Comparable burden of severe acute respiratory illness among children and older adults in Kenya: evidence from a multi-site health utilization survey

Objectives: Hospital-based influenza surveillance may underestimate the population-level respiratory disease burden in sub-Saharan Africa. Previous Kenyan estimates reported severe acute respiratory illness (SARI) rates among adults aged ≥65 years that were 4-10 times lower than those among children aged <5 years. We used community health utilization survey (HUS) data to estimate national and county-level HUS-defined SARI and influenza-associated SARI episode rates in Kenya, with adjustments for geographic variation and health care-seeking behavior. Methods: Data from 21,088 participants in a 2018 HUS conducted in Nakuru, Kakamega, and Siaya counties were included in the burden analysis. Age-specific HUS-defined SARI episode rates were extrapolated geographically using relative county lower respiratory infection disability-adjusted life-year factors from the Global Burden of Disease 2016 study. Age-specific proportions of episodes with actual admission or recommendation for hospitalization were obtained from the HUS, whereas the 2014 Kenya Demographic and Health Survey provided a relative county health care-seeking modifier. Influenza attribution used all-age positivity from seven surveillance sites or assigned regional sources during 2017-2019. Point estimates used observed or central inputs; 95% uncertainty intervals were derived from 10,000 design-aware resampling and joint uncertainty iterations. Results: The national HUS-defined SARI episode rate was 3318 (95% uncertainty interval: 2477-4326) per 100,000. Rates were 7758 (5008-11,156) among children aged <2 years, 7890 (5698-10,629) among children aged 2-4 years, and 7155 (3108-12,682) among adults aged ≥65 years, producing a U-shaped age distribution. The influenza-associated SARI rate was 365 (270-493) per 100,000. Conclusion: Among households with children aged <5 years, adults aged ≥65 years had a HUS-defined SARI burden comparable in magnitude to that of young children. The directly observed HUS age pattern persisted after national extrapolation. Influenza-associated age-specific estimates showed a similar pattern but were less directly supported by age-specific surveillance data. Community HUS data also identified substantial HUS-defined SARI that was not captured by sentinel hospital admissions.