Is Exposure to Sick or Dead Poultry Associated With Flulike Illness?A Population-Based Study From a Rural Area in Vietnam With Outbreaks of Highly Pathogenic Avian Influenza

Background  The verified human cases of highly pathogenic avian influenza in Vietnam may represent only a selection of the most severely ill patients. The study objective was to analyze the association between flulike illness, defined as cough and fever, and exposure to sick or dead poultry.

Methods  A population-based study was performed from April 1 to June 30, 2004, in FilaBavi, a rural Vietnamese demographic surveillance site with confirmed outbreaks of highly pathogenic avian influenza among poultry. We included 45 478 randomly selected (cluster sampling) inhabitants. Household representatives were asked screening questions about exposure to poultry and flulike illness during the preceding months; individuals with a history of disease and/or exposure were interviewed in person.

Results  A total of 8149 individuals (17.9%) reported flulike illness, 38 373 persons (84.4%) lived in households keeping poultry, and 11 755 (25.9%) resided in households reporting sick or dead poultry. A dose-response relationship between poultry exposure and flulike illness was noted: poultry in the household (odds ratio, 1.04; 95% confidence interval, 0.96-1.12), sick or dead poultry in the household but with no direct contact (odds ratio, 1.14; 95% confidence interval, 1.06-1.23), and direct contact with sick poultry (odds ratio, 1.73; 95% confidence interval, 1.58-1.89). The flulike illness attributed to direct contact with sick or dead poultry was estimated to be 650 to 750 cases.

Conclusions  Our epidemiological data are consistent with transmission of mild, highly pathogenic avian influenza to humans and suggest that transmission could be more common than anticipated, though close contact seems required. Further microbiological studies are needed to validate these findings.

In Vietnam, an epidemic of highly pathogenic avian influenza (HPAI) in poultry, subtype H5N1, has been ongoing since late 2003, despite efforts to limit its progress by culling. Poultry breeding is widespread in Vietnam, where about 80% of the population lives in rural areas. In addition to being an important source of income, poultry is kept by many rural households for subsistence farming. The current epidemic in poultry is thus not only a public health problem but also an economic drawback for the many Vietnamese who live in rural areas. Ministry officials in Hanoi accordingly report that interventions requiring poultry culling have been difficult to implement.

In January 2004, Vietnamese health authorities reported an outbreak of severe respiratory illness in previously healthy children.¹ In July 2005, 87 confirmed human cases of HPAI (38 with a fatal outcome) and 1838 verified outbreaks in poultry were reported from Vietnam, which makes Vietnam the country hardest hit by the ongoing H5N1 epidemic.²

The HPAI caused by subtype H5N1 has been described among humans in 3 settings: in Hong Kong in 1997 and 2003, in Thailand in 2004, and Vietnam in 2004 and 2005.^1,3,4 The disease is associated with high case-fatality rates in humans.^1,3,4 Still, little is known about the natural course of HPAI in humans, and the high case-fatality rates are derived from patients admitted to regional or provincial hospitals in larger (ie, major) cities. Whether the confirmed numbers of cases and deaths represent the true incidence and case-fatality rate is unknown because, to our knowledge, no studies assessing exposure to HPAI and disease in a population-based setting are available.

To obtain information about whether HPAI in humans is more common than has been recognized previously, we analyzed the association between flulike illness and exposure to sick or dead poultry in a rural Vietnamese population with a confirmed HPAI epidemic in poultry.

FilaBavi is a longitudinal, population-based demographic surveillance site in Bavi, a rural district of Ha Tay province in Vietnam.⁵ In the FilaBavi sampling, Bavi, with a population of 242 000 inhabitants, was divided into 352 clusters, stratified according to geographical region. Sixty-seven of the clusters were then randomly selected proportional to population size per unit and included in the FilaBavi study sample. The FilaBavi is an open cohort and comprised at the time of this study (April 1 through June 30, 2004) 11 942 households with 45 478 inhabitants. Trained female interviewers collected information about vital events, self-reported symptoms, and in-and-out-migration through quarterly interviews with household representatives. The household representative is usually the female head of the family. These general data were collected in FilaBavi 4 times yearly since 1999. Every other year, a full census with a detailed assessment of socioeconomic and demographic characteristics of all individuals in the cohort is performed. The socioeconomic data used in this study are based on a census performed in 2003.

Official reports from the Vietnamese Ministry of Agriculture confirmed an HPAI epidemic among poultry in Bavi in February 2004, with verified cases in poultry reported from 12 communities in Bavi. Ministry representatives stated that HPAI in poultry in the region was likely highly underreported because of the unwillingness to participate in culling activities. One human HPAI case was confirmed during the first quarter of 2004, outside Bavi but within the province of Ha Tay.

For the purpose of this study, 2 screening questions were added to the regular quarterly interview procedure of the FilaBavi cohort for April through June 2004—1 to identify individuals who had been sick with cough and either fever or dyspnea during the past 6 months, and 1 to identify individuals who had had any kind of contact with poultry (well, sick, or dead) during the past 6 months. If the household representative identified any individual in the household who answered yes to either or both of the questions, this individual was interviewed in person if older than 15 years. For children, a parent or caretaker was interviewed. Interviews were performed immediately if the individual was available; if not, the interviewer came back the next possible day. A pretested, structured questionnaire translated into Vietnamese was used to assess characteristics of exposure and disease during the past 6 months. Detailed questions were posed about direct or indirect contact with well, sick, or dead poultry; poultry breeding within the household or occupationally; self-reported symptoms; and severity of the most serious disease episode.

Neither the respondents nor the interviewers were informed about the research questions. Field supervisors led the work, rechecked all questionnaires, and reinterviewed a randomly selected 5% of respondents within a week after the original interview. Consistency was greater than 90%.

The combination of cough and fever was chosen as the disease indicator and henceforth will be referred to as flulike illness. Household socioeconomic status was measured by using a wealth index based on information about income, expenditure, assets, and housing.⁶

All individuals consented to take part in the study. The long-standing good collaboration between the population, the Local People’s Committee in Bavi, and the FilaBavi research group ensured high participation rates in the FilaBavi demographic surveillance. Similar high participation rates are found in other studies.⁷

We lacked the date of birth for 2 of the 45 478 individuals, so they were excluded from the analyses.

Data were processed and analyzed (SPSS version 11.5; SPSS, Chicago, Ill; and Stata 6.0; StataCorp LP, College Station, Tex). Odds ratios (ORs) with 95% confidence intervals (CIs) were used to test for statistically significant associations between exposure and disease. All univariate associations were tested, and significant associations were entered into multivariable logistic regression models to account for confounding. Only variables that contributed significantly to the models in a likelihood ratio test were kept. To determine whether age, sex, and socioeconomic status were effect modifiers on the association between poultry exposure and flulike illness, we used likelihood ratio tests to assess all models with and without interaction terms between these variables and poultry exposure. The attributable fraction of disease in the exposed group (AF_e) was calculated by using the formula AF_e = (OR − 1)/OR. From the attributable fraction of disease in the exposed group, we estimated the total number of disease cases attributed to direct contact with sick or dead poultry with 95% CIs.

The Ministry of Health in Hanoi, the Local People’s Committee in Bavi, and Umeå University in Sweden granted ethical permission for the study. Informed consent was obtained from all participants at the time of interview. There also are no conflicts of interest in this study; funding agencies have had no active role in the research.

Most (38 373 [84.4%]) of the 45 476 persons included in the study lived in households keeping poultry. Occupational exposure to poultry was also common; 15 043 (33.1%) persons used poultry feces as manure, 14 606 (32.1%) raised poultry on a commercial basis, and 11 755 (25.9%) resided in households reporting sick or dead poultry (Table 1). Having poultry in the household was in itself not a risk factor for flulike illness, but contact with sick or dead poultry was in a univariate analysis significantly (P<.001) associated with flulike illness. Low socioeconomic status, female sex, and young or old age were also risk factors for disease. Smoking, which was reported by 39.8% of men and 0.90% of women, was not a risk factor for disease, nor was household size.

In a multivariable logistic regression model, including age, sex, and socioeconomic status, we found a dose-response relationship between poultry contact and flulike illness. Having poultry in the household was not associated with a significantly increased risk (OR, 1.04; 95% CI, 0.96-1.12); having sick or dead poultry in the household, but without direct contact, was a significant risk factor (OR, 1.14; 95% CI, 1.06-1.23); and having been in direct contact with sick or dead poultry produced the highest risk for flulike illness (OR, 1.73; 95% CI, 1.58-1.89) (Table 1).

Age was an effect modifier of the association between exposure to sick or dead poultry and flulike illness (Table 2). In children younger than 7 years, there was no association between direct contact with sick or dead poultry and flulike illness. In all other age groups, there was a significant association, peaking in adults aged 19 to 45 years (OR, 2.36; 95% CI, 2.13-2.62).

The flulike illness attributed to direct contact with sick or dead poultry was estimated by assessing the attributable fraction of disease in the exposed group (Table 2). Between 650 and 750 cases with flulike illness could be attributed to direct contact with sick or dead poultry.

In addition to the symptoms of fever and cough, combined or individually, headache, vomiting, dyspnea, and sore throat were significantly associated with direct contact with sick or dead poultry (Table 3). Individuals with flulike illness and direct contact with sick or dead poultry were more restricted by their illness compared with those with flulike illness and no direct contact; illness preventing work or school attendance was reported by 81.0% and 63.7%, respectively (OR, 2.41; 95% CI, 2.09-2.78), with a median duration of 3 days in both groups. Bed rest was reported by 27.2% and 24.8%, respectively (OR, 1.13; 95% CI, 1.02-1.25), with a median duration of 2 days in both groups.

Clustering of cases in families reporting flulike illness was slightly more common in households with sick or dead poultry. The mean number of cases per household was 1.69 compared with 1.57 in households with no sick or dead poultry (P = .001; t test).

In this study, we present data concerning exposure to poultry and flulike illness in an area in northern Vietnam where outbreaks of HPAI among poultry have been recorded officially. The use of a demographic surveillance site, where regular, structured interviews have taken place 4 times yearly since 1999, provided an ideal study setting. The specially trained interviewers and the respondents were accustomed to the interview procedures, and a trustful relationship had been built over the years. The specific questions of our study were added to the regular questionnaire, and questions about disease symptoms were asked before questions about exposure. The general perception of human HPAI in the area is that of a rare, life-threatening disease, not the mostly mild respiratory infections reported here; however, some reporting bias may have occurred.

The main finding in our study was an association between direct contact with sick or dead poultry and flulike illness, defined as the combination of cough and fever. The dose-response relationship between degree of exposure and outcome suggests a causal relationship.

Given direct contact with sick or dead poultry, the risk of flulike illness was highest in those aged 19 to 45 years. No association between exposure and disease was found in young children. A likely explanation for these age-related risk differences is various degrees in the intensity of the contact within the group who reported direct contact with sick or dead poultry. Because young and middle-aged adults are more likely to have been in close contact with sick or dead poultry, the peak risk among adults is also consistent with a causal relationship. In addition, the exposure data among younger children may be less accurate because caretakers provided information. The different ORs observed in men and women may have biological causes but could also reflect differences in exposure.

Many of the HPAI cases reported in humans have been clustered in families, and there is evidence of possible person-to-person transmission.⁸ We found the mean number of flulike cases within families with sick or dead poultry was significantly higher than that in families not reporting sick or dead birds. Because clustering of cases could be caused by either a common external exposure or transmission of infection within the family, we cannot draw any conclusions regarding causality.

In the absence of serological data, we cannot state the cause of disease. The observed results could have resulted from other diseases affecting poultry and humans. Psittacosis could kill poultry and produce respiratory symptoms among humans, but outbreaks of psittacosis of this magnitude have not been reported, whereas HPAI was epidemic among poultry in the area during the study. Other entities such as Newcastle disease and botulism may cause epidemic outbreaks among birds, but these diseases do not transmit flulike illness to humans. Transmission of HPAI to humans is therefore the most likely cause of the flulike illness attributable to contact with sick or dead poultry in our study.

The flulike illness reported here is much milder than the disease described among the confirmed human cases of H5N1 in Vietnam and Thailand in 2004.^1,3 The Vietnamese health system is characterized by a growing diversity of health care providers in the private sector together with a national health care system. Results of earlier studies in patients with respiratory symptoms showed a preference for self-medication or unqualified private providers.^7,9,10 Private providers, including pharmacists and unlicensed doctors, most often supply low-quality care and treatment characterized by irrational use of drugs and low or nonexistent levels of referral to specialists.^11,12 The official number of human cases of HPAI is therefore likely to represent a small proportion of severe cases—those that have reached advanced health care with good diagnostic capacity, which is available only in the major cities. Given the existing barriers to qualified health care access, especially important to the rural population, there may have been an underdiagnosis of severe and especially milder human cases of HPAI. Our epidemiological data are consistent with HPAI in humans manifested as a relatively mild, febrile, respiratory infection that easily can go undetected. If so, there are similarities with other emerging infections (eg, Ebola hemorrhagic fever) with an overestimation of disease severity and case- fatality rates in the beginning of an outbreak.

During the current widespread Asian HPAI epidemic in poultry, the disease has been reportedly rare in humans. Our findings, however, suggest that in populations living in close contact with poultry in areas endemic for HPAI, transmission to humans may be frequent. The estimate of 650 to 750 cases attributed to direct contact with sick or dead poultry is conservative because we did not consider indirect contact such as using poultry droppings as manure. Furthermore, all asymptomatic cases and individuals without cough and fever or dyspnea were excluded.

Seroconversion after exposure to HPAI is common. In Japan in 2004, 5 (29%) of 17 unprotected poultry culling workers had antibodies against H5 without having had severe symptoms, and in Hong Kong in 1997, 3% of government workers and 10% of poultry culling workers were seropositive for antibodies against H5.^13,14 However, because seroconversion may also appear after exposure to a viral antigen without infection, these results cannot be considered conclusive for mild or asymptomatic HPAI infection.

If transmission from poultry to humans is more common than anticipated earlier, it also may imply an increased risk of viral reassortment through infected individuals serving as mixing vessels for HPAI and human influenza. Although our data suggest that HPAI in humans could be more common than has been recognized previously, the flulike illness in our study was not easily transmitted from birds to humans. The presence of sick or dead birds in the household resulted in only a slightly elevated OR for cough and fever, and only direct contact with these birds resulted in a convincingly higher OR for disease.

Our results from a large epidemiological population-based study in an area with an ongoing epidemic of HPAI in poultry are consistent with a higher incidence of HPAI among humans than has been recognized previously. The results suggest that the symptoms most often are relatively mild and that close contact is needed for transmission to humans. Our data need to be confirmed with population-based seroprevalence studies and with virology studies in patients with acute mild infection. Time-series and spatial data on respiratory disease from the FilaBavi site are being analyzed.

The findings in this study may affect the choice of design when starting a cross-country surveillance of patients manifesting respiratory symptoms and when taking samples for analyses of H5N1.¹⁵

Correspondence: Anna Thorson, MD, PhD, Division of International Health, Karolinska Institutet, SE-171 76 Stockholm, Sweden (anna.thorson@phs.ki.se).

Accepted for Publication: August 3, 2005.

Author Contributions: All authors had full access to all the data in the study, and the corresponding author takes responsibility for the integrity of the data and the accuracy of the data analysis.

Financial Disclosure: None.

Funding/Support: This study was financially supported by the Swedish Institute for Infectious Disease Control. The FilaBavi demographic surveillance site is financed by the Department for Research Cooperation within the Swedish International Development Agency (Sida/SAREC, Stockholm). The financers have had no role in the research process.