Human influenza infections are a major cause of morbidity and mortality worldwide [1–3]. Although risk factor data from tropical climates are limited, young children (<5 years), pregnant women, the elderly, and persons with underlying medical conditions have been shown to be at increased risk of severe disease [1, 4]. A recent study estimated that there were 20 million cases of influenza associated with pneumonia; 1 million cases of influenza associated with severe pneumonia; and 28000–111500 deaths associated with influenza among children aged <5 years globally in 2008, with 99% of these deaths occurring in the developing world [5]. A summary of existing direct estimates of influenza disease burden in tropical and developing countries is needed to validate global modeling efforts that suggest a disproportionate burden in these countries.

In Kenya, influenza surveillance was established partly in response to the global emerging threat of avian influenza A(H5N1)[6, 7]. As is the case with other tropical and sub-tropical countries, influenza viruses circulate in Kenya for most of the year [7–9] and morbidity (hospitalization and outpatient) burden of influenza have only recently been described [10–18].

An improved understanding of disease burden in Kenya relating to morbidity, mortality, and economic losses is needed to support decisions involving the allocation of limited resources toward influenza control programs. The Kenyan Ministry of Health (KMoH) has released its first ever influenza vaccination policy [19] and this has necessitated the publication of an overview of the burden of influenza in Kenya to inform initial vaccination pilot activities.

In this article, we review existing data on the influenza disease burden in Kenya using data collected until December 2013. We summarize published data describing the health burden of human influenza collected through population-based influenza surveillance systems in Kenya. We also discuss the various disease burden estimation methods used and provide suggestions for future research strategies that will help to generate additional data needed to inform influenza control strategies.

We have provided a comprehensive summary of available data on disease burden of influenza in Kenya and we show that influenza is an important cause of respiratory infection-associated morbidity, especially among younger children under the age of five years. Indeed, both adjusted and unadjusted incidence rates of hospitalization with influenza [10, 17, 18], and outpatient visits[18] were higher than those that have been reported in United States and European countries during similar time periods [21–32].

We also note that the published literature on the burden of influenza in Kenya is limited but expanding. Eight of the ten papers that we reviewed were published within the last three years (2012–2014)–including two studies that published data on the post- pandemic A(H1N1) period. This could be attributed to the interest generated by the threat of avian and pandemic influenza.

Our review showed that there were an estimated 2.7–4.7 cases of influenza per 1,000 among children <5 years who were hospitalized with severe acute respiratory illness (SARI). These were 7–10 times higher compared to those in persons aged ≥5 years. A study that estimated disease burden among hospitalized children <6 months in Western Kenya reported that there were 5.7 cases of influenza per 1,000 [18]. This is consistent with data from several other countries [22, 23, 33], and shows a considerable burden of disease in young infants for whom influenza vaccination is not recommended, and also highlights the rationale for targeting pregnant mothers for influenza vaccination [34]. Whereas pregnant mothers have been shown to be at increased risk of complications associated with influenza [35], vaccinating them may not only be beneficial to them but could also offer protection to their young infants—for whom no influenza vaccine is currently licensed—through breastfeeding and trans-placental antibody transfer [36, 37].

The incidence rates of hospitalization with influenza among children <5 years reported at the two refugee sites (Kakuma and Dadaab)–without adjustments for eligible cases who were not tested—were higher than rates reported elsewhere in Kenya. This could be due to the unique challenges experienced by the populations in refugee settings, such as population density, which could make them more vulnerable to exposure to respiratory infections [38, 39]. While the incidence rates of hospitalization with influenza are similar to those reported in South Africa [40], Asia [41–46], Latin America [47], some of the reported rates were up to seven times higher than rates reported in the United States [21–26] and Europe [27, 28]. The incidence rates of hospitalization with influenza children <6 months in Kenya, for example, were 2–3 times higher than rates reported in the United States [23].

The adjusted incidence of medically-attended (outpatient + inpatient) influenza among children <5 years ranged from 21.8 to 58.0, and 4.3 to 26.0 per 1,000 among persons aged ≥5 years in different studies. These rates were similar to rates reported in Asia [48], but up to 2–4 times higher than annual estimates reported in Europe [29–31], and up to 2–8 times higher than rates reported in the United States [32].

The incidence of non-medically attended severe ARI associated with influenza suggested a burden of disease that was similar to the medically-attended burden. As reported in the health utilization survey conducted in Western Kenya, 52% of children <5 years and 66% of persons ≥5 years who reported to have had pneumonia did not seek care at a hospital [49]. The similarity between the medically-attended and non-medically attended incidence not only underscores the fact that there is a considerable burden of non-medically attended influenza, but also highlights the low levels of health-care seeking for respiratory illness in Kenya [49]. These findings also suggest that surveillance limited to the health care setting will not capture the entire burden of influenza severe respiratory illness in contexts such as Kenya.

The studies reviewed included various adjustments for patients who met the case definitions but were not tested for influenza [10–12, 18]; for those who sought health-care at a facility other than the one used for estimating the incidence rates [13, 14]; and for asymptomatic detection of influenza among controls [13, 14]. The first two adjustments would serve to increase the crude incidence to account for persons who met the case definition and were not tested for influenza or those who did not seek care; while the latter would drive the rate downwards by only accounting for the cases for which the virus was the likely cause disease. Other than the case definition for ILI developed by the World Health Organization (WHO) which was more commonly applied across the different studies [50], the case definitions used for SARI, ALRI and ARI also substantially varied across the studies reviewed. In order to facilitate disease burden comparisons over time, it would have been helpful if researchers also presented their data using standard case definitions as recommended by WHO[51]; and unadjusted rates in addition to those where adjustments were applied. These standardizations, in addition to presenting data in age groups that may be aggregated to WHO recommended age categories [<2, 2–4, 5–14, 15–49, 50–64, and ≥65 years] [51] would help to facilitate comparisons across studies and across countries [52].

All the articles reviewed utilized data generated from well-defined catchment areas managed by either the KEMRI and CDC, or the KEMRI and Wellcome Trust research collaboration; indeed a majority of the articles reviewed included KEMRI and CDC co-authors who are also authors on this paper. Additionally, all the papers that we reviewed included in their analysis data that were collected year-round and a majority of them had multiple years of data used in estimating incidence rates. This consideration is important because–other than considering that influenza circulates in Kenya year-round [7]–it minimizes the risk of overestimating the disease burden by only sampling during epidemic periods.

Six of the ten studies reviewed utilized mid-year population denominators, derived from HDSS or National census data [12], for the estimation of incidence rates. As opposed to using denominators derived from individual follow-up (person-time) in population-based surveillance systems, denominators based on mid-year population could potentially underestimate the incidence rates as they may not accurately reflect the actual population dynamics relating to births, migrations and deaths, especially if smaller populations are involved [53]. However, using either of these two denominator types to estimate incidence rates would normally yield nearly identical results for large populations. Taken in the context of the resources required to set-up and run a population-based surveillance system, denominators derived from mid-year population numbers may be useful for disease burden estimation in Kenya for the foreseeable future.

Only one study reported rates among children <6 months, and a few reported data on those aged ≥50 years [11, 12] which is in part explained by the lower health-seeking behavior among older persons in Kenya [49], and perhaps also by a diminished likelihood that older patients will report the fever required to meet the WHO SARI case definition. Understanding the disease burden, especially in the high risk groups which include pregnant women, and people with underlying medical conditions; as well as understanding the socio-economic (direct and in-direct) burden of influenza in Kenya would be helpful to public-health and influenza control programs and understanding the impact of influenza. For example a recent study conducted in Western Kenya showed a substantial burden of influenza (3-times higher) among HIV-infected adults aged ≥18 compared to their HIV-negative counterparts [13]. Another study in South Africa reported 4–8 times greater incidence of acute lower respiratory tract infection (LRTI) with influenza among HIV-infected compared to HIV-uninfected persons [40].

Our study was subject to limitations. First, we may have missed some articles as we limited our review to only published data searched the PubMed and EMBASE databases. However, we believe that the likelihood of finding additional data relevant to our study in other databases is very low. Second, most of the published data summarized in our review included data from the 2009 pandemic influenza period and may have served to overestimate the seasonal influenza disease burden. Third, the reviewed papers were limited to respiratory surveillance only. For some populations (particularly young infants) presentation may be fever without respiratory symptoms. As such, the true burden among children may have been underestimated. Fourth, the clinical threshold to hospitalize in Kenya may not be comparable to US or Europe and therefore hospitalization rates should be interpreted with caution when making these comparisons. Fifth, most of the studies presented data on incidence of influenza without presenting either the age-specific denominators or age-specific numbers of cases. Taken together with the fact that there were varied case definitions and adjustments applied, this made it difficult for us to calculate meta-analytic rates of influenza disease burden in Kenya. Lastly, while not a direct limitation of our methods, the absence of data on influenza mortality remains a gap that needs to be addressed in order to inform influenza vaccine policy.

In conclusion, our literature review provides a comprehensive summary of available data on the disease burden of influenza in Kenya over the past 8 years, and shows a substantial medically- and non-medically attended disease burden among children aged <5 years. Additional research gaps identified in the review include the lack of influenza mortality and socio-economic disease burden data. While these additional data would be very helpful to policy makers and other stakeholders to inform prevention and treatment policies, the current data in Kenya indicate an important burden of influenza in young children that might be reduced with a targeted vaccination program including children and pregnant women. However, any decision about influenza vaccination must look at its burden relative to other respiratory pathogens such as respiratory syncytial virus–when a vaccine becomes available—and even non-respiratory vaccine preventable diseases.