza like illness in Ethiopia. Method Influenza sentinel surveillance data collected from 3 influenza like illness (ILI) and 5 Severe Acute Respiratory Illness (SARI) sites from 2012 to 2017 was used for analysis. Descriptive statistics were applied for simple analysis. The proportion of medically attended influenza positive cases and incidence rate of ILI was determined using total admitted patients and catchment area population. Seasonality was estimated based on weekly trend of ILI and predicted threshold was done by applying the “Moving Epidemic Method (MEM)”. Result A total of 5715 medically attended influenza suspected patients who fulfills ILI and SARI case definition (77% ILI and 23% SARI) was enrolled. Laboratory confirmed influenza virus (influenza positive case) among ILI and SARI suspected case was 25% (1130/4426) and 3% (36/1289). Of which, 65% were influenza type A. The predominantly circulating influenza subtype were seasonal influenza A(H3N2) (n = 455, 60%) and Influenza A(H1N1)pdm09 (n = 293, 38.81%). The estimated mean annual influenza positive case proportion and ILI incidence rate was 160.04 and 52.48 per 100,000 population. The Incidence rate of ILI was higher in the age group of 15–44 years of age [‘Incidence rate (R) = 254.6 per 100,000 population’, 95% CI; 173.65, 335.55] and 5–14 years of age [R = 49.5, CI 95%; 31.47, 130.43]. The seasonality of influenza has two peak seasons; in a period from October–December and from April–June. Conclusion Significant morbidity of influenza like illness was observed with two peak seasons of the year and seasonal influenza A (H3N2) remains the predominantly circulating influenza subtype. Further study need to be considered to identify potential risks and improving the surveillance system to continue early detection and monitoring of circulating influenza virus in the country has paramount importance.

Seasonal influenza typically occurs every year in the late fall or winter in temperate regions [11] while the seasonality is less clearly defined in tropical and subtropical regions [12]. Seasonality patterns of influenza in eastern Africa including Ethiopia, have not been clearly established [13]. However, some evidences from different studies indicated that Influenza A (H1N1) pdm2009, Seasonal Influenza A (H3N2) and Influenza B are circulating in different countries of sub-Saharan Africa [14][15][16].
Estimates of the national burden of influenza associated hospitalization across age groups are severely limited in Africa. It was described only in few countries for influenza and influenza-related incidence data remain inadequate [3,17,18]. World Health Organization (WHO) has highlighted the need for influenza disease burden estimates especially from low-and middleincome countries. 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 [19]. This study aimed to address this gap by estimating the burden and seasonality of influenza like illness using sentinel surveillance data collected from 2012 to 2017 in Ethiopia.

Data source
The Influenza like Illness (ILI) and Sever Acute Respiratory Illness (SARI) surveillance data collected in a period from 2012 to 2017 was applied to estimate proportion of laboratory confirmed influenza positive ILI cases (medically-attended) among patients presented to the sentinel sites (health facility) seeking medical care for any kind of illness and influenza positive ILI cases among the catchment population (incidence rate) in Ethiopia. The ILI program is a health center-based sentinel surveillance system that monitors children and adults presenting with sudden onset of fever > 38°C and cough or sore throat. The three ILI sentinel sites (Shiromeda health center which was established in 2008, Kolfe health center and Akaki health center established in 2010) are located in Addis Ababa. The SARI program is a hospital-based sentinel surveillance system that monitors children and adults hospitalized with pneumonia across five hospitals in Ethiopia. The SARI sentinel sites are located in four regions which are Yekatit-12 hospital (Addis Ababa) established in 2008, Mekele hospital (Tigray region, northern Ethiopia), Felege-hiwot hospital (Amahara region, North weastern Ethiopia), Adama hospital (Oromia region, Central Ethiopia) and Adare hospital (Southern region of Ethiopia) which was established in 2012; Fig. 1, below.
Case definitions and data collection ILI was defined as any person presenting with sudden onset of fever ≥38°C (axillary measured) AND cough or sore throat in the absence of other diagnosis with in the last 7 days [17,18]. SARI was defined as any person > 5 years old presenting with symptoms of acute lower respiratory infection with sudden onset of fever ≥38°C, AND cough or sore throat, AND shortness of breath or difficulty breathing with in the last 7 days, AND requiring hospitalization [17,18] or any child ≤5 years that fulfilled the case definition of pneumonia and severe pneumonia as per the Integrated Management of Childhood Illnesses protocol [19]. Therefore, the incidence rate of ILI was defined as the number of new ILI cases (fulfilling ILI case definition) per total population of the catchment area per year.

Epidemiological and laboratory data
Epidemiological data and specimens were collected from patients who presented at a sentinel site and fulfill the case definition for ILI and SARI. Epidemiological data including patients' demographic information was recorded on a standard case based reporting format and weekly aggregated data format. The case based reporting format has every detailed epidemiological and laboratory based information for those sampled cases. Whereas the aggregated data form was collected a weekly bases with a components of total admission, number of new case, number of sampled, number of tested, number of positive and proportion of both ILI and SARI case by specific age group weekly aggregate collected from each sentinel sites. Overall data of Influenza surveillance from 2012 to 2017 was obtained from Ethiopian Public Health Institute with an official request.
Each sentinel site has a protocol to collect throat swab specimens as part of sentinel surveillance system. The sentinel sites focal persons obtained unwritten (Verbal) consent from each patient prior to collection of throat swab specimen. Since it's a routine surveillance activity under the health care delivery system only verbal consent from the patients are the only requirement. Throat swab samples were systematically collected from out-patients of all ages who fulfilled the case definition for ILI per week at sentinel surveillance sites. Similarly throat swab samples were also collected from patients who fulfilled the SARI case definition and admitted to designated SARI sentinel surveillance sites (hospital). Specimen were collected within 7 days after the first onset of symptoms for ILI and SARI. Throat swabs samples was placed in viral transport media (VTM) and stored at 4°C until transported to the National Influenza Laboratory (NIL) at Ethiopian Public Health Institute (EPHI). Shipment of throat swab specimens in viral transport media to the NIL at EPHI was conducted within 72 h of collection using a cold chain system. Viral RNA from throat swabs were extracted and subjected to real-time PCR amplification with parameters set for influenza testing, according to Center for disease Control and Prevention (CDC) protocol using reagents obtained from Influenza Reagent Resources (IRR). Further sub-typing and characterization of influenza A-positive specimens were carried out using the CDC real-time reverse transcription PCR protocol [20].

Population denominators
The total admitted patient in each sentinel sites for health care service and the population in the lower administrative level where the sentinel sites are located (district/woreda population) was applied as a denominator to calculate the medically attended ILI positive case proportion and incidence rate. The total admitted population was obtained from the weekly aggregated data collected from the registration log book for all patients seeking any medical service in each health facilities and used as a denominator population to calculate the influenza positive case proportion of ILI and SARI associated influenza patients. Age-and year-specific total population denominators were obtained from projections of the 2007 census data for Ethiopia reported by Central Statistical Authority [CSA, 2007]. The midyear population size were applied for calculation of the Incidence All estimates were obtained overall and within the age categories of: < 1 years, 1-4 years, 5-14 years, 15-44 years and > 44 years of age. This age category has different from WHO age category of 0 to < 2 years, 2 to < 5 years, 5 to < 15 years, 15 to < 50 years, 50 to < 65. The variation in age category applied in this manuscript was done by considering the availability of

Statistical analysis
The overall data management and analysis were done using Excel v-13, SPSS V-23 and RStudio (version 3.4.3) applications. The case based surveillance data and weekly aggregated data (aggregated by age specific total admission, number of new cases, number of tested and sampled per week) from 2012 to 2017 were entered and cleaned using excel v-13. The incomplete data set (missing values) and the outbreak investigation results in 2016 were excluded and only the complete data sets recorded in the routine sentinel surveillance system were included to maintain the consistency of surveillance data.
Descriptive statistics were carried out using SPSS V-23, for the basic demographic characteristics for calculating the frequency, percentage, mean, median and range. Cross tabulation also applied to identify the association between the case classification (ILI and SARI) of influenza positive cases by the determinant factors like sex and age; and also for the influenza types and subtypes.
The estimated medically attended influenza positive ILI case proportion was calculated by taking the influenza (ILI and SARI) positive cases divided by the total admission (cumulative from eight sentinel sites) extracted from the weekly aggregated data. The estimated medically attended influenza like illness incidence rate was calculated by taking the influenza positive cases divided by the cumulative population of the lower administrative level population size where each sentinel sites were located and multiplied by a constant of 100,000. The Incidence rate was calculated for each year of the surveillance period and the cumulative Incidence rate was calculated by taking the cumulative influenza positive cases (2012-2017) divided by the mid-year population. Incidence rates were expressed per 100,000 population.
Weekly trend of seasonal influenza with in the 7 year period was applied for seasonality and predicted threshold was done by applying the "Moving Epidemic Method (MEM)" using R-Studio version 3.4.3. As initial step, the weekly trend of medically attended influenza positive cases were done using excel v-13. Then the weekly Incidence rate of medically attended influenza like illness was for each year in the surveillance period (2012-2017) was applied. Since Ethiopia is located in the Northern hemisphere (4 degree); the Incidence rate on week 40 was used as a starting period and week 20 as the end period. The seasonality and MEM model was extracted from the analysis result on the R-Studio. This includes the epidemic threshold, high and very high threshold, starting and ending season of the epidemics including the goodness and intensity parameters.
The main outcome of this manuscript includes estimation of the burden, seasonality and predicted threshold of medically attended influenza positive ILI case and can be described as follows: 1. Burden of influenza positive ILI cases (Medically attended) was expressed in terms of proportion and Incidence rate as recommended by WHO-guideline. Proportion of influenza positive ILI cases = (Laboratory confirmed influenza virus positive ILI cases divided by total number of patients presented to the sentinel sites (health facility) seeking medical care for any kind of illness within a specified period)*100 Incidence of ILI = (Laboratory confirmed influenza virus positive ILI cases divided by total population in the catchment area)* 100,000. 2. Seasonality was described as the peak season (period) of the year where the number of laboratory confirmed influenza positive cases of ILI were relatively high from the weekly trend of medically attended ILI cases. 3. Predicted threshold: this can be described as the expected incidence rate of laboratory confirmed influenza positive cases of ILI with in the specified time of the year. Epidemic threshold: defined as the predicted incidence rate of laboratory confirmed influenza positive case beyond the expected level and considered as an epidemics.
This analysis was done using data collected as part of routine surveillance activities, and as such, ethical approval was deemed not necessary from the Ethiopian Public Health Institute's Scientific and Ethical Review Office (SERO). The schematic analysis plan is presented in Fig. 2 below.

Result
Demographic characteristics and distribution of medically attended influenza (ILI and SARI) cases enrolled A total of 5715 medically attended influenza suspected patients who fulfills ILI and SARI case definition was enrolled in eight influenza sentinel surveillance sites located in four regions of Ethiopia from 2012 to 2017. The majority of the cases enrolled in the sentinel surveillance system was outpatients with ILI (N = 4426, 77%) from 3 sentinel sites and the rest were inpatients presented with SARI (N = 1289, 23%) from 5 sentinel sites. Higher number of ILI cases was from Shiromeda Health Center (n = 2268, 51%) followed by Kolfe Health Center (n = 1843, 42%); all ILI sites are located in Addis Ababa. Forty percent (n = 2273) of the cases were under 15-44 years age group, followed by age group of 5-14 year (n = 1490, 26%) and higher number of cases were Females (n = 3274, 57%). Of the total 5715 medically attended influenza suspected cases 20% (n = 1166) was found positive for influenza and 65% (n = 755) of the positive cases were found Influenza A, Table 1 above.

Distribution of medically attended ILI and SARI cases by influenza type and sub-type
Of the total 1166 medically attended influenza positive cases (patients) 65% were influenza type A and 35% were influenza type B. Of the influenza type A cases the predominantly circulating influenza subtype (n = 455, 60%) were Seasonal Influenza A(H3N2) and (n = 293, 39%) were Influenza A(H1N1)pdm09. Among 4444 outpatient cases presented to three sentinel sites with Influenza-like-illness (ILI), 25% (n = 1130) were tested positive for influenza. Of which, the majority of patients who tested positive for influenza were influenza type A (n = 724, 64%), and of those, the majority were seasonal influenza A (H3N2) (n = 433, 60%). On the other hand, among 1271 (22% of all cases) in-patients presented to five SARI sentinel sites with Sever Acute Respiratory 36 (3%) were tested positive for influenza. The overall percent positive for medically attended influenza cases were 20% (1166/5715). Off the total influenza positive cases 27% (n = 399/1166) were within the age group of 5-14 years of age followed by 15-44 age groups (n = 573, 25%). One percent (n = 78) of influenza positive cases were missing patient age (  Table 4 below.

Seasonality and threshold of medically attended influenza like illness (ILI)
The seasonality of medically attended influenza like illness had two peak seasons (Figs. 5

Discussion
Presenting country-level data from Ethiopia is incredibly useful, especially in a region where country-level data are not well described or published. The frequency distribution, influenza positive case proportion, incidence rate and seasonality of medically attended influenza (ILI and SARI) based on sentinel surveillance data (epidemiological and laboratory data) collected from eight sentinel sites over 6 years period (2012-2017) were presented. A total of 5715 medically attended influenza suspected patients who fulfills ILI and SARI case definition (77% ILI and 23% SARI) were enrolled in eight influenza sentinel surveillance sites located in four regions of Ethiopia from 2012 to 2017. The variation in the number of influenza suspected cases among sentinel sites and the case classification might be due to the difference in patient flow in the sites and effectiveness of the surveillance activities in each sentinel sites.
Over the 6 years period, the overall percent positive for medically attended influenza cases were 20% (1166/ 5715). The percent positive for ILI and SARI case were 25% (1130/4426) and 3% (36/1289), respectively. This finding is relatively higher than 16.2% posetivity rate in most of African countries and 18% positivity rate in eastern African countries as reported by WHO during 2016 [21]. Whereas, this shows higher compared to the influenza positivity rate from ILI samples in Niger (12%) and Gabon (11%), and lower compared to 23.3% posetivity rate in Madagascar [22][23][24]. The influenza positivity rate for SARI samples was 2.83% which was low compared to findings in South Africa and China which reported 8 and 6% respectively [25,26]. In addition, the presented finding was lower in comparison with 16.5 and 20.2% influenza virus detection rate among SARI cases reported in Eastern Mediterranean Region and Vietnam [27,28]. The variation if the findings might be due to the difference in accuracy of detection of cases at the sentinel sites, the timing of sampling, specimen quality and handling approach, effectiveness of the diagnostic testing in different settings. In contrast, it might be due to the true variation in circulating influenza viruses in different geographical areas as reported by different countries.
Off the total 1166 medically attended influenza positive cases 65% were influenza type A and 35% were influenza type B. Off the influenza type A cases the predominantly circulating influenza subtype 60% (n = 455) were Seasonal Influenza A(H3N2) and 39% (n = 293) were Influenza A(H1N1)pdm09. This finding is closely comparable with a finding reported by WHO in which, 56% of influenza A(H3N2) subtype and 32% A B Fig. 4 Medically attended influenza positive case among the total admission (a) and incidence rate (b) of ILI per 100, 000 population by age group, 2012-2017, Ethiopia A(H1N1) pdm subtype in most African countries [21]. Where us the present finding has different from 117 (38.0%) influenza A(H3N2) and 58 (19%) influenza A(H1N1)pdm09, reported in Madagascar [24].
In the present finding, the seasonality of influenza has two peak seasons in a period from October-December (week 39-51) and from April-June (week 7-15 of the year). Furthermore, the seasonality of influenza using cumulative Incidence rate with in the lower administrative level catchment population from 2012 to 2017 shows that the epidemic season mean start lies on week 42 (October) and the mean end lies week 2 (January) of the year. The highest threshold was recorded by the year2016/16 (R = 8.2, r = 032) followed by 2014/15 (R = 7.9, r = 0.37) and 2012/13 (R = 7.5, r = 0.43)). The average epidemic lengths might last for 11-12 weeks. Based on the applied model the weekly epidemic threshold of influenza ranges from 2.5 (minimum threshold) to 7.66 (Very high threshold) per 100,000 population in the catchment area [Mathew correlation coefficient "r" of 3.4 with 0.42 sensitivity rate and 0.67 positive predictive value]. This finding agreed with the WHO report in which the peak of transmission was observed from epi weeks 7 to 9 of 2016 and 2015, which coincided with the northern hemisphere winter and influenza type A was predominant. It also reported that there was a second increase in influenza activity from epi weeks 48 to 51 of 2016 predominated by influenza type B [21]. The present finding also aligns with the finding by Hirve et al. (2016) who reported a seasonality pattern similar to the northern hemisphere temperate regions start of the main influenza season between October and December [29].
The mean annual incidence rate of influenza (52.48 per 100,000) in the present finding is relatively higher in comparable to the finding reported in Kenya (2012-2014) with an estimated annual Incidence rate of hospitalized influenza-associated SARI was 21 per 100,000 population and the mean annual Incidence rate of non-hospitalized influenza-associated SARI was 81.7 (95% CI 74.1-89.9) per 100,000 persons [30]. Even if the present finding doesn't entail on SARI, association of the non-hospitalized influenza associated SARI with the ILI indicates that the finding in the present finding is relatively lower. In addition, The mean annual incidence rate of medically attended influenza cases showed variation in comparison to other countries report [24][25][26][27][28][29][30]. The variation in the reported incidence rate among the age group might be due to the actual difference in the incidence of influenza in different countries, the quality of surveillance system, the health care seeking behavior of the community and also it might be due to the different approach/method of analysis for calculating the incidence rate.

Limitation of the study
The presented study has limitations that warrant discussion. The very low health care seeking behavior of the community for influenza infection might affect the findings of the estimated Incidence rate of influenza in Ethiopia. The adjustment factor for this low health seeking behavior and also to the national level were not applied in the present finding. The presented findings has only based on the surveillance data from eight sentinel sites and it does not provide has limitations in providing the burden at national level. The surveillance system has also enrolled only few SARI patients whereas the influenza positive case proportion of SARI as compared with total admission cases was high in all hospitals. The surveillance system is also capturing only influenza. Hence the estimates in this study should be considered minimum estimates.

Conclusion
Influenza like illness has significant morbidity in Ethiopia, and that it is disproportionally distributed by age. The majority of medically attended influenza positive cases were influenza type A and higher number of it were seasonal influenza A(H3N2) and Influenza-A (H1N1) pdm2009. The incidence rate of influenza found to be higher in the age group of 5-14 years and 15-44 years. Incidence rate of Seasonal influenza has two peak seasons in a period from October-December and from April-June. The rate of influenza is significant, warranting a response from government (public health agencies) to prevent, detect, and treat influenza in the name of national health security. It could be useful to policy makers to consider influenza seasonality when determining timing, vaccine composition, and target age groups for influenza vaccination campaigns. Improving the surveillance system could help further define populations at risk for influenza.

Funding
The ILI and SARI sentinel surveillance work in Ethiopia was conducted through the financial support from Centers for Disease Control and Prevention (CDC) through cooperative agreement entitled "Sustaining Influenza Surveillance Networks and Response to Seasonal and Pandemic Influenza by National Health Authorities outside the United States" (CFDA 93.283). Reference: CDC-RFA-IP11-110302CNT12-CoAg: U51IP000600-01-05. CDC and WHO was involved in provision of training and technical support in the surveillance system. However, the funding agency was not involved in the design of the study and collection, analysis, and interpretation of data and in writing this manuscript.