Ebola Hemorrhagic Fever Associated with Novel Virus Strain, Uganda, 2007–2008

Case-fatality rate is lower for this strain than for the other 3 Ebola species known to be pathogenic to humans.

second report, ill persons had an acute febrile hemorrhagic illness, and tests conducted by the US Centers for Disease Control and Prevention (Atlanta, GA, USA) confi rmed the illness as EHF on November 29, 2007.

Outbreak Site
Bundibugyo district is located in western Uganda ( Figure 1). Approximately 60% of the district is covered by the Rwenzori Mountains and Semliki National Park and Game Reserve, which have a broad range of wildlife, including primates. The district has a population of 253,493 and a population density of 108 persons/km 2 (9), 1 hospital, and 26 health centers. The main economic activities are cocoa farming, fi shing, and tourism. Hunting of wild animals is common among settlements near the national park.

Epidemiologic Activities
During November 29, 2007-February 20, 2008, village health teams conducted active searches in the communities through daily door-to-door visits in their respective villages while investigation teams of local and international experts reviewed case notes at health facilities and verifi ed suspected cases reported by the village health teams. Cases were categorized as suspected EHF, probable EHF, and confi rmed EHF ( Table 1).The UMOH developed working case defi nitions and established an enhanced surveillance system for identifying suspected case-patients in the affected district and in contiguous districts.

Investigative Activities
Investigation teams actively searched for case-patients in health facilities and communities and retrospectively reviewed hospital records. Increased awareness of the disease through public education campaigns and the media facilitated reporting of suspected cases to health authorities. Village health teams assisted in case identifi cation and reports and contact follow-up at the community level. Clinical and epidemiologic data were systematically collected from persons with suspected, probable, and confi rmed cases. Proxies (usually parents, spouses, or adult siblings) were interviewed for information about case-patients who had died before an interview could be conducted. Persons with suspected cases identifi ed in the community were transported by a mobile ambulatory team to designated isolation facilities. A triage desk was established in the outpatient departments of each health facility in the district to screen for suspected cases and notify the surveillance system. Clinical specimens, including blood, were collected for laboratory testing from all persons with suspected EHF. All contacts were entered into, and follow-up schedules were drawn by using, the Field Information Management System database (10).

Laboratory Methods
Five milliliters of blood was collected from all persons with suspected EHF at least 4 days after symptom onset when possible. Specimens initially were transported to the Uganda Virus Research Institute (UVRI) and then transferred to US Centers for Disease Control and Prevention for laboratory analysis. However, on December 4, a laboratory was set up at UVRI, and subsequent specimens were tested there, as described by Towner et al. (11). Virus isolation was not attempted at UVRI.

Data Analysis
We used Epi Info software version 3.4 (12) to create a database into which information from individual case investigation forms was entered and updated daily. The age and sex population structure and projections for Bundibugyo district were obtained from the population and housing census 2002 data (9), and attack rates were computed by using the district population projections. Current geographic maps were obtained from the World Health Organization Health Mapper Mapping Software, version 4.2 (13).
We tabulated risk factors by case status and calculated odds ratios (ORs) using as the reference group persons with suspected EHF who had negative test results. To control for confounding and to test for effect modifi cation, we entered variables with p values <0.1 into a multivariable logistic regression model. By using backward elimination, we eliminated all variables that were not statistically significant at p<0.05. The model had the following co-variables: admission to hospital or visit with a sick person, consulta- tion with traditional healer, participation in funeral rituals, travel to area with cases, contact with a known suspected/ confi rmed case-patient, contact with wildlife, subcounty of residence, age, and sex. The Uganda National Council of Science and Technology expedited ethical review and clearance because the data were being collected to guide outbreak control.

Outbreak Response
UMOH declared an EHF outbreak on November 29. Consequently, the National Task Force, composed of both local and international partners, was activated to determine the magnitude of the outbreak and coordinate immediate outbreak response.
All case-patients were treated in isolation wards set up in Kampala at the National Referral Hospital, Mulago, and in Bundibugyo at Kikyo and Bundibugyo health facilities. Health workers on the case-management teams had participated in patient management during the SEBOV outbreak in Gulu in 2000, thus, their experience provided a valuable resource during the response. Countrywide social mobilization and health education during the outbreak involved developing and disseminating health education materials and messages through mobile video shows, community meetings with opinion leaders, radio talk shows, and newspaper inserts.

Index Case Investigation
The putative index patient was a 26-year-old woman from Kabango village, Kasitu subcounty, in Bundibugyo district. Hunting spears were found at her home, but hunting as a practice was denied. Fever and general weakness developed in this woman, for which she was hospitalized on August 1. She delivered a preterm infant the following day. Diarrhea and diffi culty breathing developed, but hemorrhagic manifestations did not appear. She died on August 4.
In this cluster, 9 case-patients and 6 deaths (the neonate, sister, mother, and 2 nieces of the index patient) (CFR 67%) were reported. The mother and sister were involved in nursing and handling the remains of the index patient because, at the time, barrier nursing and supervised burials had not been initiated. Because these persons were affected before the outbreak was confi rmed, the high CFR was attributable to inadequate supportive care. Three survivors were tested; 2 were positive for Bundibugyo ebolavirusspecifi c IgG.

Symptoms
The median incubation period from contact with an infected person to symptom onset was 7 days (range 2-20 days). The most frequent symptoms were fever (100%), intense fatigue (92%), headache (87%), abdominal pain A person who had slept in the same household and/or had direct physical contact with a person (dead or alive) with suspected, probable, or confirmed EHF and/or had been exposed to an infected person or to an infected person's secretions, excretions, tissues, or linens within 3 weeks after that person's onset of illness.
(87%), vomiting (83%), and diarrhea (83%). Bleeding manifestations (including hematuria; hematemesis; bleeding from the eyes, nose, and vagina; and/or bloody stool) occurred in 52% of case-patients, and most (59%) case-patients who died had some form of bleeding. Other symptoms reported were dysphagia (54% of case-patients), dyspnea (52%), and rash (50%). Median duration from symptom onset to recovery was 10 days (range 2-26 days). Median duration from symptom onset to death was 10 days (range 3-21 days). The outbreak showed 3 successive cycles of transmission ( Figure 2). Each transmission cycle lasted 6 weeks, with an interval of 3-11 days, and was characterized by a typical epidemic peak followed by gradual decline. The highest peak occurred on November 25. Despite the timely detection and initiation of investigations during the 30th epidemiologic week, the successive transmission cycles indicated delayed laboratory confi rmation and declaration of the outbreak. Outbreak confi rmation was delayed until the 48th epidemiologic week, resulting in inadequate control measures before the outbreak was confi rmed.

Geographic Characteristics and Risk Factors
The outbreak was confi ned to Bundibugyo district. More than 97% of cases were detected in 4 of the 10 subcounties in the district (Kasitu, Bundibugyo town council, Bubukwanga, and Busaru), with >54% of cases occurring in Kasitu subcounty ( Table 2).
The overall attack rate in the district was 43 cases/100,000 population. The highest rate occurred in Kasitu subcounty, followed by the rate in Bundibugyo town council (within Bundibugyo district (Table 2). By sex, the attack rate for men was higher than that for women (64 vs. 47/100,000 population). By age group, the attack rate was higher for persons 41-50 years of age than for persons 51-60 years of age (146 vs. 122/100,000 population).

Analysis of Possible Risk Factors
All case-patients were investigated for exposures within 3 weeks before development of symptoms. To better determine risk factors for the disease, we conducted a bivariate analysis using the 76 non-case-patients as the reference group to separately assess persons with probable and confi rmed EHF (Table 3). Before the institution of strict isolation policies, visitors had direct contact with patients through shaking of hands, hugging, or contact with potentially infected surfaces. Patients with confi rmed EHF (OR 8.71, 95% confi dence interval [CI] 3.03-26.30) and patients with probable and confi rmed cases combined (OR 2.56, 95% CI 1.35-4.85) were signifi cantly more likely to have visited sick persons or to have visited the hospital 3 weeks before becoming sick. Consultation with a traditional healer within 3 weeks before illness onset was not signifi cantly associated with having EHF (OR 0.16, 95% CI 0.01-1.15) ( Table 3).
The other risk factor identifi ed was participation in funeral rituals before onset of illness. The funeral rituals, performed by close relatives, involved washing and dressing the body of the decedent. Patients with probable EHF (OR  3.20, 95% CI 1.55-6.64) and confi rmed EHF (OR 7.37, 95% CI 2.89-19.27) were signifi cantly more likely than persons in the reference group to have participated in funeral rituals before they became sick (Table 3). Similarly, when both groups of patients with probable and confi rmed EHF were combined (OR 4.22, 95% CI 2.17-8.24), the members of the new group were signifi cantly more likely to have participated in funeral rituals before becoming sick (Table 3). Ebola case-patients were asked whether they had had any form of contact with a person known to have suspected or confi rmed EHF. Patients with probable and confi rmed EHF combined were signifi cantly more likely (OR 2.66, 95% CI 1.35-5.24) to have had contact with a person known to have suspected or confi rmed EHF before becoming sick (Table 3).
We performed multivariate analysis using binary logistic regression to control for confounding and to test for effect modifi cation. After backward elimination, participating in funeral rituals remained as the sole signifi cant risk factor associated with being a probable/confi rmed casepatient (adjusted OR 3.83, 95% CI 1.78-8.23).

Discussion
The EHF epidemic in Bundibugyo district in western Uganda during August 1, 2007-February 20, 2008, was caused by a new species of the virus, Bundibugyo ebolavirus. It most likely was transmitted from wildlife located within Semliki National Park. The median incubation period of 1 week was less than that reported previously (5) but within the range reported in other Ebola outbreaks (1,2). In addition, patients' symptoms were similar to those reported previously, except for the manifestation of bleeding, which occurred >2× more often than during the SEBOV outbreak in Gulu in 2000 (5).
The Bundibugyo ebolavirus outbreak (CFR 34%) caused a lower proportion of deaths than did the SEBOV outbreak in Gulu (CFR 53%) (5). Similarly, the CFR for the Bundibugyo outbreak was lower than that reported from other outbreaks outside Uganda (1,2), which indicates either that the new virus strain may be less virulent or that improved interventions led to more timely case identifi cation and better case management.
During the Bundibugyo outbreak, vital functions were sustained by supportive treatment, including administering antipyretics, monitoring fl uid balance, and giving antibacterial or antimalarial drugs for concurrent bacterial or protozoal infections. Oral rehydration and oral administration of antibacterial drugs were encouraged for all patients, provided they were conscious and not vomiting; otherwise, fl uids and antibiotics were administered intravenously. To streamline case detection, village health teams and ambulance teams were trained early to conduct active case search and referral. Strict isolation measures included the establishment of triage in all health facilities, designation of isolation wards, training of healthcare workers in adherence to standard precautions, barrier nursing, supervised burial, and ambulance services. During the Bundibugyo outbreak, case-patients were more likely than non-case-patients to have participated in funeral rituals. The practice exposes contacts to infectious body fl uids that have been associated with acquiring EHF (4).
Fourteen health workers were infected during the Bundibugyo outbreak before strict isolation procedures were initiated. During the SEBOV outbreak in Gulu, 64% of health workers were infected after isolation wards were established (14).
The outbreak response had 2 challenges and at least 1 limitation. Investigations at the local and national levels were conducted in a timely manner, but the lack of capacity for laboratory confi rmation delayed initial outbreak confi rmation and therefore the initiation of an appropriate response. Because hunting in the national parks is illegal, attempts to link the Bundibugyo outbreak to wildlife were futile because none of the families investigated admitted to participating in hunting. The use of hospitalized patients as a comparison group was economical, but their illnesses could have been related to risk factors for Bundibugyo ebolavirus infection, hence rendering those risk factors undetectable.
We recommend that an index of suspicion for Ebola viruses (and Marburg virus) be maintained for clusters of cases with fever of sudden onset, intense fatigue, abdominal upsets, and evidence of person-to-person transmission. In concert with timely initiation of active case searching, use of ambulance and burial teams, and strict adherence to patient isolation practices, an index of suspicion should ensure mitigation of the identifi ed risk factors.