Study Area and Population

The epicenter of the 1998–2000 MHF outbreak, Durba, is 14 km from Watsa town, the administrative center of the subdistrict of Watsa. Watsa Subdistrict is located near the border with Uganda and Sudan. Watsa's health system was seriously compromised during the outbreak by economic decline and the ongoing war in eastern DRC.

Our survey was of lay persons, referred to as household contacts, whose contact with an MHF patient occurred during lay activities, such as nursing a patient (supporting, feeding, washing, and the like, whether at home or in health facilities), transporting a patient or body, or preparing a body for burial. Healthcare workers whose contact occurred during their professional duties were not eligible.

Cases of MHF were either laboratory confirmed (positive by PCR, antigen-capture ELISA, virus isolation, or a combination of IgG ELISA, IgG indirect immunofluorescence assay [IFA], and clinical and epidemiologic evidence [48 cases]) or epidemiologically linked (persons for whom laboratory confirmation was not attempted who had acute fever, hemorrhage, and contact with a laboratory-confirmed patient [25 cases]). Forty-five cases were known from surveillance during the outbreak; we identified 28 retrospectively. Contacts of suspected case-patients whose conditions were not laboratory confirmed or epidemiologically linked as defined above were not eligible because their diagnosis lacked certainty.

We attempted to visit the households of all 73 MHF patients and to prepare a list of persons who had direct contact with the patient or his or her body fluids or remains. If contacts were temporarily absent, we undertook at least 2 repeat visits. If they had moved away, we tried to locate them at their new address, unless distance or lack of security (e.g., rebel activity, bandits) hindered us, in which case we interviewed former neighbors about the contacts' disease episodes in the 4 weeks after the patient's illness. We asked all contacts we met to give verbal informed consent; if they agreed, we interviewed them and took blood samples. This study was approved by the ethics committee of the Antwerp Institute for Tropical Medicine and the representative of the Ministry of Health in Watsa.

Interviews

After establishing the identity of the contact and the relationship to the patient, we asked an open-ended question about the role the contact played during the patient's illness. We also asked closed-ended questions on whether the contact had touched, carried, or embraced the patient (and whether the patient at that point had diarrhea, vomiting, or bleeding) and whether the contact had touched the patient's clothes or linen (and whether these were soiled with stool, vomitus, or blood). Since patients who died often had had diarrhea, vomiting, and bleeding in the final stages of disease, we also asked whether the contact had touched, carried, embraced, or washed the person after death. While field testing the questionnaire, we found that protective gear such as gloves was unavailable to lay persons; thus, all contacts were assumed to be unprotected. We asked about symptoms the contact had experienced during the 4 weeks after exposure; these symptoms (Table 1) correspond to the ones used during the epidemic to define a clinically suspected case.

Blood Sampling and Testing

After the interview, 5–10 mL venous blood was taken from contacts. After 12 to 24 hours, serum was separated from the blood clot, refrigerated at ≈4°C, and transported to the Uganda Virus Research Institute (within 1 to 2 weeks). There it was frozen at –70°C and shipped on dry ice to the National Institute for Communicable Diseases, Johannesburg. Serum was examined by ELISA and IFA and considered positive if anti-Marburg IgG was found in both tests.

Data Analysis

Data were entered with EpiInfo version 6.0 (Centers for Disease Control and Prevention, Atlanta, GA, USA), and analyzed with Stata version 8.2 software (StataCorp LP, College Station, TX, USA). Depending on symptoms associated with increased virus shedding and on the intensity of the contact, level of exposure was categorized. Low-level contact included any direct contact with a living case-patient without diarrhea, vomiting, or bleeding; or touching clothes or sheets not soiled with stool, vomitus, or blood. Medium-level contact was defined as touching a living case-patient with diarrhea, vomiting, or bleeding; touching clothes or sheets soiled with stool, vomitus, or blood; or touching remains. High-level contact included carrying or embracing a living patient who had diarrhea, vomiting, or bleeding; or carrying, embracing, or cleaning remains.

We established transmission chains and generations for all patients, taking into account work as a gold digger, exposure to other patients, incubation period, and date of onset (M. Borchert, unpub. data). When a patient had been working as a gold digger and had been exposed to another patient, we gave priority to the confirmed human-to-human exposure over the possible primary exposure and classified these cases as nonprimary ones.

We calculated the secondary attack risk as the proportion of household contacts of primary case-patients who then became secondary case-patients themselves, including only primary case-patients whose contact list could be established fully and who did not share contacts with another case-patient; we used the analogous approach to estimate the tertiary attack risk. We computed the reproduction number (R_p) as the product of the secondary attack risk and the average number of contacts per primary case-patient. As most cases in our survey had occurred after control measures were implemented, we consider this number to be the postintervention R_p, not the basic reproduction number R₀.

Completeness of Data

Household contacts of 73 MHF patients were eligible to participate in the survey. We completed contact lists for 48 patients (66%). For 7 patients, Watsa health authorities had listed some contacts during the epidemic. Because we could not meet these patients, their contacts, or others who could verify the list's completeness we likely missed some contacts. For 18 patients, no contacts had been listed by the health authorities. Since we did not speak to anyone who had witnessed the case during the epidemic, contacts have also probably been missed for these cases.

Existing surveillance records listed 141 contacts. For the 48 cases we could investigate fully, 96 additional contacts were found. Seventy-one of these were contacts of patients identified by surveillance during the outbreak. The total number of identified contacts therefore was 237 (141 + 96), relating to 55 (48 + 7) of 73 cases.

A patient whose case was fully investigated had, on average, 4.46 contacts; on this basis, one would expect 326 contacts for all 73 patients. The 237 identified contacts correspond to 73% of this expected number. A total of 143 contacts could be traced, and 124 consented to being interviewed and giving a blood sample, representing 52% of the 237 identified contacts and 38% of the 326 expected contacts. Three persons listed by surveillance denied any physical contact with the patient and were excluded from analysis. Therefore, results refer to 121 study participants.

Characteristics of Contacts

The median interval between the onset of the patient's disease and the contact's interview and blood sample collection was 24 months (range 11–48). Half of the contacts were female, and three fourths were 15–49 years of age. Most contacts were family members (88%), while colleagues accounted for 11% (Table 2).

Half of the contacts held or carried a patient, a third fed or washed a patient, and a tenth reported sharing a bed with a patient (Table 3). Exposure to a living patient was almost universal; three quarters of contacts had exposure to body fluids and excreta. Forty-three percent of contacts had exposure to remains. The exposure level was low in 13%, medium in 19%, and high in 68% of contacts and did not differ between the sexes.

For 43 of the 50 contacts known to surveillance, we compared exposure reported in our survey with exposure documented by surveillance officers during the outbreak. For 88% of contacts, surveillance and study information agreed.

Two study participants were positive for anti–Marburg IgG: a 21-year-old brother and a 27-year-old male neighbor of MHF patients. Both contacts were highly exposed to their respective primary case-patients. These contacts were also, as unofficial gold miners, at risk for primary transmission themselves (6). The 21-year-old reported 6 general symptoms within 4 weeks after exposure, including fatigue, abdominal pain, nausea/vomiting, hiccoughs, chest pain, and difficulty breathing, but he did not fulfill the definition of a suspected case because he did not exhibit fever or bleeding. The 27-year-old reported a hemorrhagic fever syndrome, including vomiting and coughing blood and bloody or black stool. Neither contact sought medical care. We consider them to be additional confirmed patients and classified them as secondary cases because of the combination of high exposure and postexposure symptoms compatible with MHF. The 119 seronegative contacts were considered nonpatients. Thus, the overall seroprevalence in our study population is 1.65% (95% confidence interval [CI] 0.2%–5.8%), the same as in the general population (1.64%) (6).

Although almost all contacts were seronegative, one third reported fever within 4 weeks of contact with a patient (Table 1), more than one half reported a general symptom (headache, fatigue, and loss of appetite most frequently), and 3.3% reported hemorrhage. Thirty-three (27%) contacts would have qualified as clinically suspected case-patients during the epidemic and should have been taken to an isolation ward for assessment by an experienced healthcare worker. This did not happen, and 23 of these persons were not even known by authorities to be contacts.

On the basis of surveillance records and interviews with family members, neighbors, or colleagues of the 113 eligible contacts we could not interview or obtain blood samples from, we identified 1 epidemiologically linked patient, 1 suspected MHF case-patient, and 13 noncases; for the 98 remaining contacts, information was insufficient to classify them. The total of MHF cases thus increased to 76 (50 laboratory-confirmed, 26 epidemiologically linked).

Secondary Attack Risk and Postintervention R_p

Thirty-one of 76 cases were identified as primary, 21 as secondary, 15 as tertiary, and 5 as quaternary. Four cases could not be classified. Eleven patients with secondary cases acquired their infection as a household contact and had only 1 patient with a fully investigated primary case as a possible source. These constituted the numerator for the secondary attack risk and contributed to the denominator. Forty-two healthy contacts with only 1 patient with a fully investigated primary case as possible source also contributed to the denominator. The secondary attack risk was thus estimated as 21% (11/[11 + 42], 95% CI 11–34) for household contacts. Restricting the calculation to confirmed primary cases did not significantly change the secondary attack risk estimate. The average number of household contacts per fully investigated primary case was 4.46, so that R_p for household contacts was estimated as 0.93. The tertiary attack risk (6/32 = 19%, CI 7–36) did not differ from the secondary one.