Although there have been a number of studies of risk factors for RVF infection and illness, this is the first published study to focus on risk factors associated with severe RVF disease and death. A potentially important finding was that consuming or handling products from sick animals was significantly associated with acute RVF infection, severe illness, and death. Mosquito-related exposures were difficult to quantify and were not associated with infection or severe disease in the multivariable analysis. During bivariable analysis, contact with animals (cows, sheep, or goats) was significantly associated with acute infection with RVFV and with severe RVF disease. In our study, similar to studies conducted during previous RVF outbreaks,13–15,19
exposures associated with animal contact, including consuming or handling products from sick animals (milk, meat, or blood) and being a herdsperson were significantly associated with acute RVFV infection. In the current study, these exposures were also associated with severe disease in the bivariable model. However, contact with an aborted animal fetus was the only independent factor (in the multivariable model) associated with increased likelihood of severe RVF disease. It is possible that aborted fetuses contain high quantities of RVFV, increasing the risk either through direct secretions or aerosolizing of virus after touching the animal. Touching an aborted fetus was also shown to be a risk factor for infection during a study of sporadically occurring RVF during a non-epidemic period.32
Previous studies have shown that certain types of exposures to animals and their secretions (through slaughtering or sick animals) were associated with infection with RVFV, and transmission is also felt to occur by bites from infected mosquitoes.11,12
Because most RVF infections are subclinical,33
we set out to determine whether specific factors were associated with severe disease. It is likely that exposure to infected animals and their secretions provide greater opportunities for infection with large inoculum of RVFV. This contrasts with the presumably much smaller inoculum of virus likely associated with one or more bites from infected mosquitoes. Previous studies of other infections in animals have documented that the probability of severe disease is determined by the size of the inoculum, the route of inoculation, and the frequency of naturally occurring immunizing inoculations (infections that do not cause symptoms but do induce protective immunity).34
In humans, there are a variety of diseases for which inoculum affects clinical presentation and outcome including hepatitis B,35
hemolytic uremic syndrome caused by verocytotoxin-producing Escherichia coli
and yellow fever.39
Exposures to infected mosquitoes might be sufficient to stimulate immunity, but perhaps more rarely to lead to severe outcomes (like mortality). Consistent with this notion are data showing that while the highest prevalence for acute RVF infection during this outbreak was in Baringo District, the case-fatality ratio was lowest there.24
In Baringo District, mosquito densities were higher than in Garissa and Kilifi,40
and the proportion of cases involved with occupations linked with animal care practices were lower.24
However, an important limitation in reaching this conclusion is that enhanced surveillance in Baringo district, based on lessons learned in Garissa district and Kilifi district (where RVF occurred a few months earlier) may have led to more complete detection of cases in Baringo. This limitation may also be responsible for the greater proportion of acute infections associated with severe disease in Baringo when compared with Kilifi District.
The association of acute RVF infection and consuming or handling products from sick animals has been documented in other studies.6,19,41,42
Previous studies have also reported high-risk occupations for RVF infection, including abattoir workers,43
However, earlier studies have not focused on factors associated with severe disease or death among persons infected with RVF virus; thus, the findings of this study linking consuming or handling products from sick animals to death will provide further impetus to focus on minimizing the potential for these exposures during future RVF outbreaks. The differences between districts in terms of access to health care or hospitals, active case finding and modalities of supportive care available and provided to RVF patients, not studied during this investigation, need to be considered during future studies.
The PAR% quantifies the contribution of a risk factor to the outcome of interest. The PAR% for death among individuals who handled or consumed sick animal products was 47%. This means that 47% of the deaths among individuals acutely infected with RVFV can be attributed to consuming or handling sick animal products, and if this exposure had been eliminated, then 47% of the deaths among the study population could have been prevented. Likewise, the PAR% of consuming or handling sick animal products for acute infection was 19%, suggesting that many mild and severe infections could have been prevented in the study population if consuming or handling products from sick animals had been eliminated.
The high proportion of acute RVF infection and severe disease in males and the association of being male with death during bivariable analysis suggest exposure to a higher dose of infectious virus among males or potential susceptibility of males to RVF infection and disease. However, unlike the RVF outbreak in Northeastern Kenya in 1997–199819
and a recent study in northeastern Kenya of risk factors for sporadically occurring RVF during a non-epidemic period,32
gender was not an independent risk factor for infection in our study. Association of male gender with acute RVF infection in bivariable analysis was likely related to the occupation of herding, predominantly performed by males, which involves increased animal-related exposures, such as consuming or handling sick animal products during slaughter, milking, or skinning. Because of their close proximity to animal herds, herdspersons may also be at greater risk of being bitten by mosquitoes that have bitten infected animals.
A high proportion of acute RVF infections were in housewives. This could be related to their handling sick animal products during food preparation procedures. In addition, the findings of our study suggest that younger and older persons may be more susceptible to severe disease and infection, respectively, as evidenced by the high proportion of severe disease among those less than 14 years of age and the high proportion of acute RVF infection among persons greater than 50 years of age.
In addition to the difficulties in assessing mosquito exposures by interview with much precision, limitations of this study included temporal challenges in linking exposures to infection and in case classification. It was not possible to carry out this study prospectively to establish a temporal relationship between exposure and outcome, as both exposure and outcome had occurred by the time the study was initiated. Second, because the outbreak was still ongoing at the time the investigation was conducted, some controls may have been misclassified if they had been exposed and were in a pre- or subclinical stage of the disease and had not yet sero-converted. Of the few clinical infections that have been followed closely for serologic conversion, RVFV-specific IgM antibodies appear around Day 5, peak by Day 30, are absent in 50% of patients by Day 45, and are undetectable at 4 months.6
We could also have missed people who had been exposed to RVFV > 45 days before having their blood drawn for the survey. They would have been excluded from these analyses because they had IgG antibodies with no IgM antibodies, and it is possible that earlier in the outbreak their exposures resulting in infection were different than later in the outbreak. Finally, during the household interview some relevant family members (e.g., herdsmen or household heads) were in the field grazing livestock or at work; while an attempt was made to revisit the herders or household heads, contact was not always possible. Thus, we may have missed the opportunity to characterize other important exposures, and we may have underestimated the magnitude of the proportion of residents who were infected.
The findings of this study emphasize the potential importance of effective health education campaigns to prevent transmission through handling or consuming infected animals. Pictorial narratives could include messages such as “Do not slaughter, skin, milk, or provide birthing care to sick animals;” “Bury or burn carcasses during an outbreak;” “Boil all milk;” “Wear personal protective equipment -gloves, coveralls, boots, eyewear, and mask when handling sick animal product;” “Avoid contact with infected tissues, blood, milk, meat, and aborted fetuses, or sick animals.” During this RVF outbreak several control measures were put in place, which included a ban on raw milk, home slaughter, and animal quarantine; however, slaughter bans and animal quarantine were difficult to enforce because of the critical role of livestock in the livelihoods of residents in many of the affected areas. Another strategy might be to place community health workers, armed with health messages, at points of congregation of high risk individuals, such as watering holes for livestock and market places.
Although mosquitoes may play a role in amplifying and spreading virus among animals, their role in transmission to humans, and especially in causing severe RVF illness, was difficult to evaluate in this study. It was impossible to identify persons bitten by different species of mosquitoes, to detect whether the biting vector had RVFV, and the dose of inocula per bite. Further studies are needed to evaluate the role of mosquitoes and other vectors in transmission of RVFV to humans and animals.