Severe acute respiratory illness surveillance for influenza in Kenya: Patient characteristics and lessons learnt

Abstract Background We describe the epidemiology and clinical features of Kenyan patients hospitalized with laboratory‐confirmed influenza compared with those testing negative and discuss the potential contribution of severe acute respiratory illness (SARI) surveillance in monitoring a broader range of respiratory pathogens. Methods We described demographic and clinical characteristics of SARI cases among children (<18 years) and adults, separately. We compared disease severity (clinical features and treatment) of hospitalized influenza positive versus negative cases and explored independent predictors of death among SARI cases using a multivariable logistic regression model. Results From January 2014 to December 2018, 11,166 persons were hospitalized with SARI and overall positivity for influenza was ~10%. There were 10,742 (96%) children (<18 years)—median age of 1 year, interquartile range (IQR = 6 months, 2 years). Only 424 (4%) of the SARI cases were adults (≥18 years), with median age of 38 years (IQR 28 years, 52 years). There was no difference in disease severity comparing influenza positive and negative cases among children. Children hospitalized with SARI who had an underlying illness had greater odds of in‐hospital death compared with those without (adjusted odds ratio 2.11 95% CI 1.09–4.07). No further analysis was done among adults due to the small sample size. Conclusion Kenya's sentinel surveillance for SARI mainly captures data on younger children. Hospital‐based platforms designed to monitor influenza viruses and associated disease burden may be adapted and expanded to other respiratory viruses to inform public health interventions. Efforts should be made to capture adults as part of routine respiratory surveillance.


| INTRODUCTION
Globally, each year, influenza epidemics result in 3 to 5 million cases of severe illness and approximately 290,000 to 650,000 influenzarelated deaths. 1 A recent modeling study estimated that the highest burden of annual influenza-related deaths was in Sub-Saharan Africa, South East Asia, and Western Pacific regions. 2 The rate of severe disease is highest in young children <2 years, adults ≥65 years, pregnant women, and individuals with underlying chronic medical conditions. 3,4 In temperate areas, there are distinct influenza seasons during the cold months, but in the tropics, influenza circulates year-round, with multiple, often erratic, periods of increased influenza activity, 5 a pattern also seen in Kenya. 6 Due to increased globalization, respiratory disease pandemics are likely to spread widely fast, as seen currently with the COVID-19 pandemic. 7 In 2006, in response to the pandemic threat of avian influenza A (H5N1), influenza surveillance in Sub-Saharan Africa was expanded. 8 Since then, data collected through clinical and virologic surveillance in Africa have contributed to the description of influenza seasonality, characterized genetic make-up of circulating viruses, and provided viruses for vaccine production. 9 In Kenya, influenza sentinel surveillance has been ongoing since 2007 (2008) and over the years has evolved to focus exclusively on hospitalized cases. This paper aims to describe the epidemiology and clinical features of patients hospitalized with influenza in Kenya and highlight the importance of yearround surveillance for severe acute respiratory infections (SARI), especially in tropical countries.

| Surveillance procedures
Trained surveillance officers at each surveillance site identified patients with any signs or symptoms of acute respiratory illness by review of inpatient ward admission registers (new admissions on Monday reviewing those admitted during the weekend and Tuesday through Friday to capture new admissions). The surveillance officers approached all potential patients to assess eligibility, specifically if the patient met the case definition for SARI defined as history of fever (or measured fever of ≥38C ) and cough, with onset of symptoms within 10 days prior to hospitalization.
Patients who met the SARI surveillance case definition were approached for verbal consent, and once consent had been obtained, the surveillance officers assigned a unique number to the patient and interviewed either the patient or guardian using an electronic structured questionnaire stored on a password-protected netbook. Data were collected on demographic and clinical characteristics of patients, including underlying conditions and medical history. A follow-up exit questionnaire was completed for each patient by abstracting data on clinical outcomes of hospitalization from medical charts upon discharge, death, or transfer. All data were stored on a secure central server maintained by the Kenya Medical and Research Institute (KEMRI). No personal identifiers were stored in the data base.

| Specimen processing
An aliquot of each specimen was tested by real-time reverse transcription polymerase chain reaction (rtRT-PCR) for influenza virus types A and B at the NIC, using protocol, primers, and probes provided by the Influenza Division, CDC-Atlanta, USA, as previously described. 10 Briefly, total ribonucleic acid (RNA) was extracted from 140 mL aliquots of each specimen using a QIAamp viral RNA mini kit (Qiagen GmbH, Germany) according to the manufacturer's instructions. One-step rtRT-PCR was carried out using the AgPath kit (Applied Biosystems, California, USA). Values with a cyclic threshold (C T ) reading <40 were recorded as positive. All specimens positive for influenza type A were subtyped for H3 and pH1N1 using rtRT-PCR.
Specimens that were positive for influenza A virus by rtRT-PCR but failed to subtype were sent to the WHO Influenza Collaborating Center at CDC-Atlanta, GA, USA, for further antigenic characterization. Because children and adults would likely differ in risk factors and severity, the analysis was done separately for children (<18 years) and adults. We used proportions to describe the demographic and clinical characteristics of SARI cases separately for adults and children. We compared influenza positive cases with those testing influenza negative, for children only due to limited sample size among adults, using Chi-square test statistic (or Fisher's exact test if applicable) for various demographic, clinical characteristics, and outcomes. Children with no recorded laboratory result were excluded from this analysis. We assessed the severity of disease among children by comparing clinical and treatment characteristics (tachypnea, hypoxia, high fever, intravenous fluids and blood transfusion treatment, admission to intensive care unit [ICU], and death) among influenza positive cases and those testing negative. We used graphs to display trends of influenza positivity for hospitalized children with SARI who were tested for influenza in the 5-year period and showed the types and subtypes detected. Those who had absconded care or refused treatment were excluded as their outcome status could not be confirmed. We then pre-selected variables that were potential confounders (age, sex, length of time from illness onset to hospitalization, and any chronic illness) or with a p value of <0.2 to include in a multivariable logistic regression model to assess predictors of death. Variables that had a p value of <0.05 in the final model were considered statistically significant.

| Ethical considerations
Data from this surveillance were regarded by the Kenya MOH as a routine public health activity and received a non-research project determination by CDC and thus did not require an ethical review. Verbal consent was obtained from all participants before administration of questionnaires and collection of specimens. For children, verbal consent was obtained from the parent/guardian and assent for children ≥7 years. (53.2%) were aged 6-23 months, and 5979 (55.6%) were males. Hospitalizations associated with SARI among adults mostly represented F I G U R E 1 Sentinel surveillance sites for severe acute respiratory illness (SARI) in Kenya people 18-64 years (85%), and 61.8% of all adults hospitalized were males. The median age among adults was 38 years (IQR = 28 years, 52 years). Underlying medical conditions were present among 2142 (19.9%) children and 214 (50.4%) among adult SARI cases. The overall influenza positivity among samples tested from two thirds of children and adults was comparable (10.0% and 11.7%, respectively) ( Table 1).
Because the number of adult cases was limited, the rest of the analysis was focused only on children from this point forward.

| Comparison of children with SARI among those with and those without lab-confirmed influenza
When comparing children with influenza and those without, there were significant differences observed by age group (p < 0.001); the <6 months age group was underrepresented among influenza positive cases compared with influenza negative (9.5% vs. 21.1%). Nonetheless, most clinical outcomes and severity were similar, with a small difference in the time from disease onset to hospitalization where influenza negative cases seem to be hospitalized earlier in the course of the disease compared with influenza positive cases (68.8% vs. 63.3%, respectively; p value <0.001), and a lower frequency of tachypnea was reported among the influenza positive cases (36.0% vs. 43.8%; p < 0.001) ( Table 2).
Overall, the median influenza positivity rate of SARI samples tested for the entire period was 7% (IQR = 3.6%-13.9%), with the highest positivity rate of 23% detected in July 2018. There was not any clear pattern of seasonality observed ( Figure 2). Overall, highest percentage of influenza positivity was in those aged 5-17 years (16%; 95% CI 12.6, 19.5) and 2-4 years (15%; 95% CI 13.0, 16.6). The temporal distribution of the types and subtypes of influenza viruses among children that were detected during the study period is shown in (supporting information Figure S1).

| Comparison of case fatality proportion (CFP) among children with and those without lab-confirmed influenza
Overall, there was no significant difference in the CFP between hospitalized children who tested positive for influenza (3.3%, 95% CI 1.9, 4.5) and those who tested negative (3.6%, 95% CI 3.2, 4.2). Young children aged less than 6 months who were influenza positive had a higher but not statistically significant CFP of 8.8% (95% CI 2.1, 15.6) compared with 4.7% (95% CI 4.  Table S1).   Among the hospitalized adults, 15% were older adults aged ≥65 years, an age group recognized as being at increased risk for influenza complications and death. 11 Among children with a reported underlying illness, in-hospital death was twice as likely to occur compared with those who did not have any reported. Nonetheless, the data do not show a particularly high risk of death among children with influenza compared with those without. Overall, 10% of children hospitalized with SARI had laboratory-confirmed influenza, with the peak percentage positivity among school age children (aged 5-17 years). SARI surveillance may need to encompass monitoring of other respiratory pathogens that are responsible for healthcare utilization and disease burden to optimize resources.
The surveillance system in Kenya mainly captured children aged <5 years, representing 94% of all SARI cases. In public health facilities countrywide, hospital user fees for children aged <5 years are subsidized by the government, and as a result, young children may be overrepresented in this surveillance due to differentiated care seeking. 12 Adults may be more likely to delay healthcare seeking, 13  South Africa (3%) 20 and Egypt (2%). 21 The overall CFP in influenza- groups not easy to reach as part of the healthcare system. 28,29 The burden of healthcare utilization associated with school aged children may need to be taken into account as the Kenya government considers recommendations for the use of influenza vaccines.
Our study had some limitations. We did not have information on the total number of hospital admissions to determine the burden of influenza and SARI hospitalizations in the context of overall hospital admissions. Estimates of disease burden using hospital data in this study should be interpreted cautiously, because in Kenya, there is limited access to care due to economic and logistic reasons. Previously, studies have shown that in countrywide, only 20% of influenzaassociated SARI cases are hospitalized 30 even though many severe cases occur in the community. Moreover, the surveillance system captured a small proportion of older patients; therefore, counting influenza cases and associated deaths in the hospital setting may lead to underestimation of disease burden. In this surveillance, we only tested samples for influenza viruses and a large proportion of SARI cases were likely associated with non-influenza etiology which remains unknown. Expanded testing would be helpful to understand the etiology of severe acute respiratory disease in Kenya.

| CONCLUSION
We found that the surveillance system mainly captured children aged <5 years. Hospitalized children with SARI who had a reported underlying illness had higher likelihood of in-hospital death compared with those without. A high influenza-associated CFP was reported among infants aged <6 months. These groups are disproportionately affected and may be priority targets for influenza control programs. With the emergence of new respiratory pathogens and the development of new vaccines targeting respiratory illnesses such as the COVID-19 vaccine, hospital-based platforms designed to monitor influenza viruses and associated disease burden should consider adapting their tools and expanding surveillance to capture systematically respiratory viruses that can inform public health interventions and future investments in respiratory disease prevention.