DISCUSSION
This study describes the nationwide enteric disease sentinel surveillance system, with an emphasis on cholera in Bangladesh.
We show that cholera is pervasive throughout the country, with substantial heterogeneities within and between geographic areas. Among all Divisions, the sites in Chittagong had the highest proportion of cholera among cases with acute watery diarrhea (AWD) (>12%) followed by Dhaka and Barisal (6.5%). The highest cholera positivity (>18%) was documented in the BITID site in the Chittagong Division, although this may be due to the fact that BITID is a well-known referral hospital in the area (Table 1, Figure 1).
In the Dhaka Division, 9% of patients with diarrhea in the Narayanganj District were V. cholerae positive every year during the study period. Of note, the icddr,b diarrheal hospital, a renowned cholera hospital with over 60 years of activity, is well known to people living in Dhaka. On average, up to 22% of hospitalized patients with diarrhea are annually found to be cholera positive at the icddr,b hospital \citep{Andrews2017}. Given that the icddr,b hospital attracts people from far away and is also a referral hospital for diarrhea, we did not include this site in our surveillance network due to potential biases in extrapolating findings to the general population. The presence of this hospital may have led to underestimates of the prevalence of cholera among diarrhea cases in other clinics in Dhaka. However, based on the Diarrheal Disease Surveillance System data of the icddr,b, Dhaka Division can be extrapolated to be the area with the highest rates of cholera. In the Barisal Division, the highest burden was observed in Bakerganj, a subdistrict, which is supported by previous studies conducted by the icddr,b in the same area earlier \citep{barua1992}.
Analysis of the demographic characteristics of patients showed that the case distribution was approximately equal by sex, which is comparable to other studies conducted in Bangladesh as well as other cholera-endemic countries \citep{Sarker2013}\citep{Deepthi2013}. Children less than 5 years old, and especially those less than 2 years old (n = 183; 1.8%), had a higher proportion of noncholera diarrheal episodes (the majority may have been be due to rotavirus) than those in older age groups (Table 1, Supplementary Table 1). This is consistent with findings from previous studies \citep{Desai2014}\citep{Khan2018}. In an earlier analysis carried out at the icddr,b diarrheal hospital in Dhaka\citep{Qadri2015} and other surveillance sites in Bangladesh, the diarrheal rates were found to be higher in older children and adults than in the younger age group \citep{Sack2003}\citep{Chowdhury2011}. Our observations also show a lower prevalence of cholera in children under 2 years of age, although 40% of cases of diarrhea are seen in this age group. This observation also lends support to the GTFCC 2017 recommendations that cholera should be suspected among persons aged 2 years and older, with AWD and severe dehydration or dying of AWD in areas where a cholera outbreak has not been declared.
Our sentinel surveillance shows that cholera is endemic in Bangladesh, with a distinct seasonality across the country. We observed different seasonality for cholera in the sites, especially between the eastern and western regions of Bangladesh. This observation is similar to those in previously conducted studies reported from Bangladesh \citep{Paul2016}\citep{Alam2011}. This variation may be related to local ecology, such as the location of the major rivers, which requires further evaluation.
The primary catchment area was approximately 50 km from each surveillance site, where 80% of the patients with diarrhea resided. Cases with severe cholera traveled farther distances seeking treatment compared with non–severe cholera cases and other patients with severe diarrhea; only 55% of cases with severe cholera traveled within 50 km of a hospital site. It is important to note that our observation is limited by the fact that the majority of sentinel sites are referral hospitals in the area, and we cannot exclude the possibility that patients were referred to a surveillance hospital by the closest treatment facility. Healthcare-seeking behavior for other severe illnesses, however, shows that patients travel far distances to attend a healthcare facility \citep{Hegde2019}.
Due to funding constraints, the study was limited to 22 sentinel sites, which may not fully reflect the diversity of cholera epidemiology within the country. Establishing a higher-density surveillance network at all levels of the healthcare system, and across areas that we expect to have both high and low cholera incidence, may allow for more detailed insights. Another major limitation was that the data were from a facility-based surveillance system for which the exact catchment area (eg, the denominator) and the sensitivity of detecting all cholera cases in the catchment area are unknown. In future expansions of this surveillance system, healthcare utilization surveys and possibly cross-sectional serological surveys may help improve our ability to estimate the true incidence of cholera disease and V. cholerae O1 infection \citep{Azman2019}.
Furthermore, the patients enrolled were not followed up after treatment or discharge; therefore, data on clinical consequences and mortality are lacking. Finally, for the confirmation of cholera cases, this study used a conventional culture method and not PCR, which may improve the sensitivity, especially in the presence of antibiotics \citep{Alam2010}\citep{Faruque2006}.
This nationwide hospital-based surveillance shows the presence of cholera in all geographical regions in Bangladesh that were under surveillance. Although describing cholera epidemiology is complex in Bangladesh, we show that different frequencies of disease exist across this small geographical area. Our study identified at least 8 geographical areas (health facilities of district and subdistrict levels) where cholera was consistently higher over the reporting period. Dhaka remains as one of the areas with a moderate-to-high prevalence of cholera among cases of diarrhea in accordance with data from the icddr,b systematic surveillance system (where 2% of all diarrhea cases are tested for V. cholerae and other pathogens) \citep{Alam2010}. Given the population size, the absolute impact of controlling cholera in Dhaka would be substantial, both within the city and likely elsewhere in the country.
One of the most immediate ways to protect populations against cholera is to provide OCV. Given our findings, large OCV campaigns are justified in at least a subset of the surveillance areas. While there is no well-defined threshold of cholera incidence or prevalence (among diarrhea cases), targeting areas with a high prevalence may be a reasonable place to start. Based on the data, it would be judicious to plan for OCV campaigns around Comilla, BITID, Cox’s Bazar, and Narayanganj, as well as Bakerganj. Since Dhaka city itself has a high burden of cholera based on icddr,b cholera surveillance data, it should also be targeted for OCV rollout to control epidemics of cholera.
The study provides critical insights for the Bangladesh National Cholera Control Plan and points towards key geographic areas within the country where cholera- prevention and -control activities, including vaccination, should be prioritized. For long-term control of cholera, massive investments in sustainable water and sanitation infrastructure are needed, although universal access may be years, if not decades, away. Continued, and expanded, national disease surveillance will be critical in the years to come to monitor progress on the road to elimination and to quantify the impact of interventions like OCV. The multisectoral support of different ministries of the government of Bangladesh and international partners for the improvement of water and sanitation measures, strengthening the national health systems, and targeted use of OCV will be critical to meeting the WHO-backed goal of ending cholera by 2030 in Bangladesh.