This is the first study to describe risk factors for seasonal influenza hospitalization among older children and adults in Africa 
. Our findings suggest that persons with chronic underlying illnesses, particularly HIV and chronic lung disease, are at increased risk of influenza-associated hospitalization, and would be priority candidates for influenza vaccination. While these factors are not unique to Africa, they are more common in Africa and therefore may play a larger role in the epidemiology of severe influenza.
Few studies have investigated the risk of influenza hospitalizations in HIV-infected persons 
. In South Africa, HIV-infected children with confirmed influenza had an estimated eight-fold increased risk of hospitalization than did HIV-negative children 
. In the same population, HIV-infected children with influenza-associated pneumonia had a higher case-fatality ratio (8%) than did HIV-negative children (2%) 
. No data are available before this study on the contribution of HIV infection to influenza hospitalization in African adults 
. In the U.S., one study showed that HIV-infected women had the highest influenza-attributable risk for severe cardiopulmonary events of any high risk group 
. In an outbreak of influenza in a rehabilitation center in Italy, HIV-infected persons had an increased risk of influenza-like illness (ILI), as well as undefined complications 
. In an outbreak in a substance abuse rehabilitation center in New York, there was no increased risk of ILI among HIV-infected residents, although 5% of them were hospitalized versus none of the HIV-negative ILI residents 
. In our study, like previous studies, CD4 count was not associated with risk of hospitalized pneumonia among HIV-infected persons 
Whether the increase in influenza hospitalization among HIV-infected persons is due to an increased susceptibility to influenza virus infection, a greater chance of developing severe complications of influenza or a lower threshold of clinicians to hospitalize HIV-infected persons is not clear. HIV-infected persons do appear to have higher rates of severe influenza complications, hospitalization and death than HIV-negative people, based on findings from developed countries 
. This is also likely the case in Africa, where co-morbidities among HIV-infected persons, such as tuberculosis, are even more prevalent 
. It is also possible that HIV-infected persons with influenza are more likely to develop secondary bacterial pneumonia leading to their hospitalization, as suggested by an increase in suspected bacterial pneumonia among HIV-infected children hospitalized with influenza in South Africa 
. Another potential explanation is that clinicians who know that someone with respiratory illness is HIV-infected might be more likely to hospitalize that person or that persons with HIV-infection who develop influenza are more likely to seek care at hospitals. We believe that the latter is likely not a large contributor in rural western Kenya, where more than two-thirds of HIV-infected persons did not know their HIV status 
. Lastly, it is also possible that HIV-infected persons are admitted for other concomitant HIV-related diseases rather than for influenza. Fewer than 30% of case-patients had an admitting diagnosis of pneumonia; however, the influenza testing results were not available to hospital clinicians, who make most diagnoses empirically without radiologic or laboratory data, including blood smears for malaria, which tends to be over-diagnosed in this setting 
Characterizing an increased risk of influenza complications among HIV-infected persons would be especially important in sub-Saharan Africa, where 22 million people are estimated to have HIV/AIDS 
. Several interventions may decrease the influenza burden in HIV-infected persons. Highly active antiretroviral therapy (HAART) has been shown to decrease the risk of influenza complications 
. Although most HIV-infected Africans who need HAART are still not receiving it, the percentage is increasing as efforts to expand HIV testing and implement widespread HIV treatment increase in Africa 
. Second, several studies suggest that influenza vaccine is effective in HIV-infected persons who are not severely immunocompromised 
. The only randomized controlled trial designed to address efficacy of influenza vaccine in HIV-infected persons showed a 20% reduction in influenza-like illness and 100% efficacy against lab-confirmed influenza 
. Currently, influenza vaccine is not available in the public sector in most of sub-Saharan Africa 
We found persons with chronic lung disease were at increased risk of influenza hospitalization, which is consistent with previous studies done in the U.S. 
. Pulmonary TB also had a significant association with hospitalized influenza in bivariate, but not multivariable, analysis. As with HIV, because we focused on influenza among hospitalized patients, we were unable to distinguish whether persons with TB and chronic lung disease are more likely to become infected with seasonal influenza viruses, or whether once infected they are more likely to develop severe disease requiring hospitalization. In addition, healthy controls might have had fewer contacts with the healthcare system in the past year, therefore lacking the opportunity to be diagnosed with pulmonary TB or chronic lung disease that cases had, leading to a misclassification bias.
We found that influenza hospitalization was associated with ownership of cattle and chickens. This association most likely reflects confounding, as animal-to-human transmission of influenza viruses, although documented, is very rare, especially for seasonal influenza viruses 
. What the potential confounders might be are not clear. People who own domestic animals may have more contact with other people in crowded settings, such as animal markets, increasing the opportunity for influenza transmission. There might also be selection bias in that animal owners who were eligible to be controls might have been less likely to be at home or take the time to participate in the study. Additionally, animal owners might represent a different category of persons, with greater social mobility or access to healthcare, more likely to go to hospitals when ill, making them more likely to be cases.
Several limitations might have affected our study. First, HIV status at the time study enrollment was assumed to be the status at the time of influenza infection. While this was likely the case, it is possible that a few people had sero-converted in the interval between influenza virus infection and study enrollment, although this misclassification should have been non-differential between cases and controls. In addition, two cases died between hospital discharge and study enrollment and could not be HIV-tested by the study team. Because they died, they had a higher probability of being HIV-infected, which might have led to an underestimation of the true risk of HIV. Second, inherent in the design of case-control studies is the possibility of recall bias. Cases might have remembered some exposures more than controls. Moreover, many months often elapsed between illness and interviews, so that exposures in the week prior to illness might not have been remembered well. Nonetheless, many factors we evaluated are unlikely to change over time, and both cases and controls were asked to recall exposures during time periods similarly distant. Third, not all hospitalized patients with respiratory illness met the screening criteria, as we used an established severe acute respiratory illness surveillance case definition, so that some inpatients with influenza were not included 
. In particular, we did not include patients without documented fever or hypoxia. Therefore, it is possible that our findings are not generalizable to all influenza-associated hospitalizations.
This study underscores the role of underlying conditions, particularly HIV infection, in hospitalized seasonal influenza patients in Africa. Persons with such conditions should be prioritized for influenza vaccination and other preventive strategies. The effect of these conditions on the burden of severe influenza disease during the recent 2009 H1N1 pandemic needs further exploration.