Cholera Incidence and Mortality in Sub-Saharan African Sites during Multi-country Surveillance

Background Cholera burden in Africa remains unknown, often because of weak national surveillance systems. We analyzed data from the African Cholera Surveillance Network (www.africhol.org). Methods/ Principal findings During June 2011–December 2013, we conducted enhanced surveillance in seven zones and four outbreak sites in Togo, the Democratic Republic of Congo (DRC), Guinea, Uganda, Mozambique and Cote d’Ivoire. All health facilities treating cholera cases were included. Cholera incidences were calculated using culture-confirmed cholera cases and culture-confirmed cholera cases corrected for lack of culture testing usually due to overwhelmed health systems and imperfect test sensitivity. Of 13,377 reported suspected cases, 34% occurred in Conakry, Guinea, 47% in Goma, DRC, and 19% in the remaining sites. From 0–40% of suspected cases were aged under five years and from 0.3–86% had rice water stools. Within surveillance zones, 0–37% of suspected cases had confirmed cholera compared to 27–38% during outbreaks. Annual confirmed incidence per 10,000 population was <0.5 in surveillance zones, except Goma where it was 4.6. Goma and Conakry had corrected incidences of 20.2 and 5.8 respectively, while the other zones a median of 0.3. During outbreaks, corrected incidence varied from 2.6 to 13.0. Case fatality ratios ranged from 0–10% (median, 1%) by country. Conclusions/Significance Across different African epidemiological contexts, substantial variation occurred in cholera incidence, age distribution, clinical presentation, culture confirmation, and testing frequency. These results can help guide preventive activities, including vaccine use.


Introduction
Background/rationale 2 Explain the scientific background and rationale for the investigation being reported

[page 4 ]
From 2007 to 2012, at least 20 African countries reported more than 100,000 cases of cholera (World Health Organization (WHO) weekly epidemiological records, [2007][2008][2009][2010][2011][2012]. However the surveillance of cholera has weaknesses. Reporting is nonexhaustive for various reasons such as individual and community fears of stigmatization and economic loss. Reporting from district to national levels may be delayed or incomplete. According to WHO, only 3% to 5% of all cases are laboratory confirmed. A variety of case definitions are used across countries, which could lead to cholera over or under-reporting. Finally, few countries have implemented case-based surveillance, with information at national level provided in the form of weekly summaries limited to cumulative case numbers and deaths.
Objectives 3 State specific objectives, including any prespecified hypotheses

[page 4 ]
We present cholera incidence results and the associated case fatality ratio from eleven geographical zones located in six Africhol countries having the strongest performing surveillance systems.

Study design 4
Present key elements of study design early in the paper

[page 4 ]
We conducted a population -based cholera surveillance in all cholera treatment facilities in given geographic zones.
Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection

[pages 5-6]
We included eight enhanced surveillance zones located in areas of known recent cholera occurrence were included in the analysis. Their location and starting dates were as follows: 1) Togo: five districts of Lome and Golfe district, Jun 2011; 2) Togo: Lake district in the Maritime region, Jun 2011; 3) Democratic Republic of Congo Cohort study-Give the eligibility criteria, and the sources and methods of selection of participants. Describe methods of follow-up.

[page 5 ]
Patients were followed in all the cholera treatment facilities of a given surveillance area. In area without known ongoing cholera, a suspected cholera case was defined as a patient aged two years or more that developed severe dehydration or died from acute watery diarrhea. In areas with known cholera, a suspected case was defined as a patient aged two years or more that developed acute watery diarrhea, with or without vomiting. A confirmed case was defined as a suspect cholera having a stool culture positive for Vibrio cholera.
Case-control study-Give the eligibility criteria, and the sources and methods of case ascertainment and control selection. We presented the proportions of suspected and confirmed cholera cases by age group, sex, occurrence during the rainy season and clinical symptoms. We calculated the crude incidence rate of cholera and corrected incidence rates for confirmed cases. We also presented case fatality ratios for suspected and confirmed cholera.
Data sources/ measurement 8* For each variable of interest, give sources of data and details of methods of assessment (measurement). Describe comparability of assessment methods if there is more than one group

[pages 6-7]
In the enhanced surveillance zones and outbreak sites, the MoH teams collected data at health centers level using the same standardized data collection forms, which included sex, age, location, date of symptoms, culture results but also clinical information such as watery diarrhea, rice water stool, vomiting, dehydration. We identified all deaths among patients admitted to a cholera treatment facility. We did not include deaths occurring in the community or after treatment center discharge. In parallel, the MoH continued to register the overall number of suspected cases in their routine surveillance system using line lists with a limited number of variables (date of onset, district, age and sex). We used district-level population estimates for 2011 or 2012 that corresponded to the geographic area under surveillance. The 2011 and 2012 population estimates were derived from the last census data (Uganda, 2002;DRC, 1983;Togo, 2009;Guinea, 1996;Cote d'Ivoire, 1998;Mozambique, 2007), updated each year by district health officers based on estimated national annual population growth rates. We adopted the definition of rainy season from the World Bank climate portal (sdwebx.worldbank.org/climateportal; accessed 2013) as follows: Uganda, Mar-Jun and Sept-Nov; Goma, DRC, Jan-May and Sept-Dec; Kinshasa, DRC, Jan-May and Oct-December; Mozambique, Oct-Mar; Cote d'Ivoire, May-Jun and Oct-Nov; Guinea (Maritime region), May-Nov; Togo (Maritime region), Apr-Jul and Sept-Nov.
Bias 9 Describe any efforts to address potential sources of bias

[page 7]
We calculated corrected incidence rates taking into account the lack of culture testing in each given surveillance zone and the lack of sensitivity of culture test. We tested 26% (3536) of all suspected cases by culture (table 1), a figure that increased to 49% when excluding zones in Goma and Conakry, which both experienced large outbreaks in August 2012 and which respectively had testing only 7.4% and 0.5% of cases during this period (Figure 1).

(c) Cohort study-Summarise follow-up time (eg, average and total amount) [see table 4]
Outcome data 15* Cohort study-Report numbers of outcome events or summary measures over time [pages 12-13-14, tables 4 and 5 and figure 1] Annual confirmed incidence of cholera presenting to a treatment facility per 10,000 population was <0.5 in surveillance zones, except in Goma where it was 4.6. Goma and Conakry had corrected incidences of 20.2 and 5.8 respectively, while the remaining surveillance zones had a median corrected incidence of 0.3. During outbreaks, the annualized confirmed incidence of cholera presenting to a treatment facility ranged from 0.3-3.3 and corrected incidence from 2.6 to 13.0 per 10,000 population. The ratio of the mean annual corrected incidence of confirmed cholera to the incidence of suspected cholera varied from 0.1 in Abidjan to 0.6 in Conakry while it was of 0.5 (SD 0.1) in outbreak sites. Of 5980 suspected cases identified in a treatment facility with a documented outcome, 69 died. The median CFR was 1.1% [IQR: 0.7-4.3]. The CFR varied from zero percent in Abidjan, Cote d'Ivoire to 10% in Lake district, Togo. We found no statistical differences in the CFR between confirmed and non-confirmed cases. However we observed that deceased patients were less likely to have received culture testing than those alive at discharge (35.3% vs. 55.6%, chi-square p. value= 0.001). Other analyses 17 Report other analyses done-eg analyses of subgroups and interactions, and sensitivity analyses [page 12, figure 1 ] We identified three epidemiological patterns. In surveillance zones in Goma (DRC), confirmed cases were seen continuously throughout the surveillance period. In zones in Lome (Togo), Mbale (Uganda) and Conakry (Guinea), there were sporadic confirmed cases plus additional outbreaks at irregular intervals. Lastly, in Beira, Mozambique and Abidjan, Cote d'Ivoire, there was a history of recurrent cholera epidemics in the period leading up to Africhol implementation but as of the end of 2013, no confirmed cases had been identified for 30 months and 17 months, respectively.

Discussion
Key results 18 Summarise key results with reference to study objectives

[page 14 ]
In the Africhol surveillance zones, we found an overall annual corrected incidence of confirmed cholera presenting to a treatment facility of 0.3 cases per 10,000 population, which increased to 20 cases per 10,000 during large epidemics. Strong spatial and temporal clustering occurred, with most cases from surveillance zones in Conakry, Guinea and Goma, DRC. Within our study many suspected cases were not cholera confirmed by culture. Furthermore the CRF measured at clinic level remained low in our surveillance sites. From the surveillance data collected in our sites, we were able to identify three epidemiological patterns of cholera: confirmed cases throughout the year such as Goma (DRC); sporadic cases plus additional outbreaks at irregular intervals such as in Lome (Togo), Mbale (Uganda), and Conakry (Guinea); and history of recurrent cholera epidemics but no cases during the surveillance period, such as Beira (Mozambique) or Abidjan (Cote d'Ivoire). Whatever the location, we found that most cholera cases occurred during the rainy season.
Limitations 19 Discuss limitations of the study, taking into account sources of potential bias or imprecision. Discuss both direction and magnitude of any potential bias

[pages 15 and 16 ]
The wide variation of culture confirmed cases may have resulted from differences in health care seeking behavior, health care access, type and extent of available health structures, health work training, and adherence to case definitions. For instance, treatment centers in Goma, DRC provided care for patients with any diarrheal disease regardless of etiology, did not charge fees, and treated persons of all ages. In other Africhol sites, cholera treatment centers offering free treatment were established only when authorities declared the outbreak. These issues also may have led to the differences in health care access behaviors and therefore to clinical presentation across sites. Other factors may lead to underestimation of incidence. For example, not all patients will present for care at a medical facility and data collection and reporting may be incomplete. However, our system was not designed to assess these issues.
Our CFR estimates were limited by our inability to assess deaths in the community which contribute to potential underestimation. Both our CFRs and overall incidence rates were limited by lack of active community-based surveillance, an objective for which our work was not funded. It is likely that this problem was particularly large for deaths: for example, a study from Kenya found that most deaths occurred among persons who had not sought treatment. Future geographically focused studies might address this issue. In theory, health utilization surveys and capture-recapture analysis could help with estimation of surveillance system sensitivity. However, in epidemic cholera prone settings in Africa, health care utilization surveys are seldom appropriate given the lack of human resources relative to the immediate priority of outbreak control. Capture-recapture analyses similarly are not feasible, given the fluid nature of a surveillance system in which cholera treatment centers are established and dismantled relative to cholera case counts.
Our study had several limitations other than those mentioned above. We report data from only eleven geographical sites located in six countries and this may not be generalizable to other African settings. Our correction of incidence based on the lack of testing was applied uniformly across the surveillance period without taking into account seasonal variations. We used a single value to correct for culture sensitivity although culture results may vary by setting based on factors such as laboratory technician skills and stool collection and transportation methods. Finally, CFRs were difficult to assess for confirmed cholera cases because of lack of testing.
Interpretation 20 Give a cautious overall interpretation of results considering objectives, limitations, multiplicity of analyses, results from similar studies, and other relevant evidence

[pages 15-16 ]
Our incidence estimates for confirmed cases showed similar fluctuations by place and time as those reported previously for suspected cases but are substantially lower than estimates modeled from WHO mortality strata. In most national cholera surveillance systems, etiologic confirmation occurs only for the first suspected cases, before outbreak declaration. Subsequently, any person with acute watery diarrhea usually would be reported as a cholera case, even though some of these will have other etiologies. Consequently, syndromic surveillance -as reported by most previous studies -likely overestimates cholera incidence. Moreover, the proportion of culture confirmed cases varied widely by site emphasizing the utility of laboratory based studies. At the extreme, in Beira, Mozambique, where a history of large outbreaks likely led providers to have a high index of suspicion for cholera, all sampled suspected cases remained negative for V. cholera [11]. While our incidence rates were lower than those from early reports, CFRs for confirmed cholera cases were consistent with those for suspect cases attending health facilities [5,11].
The low identified CFRs emphasize the great strides some cholera endemic countries have made in identifying outbreaks rapidly and improving clinical management. They might also reflect the sensitization of populations in high-risk areas to the importance of seeking timely medical care. We identified three epidemiological patterns of cholera in our sites: those with confirmed cases throughout the year such as Goma (DRC); those with sporadic cases plus additional outbreaks at irregular intervals such as in Lome (Togo), Mbale (Uganda), and Conakry (Guinea); and those with a history of recurrent cholera epidemics but no cases during the surveillance period, such as Beira (Mozambique) or Abidjan (Cote d'Ivoire). The presence of sporadic cases without ensuing outbreaks may occur from occasional introduction of infected persons into a low risk community, e.g., a community with recent cholera and a high degree of population immunity or a community with good water and sanitation infrastructure. By contrast, sustained occurrence of confirmed cases may result from ongoing environmental source contamination from which a continuously renewed susceptible, non-immune population is infected; this may have occurred in Goma, which has experienced several waves of immigration due to regional conflicts. We found that most cholera cases occurred during the rainy season. However the presence of cases before the rain start suggests that the rainy season may play a role of outbreak amplificatory. Previous studies have found similar results [16]. Substantial precipitation can cause flooding and subsequent mixing of drinking water (pond, well, lake, river) with sewage in areas with poor sanitation [17]. Alternatively, the rainy season may trigger human movement, such as the seasonal migration of fishermen along the West African coast or in interior lakes [16,[18][19][20].
Generalisability 21 Discuss the generalisability (external validity) of the study results

[pages 17-18]
While limited to health care facilities, our study presents some of the only prospectively obtained incidence data currently available for Africa. Our findings suggest that confirmed cholera burden is substantially lower than that reported from previous studies based on suspected cholera cases, and that incidence varies substantially over time and place. Efficient use of resources, such as vaccines, could be enhanced by better definition of cholera hot-spots, community behaviors that contribute to cholera spread, and high risk populations, particularly those likely to contribute to seasonal cholera spread. Because of the frequent occurrence of non-cholera causes of diarrhea in cholera endemic zones, development of public health strategies would benefit from reinforcement of local laboratory capacities for diagnosing Vibrio cholerae, something that also would benefit from development of better low-cost diagnostic methods. Environmental reservoirs should be identified and mitigation strategies developed. Determination of other diarrheal disease etiologies across all age groups will help determine the utility of etiology specific interventions. OCV interventions must be conducted, monitored and evaluated to better assess their cost-effectiveness and their health impact among at-risk populations in African contexts. Finally, there is a role for evaluation of low-cost water and sanitation improvements within an integrated strategy for cholera prevention and control.

Other information
Funding 22 Give the source of funding and the role of the funders for the present study and, if applicable,