A prolonged, community-wide cholera outbreak associated with drinking water contaminated by sewage in Kasese District, western Uganda

Kasese District is located in the western part of Uganda. It has 20 rural sub-counties and 4 towns with 115 parishes/wards and 656 villages. According to the 2014 National Census, Kasese District has a population of 702,029 people, including 338,796 males (48%) and 363,233 females (52%) [10]. The majority (75%) of the population lives in rural areas. Bwera Sub-County, where the first batch of cases reportedly occurred, is located about 5 km from the border with DRC, and has multiple markets with bustling cross-border trading.

To find cases systematically, we reviewed patient records kept at Bwera Hospital from 1 March 2015 onward to identify suspected cases. The patient records contained basic information for each patient, i.e., name, age, sex, residence, date of admission, date of hospitalization, and symptoms. We also reviewed data in the Health Management Information System, an electronic health data reporting system managed by MoH, on cholera cases reported in the area [9]. We then interviewed suspected case-patients to collect information on their food and water consumption history using a standardized questionnaire. We also worked with other health facilities and members of Village Health Team to actively identify cases using the standard case definition. The local public health authorities set up 3 cholera treatment centres in the most affected areas, where all suspected cases were managed with oral and intravenous fluid rehydration and antibiotics. We line-listed all patients at the cholera treatment centres. We performed descriptive analysis of the data in the line-list on patients’ clinical presentations, as well as distribution of the cases by patients’ age, sex, place of residence, education level, and used epidemic curves to describe case-patients’ dates of onset.

Based on the findings from the descriptive analysis, we developed a hypothesis about the most likely mode of transmission. To test the hypothesis, we conducted a case-control study in the most affected areas, i.e., Bwera and Kitswamba sub-counties.

We recruited confirmed case-persons in the case-control study. If a household had 2 or more eligible case-persons, we only interviewed one. For each case-person, we selected 4 control persons. A control-person was an individual who never had diarrhoea from February 2015 to the time of the investigation, resident in the same village as the case-person and within the same 5-year age group as the case-person. To select control-persons randomly, we obtained a list of the households in each village with eligible cases, and wrote the names of the household heads on paper lots. We randomly drew out four times as many paper lots as the number of cases in the village, and attempted to identify an eligible control-person from each of the selected households. If not enough controls were found, we repeated the process until we had found enough controls.

We administered a questionnaire to the eligible case- and control-persons to obtain information on their food and water exposures. Data were collected on the respondents’ water source, whether the respondents treated (using chlorine tablets) or boiled their drinking water, what kind of food they usually ate, whether they usually ate hot or cold food, as well as demographic variables (i.e., age, sex, place of residence, place of work, occupation). We used the 2014 Uganda Population and Housing Census data on populations up to parish level to calculate attack rates. We stratified the data by age-group and village to obtain Mantel-Haenszel odds ratios (OR_M-H). We trained 6 health workers from Bwera Hospital on how to identify case- and control-persons and how to administer the questionnaire.

We collected stool samples from suspected case-persons and placed the samples in plastic containers. Using swabs, stool samples were placed in Cary–Blair medium, and transported to the Bwera Hospital Laboratory in Kasese. Stool samples were first tested using a cholera RDT kit (Crystal VC™, EIKEN, Japan) and later by stool culture. The RDT was performed according to the instructions that came with the package [2]. Approximately 4–6 drops of liquid stool was transferred to a test-tube using a pipette that comes with the dipstick package. The dipstick was then inserted into the liquid stool, and the results were read after approximately 15 min. The dipstick had a positive test band and a control band. A test was judged as positive if both the test and control bands appeared. If only the control band appeared, it was judged as negative; if the control band did not appear, the test was judged as invalid [2].

To perform microbial analysis of the stool, the swabs were first cultured on Thiosulphate-Citrate-Bile-Salts Sucrose (TCBS™; EIKEN Japan) agar inoculated in alkaline peptone water at 37 degrees Celsius for 18–24 h. Upon identification, the V. cholerae isolates were further evaluated to ascertain the serogroup and serotype by agglutination with Polyvalent O1 and monospecific Ogawa and Inaba antisera [14, 15]. The RDT provided rapid diagnosis and was crucial in the initial phase of the investigation. The culture method provided confirmative diagnosis. Both were performed by trained laboratory personnel.

Water contaminated by sewage has been implicated in several previous cholera outbreaks [16, 17]. Hence we inspected drinking water sources in the most affected areas for possible faecal contamination. The main piped water source in Bwera, the most affected Sub-County, was a GWFS. The scheme started from the top of one of the hills, where the water was chlorinated for consumption. Driven by gravity, the water ran down a pipeline system and was distributed to public and private (household taps) users throughout Bwera Sub-County.

We systematically collected water samples along the GWFS pipeline and in households using sterile containers. Samples were collected from the starting point of the GWFS, and at 2 other points along the pipeline. We also collected samples at 2 public and 2 household water taps and from water containers in 6 randomly selected households. We used the Most Probable Number (MPN) method to determine the presence of faecal coliforms in the water samples, as indicators of sewage contamination [18, 19]. To perform the test, we pipetted 10 ml of water sample into a dilution bottle containing 90 ml of phosphate-buffered dilution water. To prepare a 1/100 dilution, we mixed the 1/10 dilution bottle well and pipetted 10 ml of its contents into a bottle containing 90 ml of dilution water. Subsequent dilutions were then made in a similar way resulting in dilutions from 1/10 to 9/10. The tubes were labelled with the sample reference number and the volume of sample (or dilution) was added to the tube. They were shaken gently and the rack was placed in a water-bath for 48 h at 37 degrees Celsius. After 18–24 h tubes that displayed colour change were regarded as presumptive positive. The number of positive tubes at each dilution was recorded. After 48 h, we prepared tubes of coliform culture medium for confirmation of faecal coliforms. Using a sterile wire loop, we transferred inoculant from each presumptive positive tube into coliform medium tubes. We labelled and incubated these tubes as before. We then noted the tubes that showed growth with the production of gas as positive. Lastly, we compared the pattern of positive results with a most probable number table to obtain the final result for each sample [20].

The outbreak lasted 5 months, from February to July 2015. By 30 June 2015, 183 suspected cases had been identified in the district, including 2 deaths (Fig. 1). Cases occurred in more than 80 villages throughout the district. Most of the cases were from 2 sub-counties: Bwera and Kitswamba. The commonest symptoms were diarrhoea (100%), vomiting (46%), abdominal pain (40%) and fever (2%) (Table 1).

The median age of the case-patients was 18 years (range: 1–90 years). The age group 5–14 was the most affected (attack rate [AR]: 4.1/10,000), followed by the age groups <5 years (AR: 3.7/10,000) and 15–24 years (AR: 2.7/10,000). Males (AR: 2.6/10,000) and females (AR: 2.6/10,000) were equally affected (Table 2).

The outbreak started in villages in Bwera Sub-County near the DRC border, spread eastward through Kasese municipality to Kitswamba Sub-County. In total, more than 80 villages across Kasese District were affected. The highest attack rates were found in parishes within Bwera and Kitswamba sub-counties (Fig. 2). Stratified epidemic curves revealed small clusters of cases occurring in the villages as the outbreak spread eastward.

In the case-control investigation, 34% (17/49) of case-persons compared to 32% (64/201) of control-persons obtained their drinking water from a river or stream (OR_M-H=1.3, 95% CI: 0.57–2.8). However, 94% (46/49) of case-persons compared with 76% (160/201) of control-persons drank water deemed unsafe, i.e., water that was neither treated nor boiled (OR_M-H = 4.8, 95% CI: 1.3–18). None of the other potential exposures, e.g., eating fish, and from which market were the fish usually bought, or eating cold food, were associated with cholera (Table 3).

In total, 78 stool samples collected from suspected case-patients tested positive for cholera by RDT, of which 61 were culture-positive for V. cholerae O1, biotype El Tor serotype Inaba. We did not test all suspected cases because of insufficient supplies of the RDT kits; also, according to WHO guidelines, in resource limited settings like Uganda, it is not necessary to test all suspected cases after the initial cases have been confirmed [1].

Water testing showed that, in samples taken from the water chlorination point of the GWFS, the total coliform count was ≤10 MPN/100 ml [18]. This level meets Uganda’s standard for safe drinking water (≤ 10MPN/100 ml). However, samples collected further along the pipeline from the public water taps in the communities had a total coliform count of 200 MPN/100 ml, while samples from household containers had a total coliform count of ≥800 MPN/100 ml.

Bwera sub-County, the most affected sub-County, borders DRC. High volumes of cross-border trading occur daily. The main water supply for this area is GWFS water from River Rubiriha, which marks the border between Uganda and DRC. Many people collected their drinking water from this river. We inspected the GWFS and found that the water was insufficiently chlorinated for safe drinking. We also found points of leakage and potential contamination along the pipeline. We observed that many families were washing their clothes and kitchen utensils in the rivers and streams. Most homes did not have proper toilet or pit latrine facilities. Many people, especially young children, swam and defecated in the water.

We did not follow WHO guidelines for sampling of environmental water points for testing; instead, we selected the water points purposively [30]. Therefore the water samples tested were not representative of all water points in the area. We did not test the water samples for V. cholerae due to logistical and laboratory-capacity constraints. Also, we did not demonstrate a geographical association between the cases and the area served by the GWFS pipeline because we were unable to obtain the coordinates of the pipeline system. Such data would have strengthened our findings.