This study was conducted as part of a campaign to control and eliminate cholera transmission in 2 communities in Haiti. The first, Cerca-la-Source, is a rural, mountainous community of ≈50,000 persons. The second, Mirebalais, is an urban commune of ≈100,000 persons. Both communities are located in the Centre Department of Haiti, a historically underserved and particularly impoverished region of the country.

We conducted a census of both communities. During the census, a subset of households was invited to participate in a household survey and a serologic survey at fixed sampling intervals during March–August 2017 (Appendix Table 1). Trained study enumerators implemented study procedures in their native language of Haitian Creole; the procedures included surveys to measure self-reported sociodemographic and cholera risk factors.

We obtained dried blood spots from consenting adults >18 years of age and shipped them to a laboratory in Boston, Massachusetts, USA, where we performed vibriocidal assays by using a drop-plate method from dried blood spots specimens, as described previously (10), except we used Advance Dx100 Serum Separator cards (Advance Dx, Inc., https://adx100.com) instead of the cards used in that study. We used target V. cholerae strains 19479 El Tor Inaba and X25049 El Tor Ogawa.

To ensure estimates were representative of the populations of Mirebalais and Cerca-la-Source, we used a raking procedure to apply survey weights on the basis of the population distribution of age, sex, and communal sections from the census in those regions. We used a random intercept to account for clustering by household. The primary outcome was the overall seroprevalence (either Ogawa or Inaba) of vibriocidal antibody responses against V. cholerae for each community. We defined seropositivity as a vibriocidal antibody response titer threshold of >320 on the basis of the best available evidence, a recent study in Bangladesh that estimated that a vibriocidal modal titer of 320 had a sensitivity of 80.6% and specificity of 83.0% for infections within the preceding year (9). We also calculated serotype-specific seroprevalence estimates for each region. For potential risk factors for seropositivity, we provided descriptive statistics, weighted seroprevalence estimates, and 95% CIs (for categorical variables) and odds ratios (ORs) with 95% CIs. We calculated ORs by using univariable logistic regression followed by multivariable logistic regression analysis, including only those risk factors associated with cholera at a significance level of p<0.20 in univariable analysis. We conducted analyses using the survey package in R 4.2.2 (The R Project for Statistical Computing, https://cran.r-project.org) (11).

The study was approved by the Partners Healthcare Institutional Review Board (protocol 2016P002781) and the Zanmi Lasante Institutional Review Board (protocol ZL IRB ID AK). All study participants provided written informed consent.

Overall, we enrolled 265 (27.6%) of 960 invited households in the study. Samples from 48 households were lost during tumultuous sociopolitical events, resulting in samples from 217 households available for analysis: 99 households with 156 persons in Cerca-la-Source and 118 households with 121 persons in Mirebalais.

Figure

Figure. Serosurvey participants with vibriocidal antibody titers for Ogawa (A) and Inaba (B) Vibrio choleraeserotypes in 2 communities, Centre Department, Haiti, March–August 2017. Samples came from 217 total households,...

We analyzed unweighted demographic characteristics for the census population and for serosurvey participants (Table 1). Serosurvey participants were representative of the census population. The weighted seroprevalence of V. cholerae was 12.4% (95% CI 6.76%–20.0%) in Cerca-la-Source and 9.54% (95% CI 4.91%–16.0%) in Mirebalais (Table 2). We analyzed the frequency distribution of vibriocidal antibody titers for both serotypes (Figure). Only 4 of 277 persons reported having received oral cholera vaccine, consistent with the fact that no major public health oral cholera vaccine campaign had been undertaken in those regions before sample collection.

We calculated seroprevalence estimates for potential risk factors for cholera (Appendix Table 2). Seropositivity varied across multiple subgroups; however, 95% CIs were wide. Only the poverty likelihood index (OR 2.33, 95% CI 0.93–5.84) and reporting having an unimproved toilet compared with open defecation (OR 0.26, 95% CI 0.04–1.53) met our predetermined p value threshold for inclusion into a multivariable model, so we did not perform multivariable analysis.

The vibriocidal antibody is a complement-dependent, bactericidal antibody directed against the lipopolysaccharide O-antigen of V. cholerae and is the best characterized immunologic marker of recent exposure to cholera. However, there is no widely agreed-upon threshold to quantify exposure over a given period, and our understanding of the relationship between symptom severity and antibody kinetics is limited. In the study from Bangladesh, a vibriocidal titer of >320 was the best marker of infection in the preceding year (9).

Limited serologic data on V. cholerae are available from Haiti. One prior serosurvey, conducted during March–April 2011, within the first 6 months of the onset of the epidemic in the Artibonite Department of Haiti, estimated that 39% of persons had a vibriocidal titer >320, whereas 64% had titers >80, which suggested extensive infection and was consistent with high early case counts (12).

The findings from this study should be interpreted considering several limitations. Only adults >18 years of age in 1 department of Haiti participated, so the data cannot be directly extrapolated to younger age groups and other regions; however; during 2017–2018, that department was the most affected according to case counts (13). We were unable to account for uncertainty in vibriocidal assay performance characteristics. Ideally, seroprevalence estimates should integrate data on the local sensitivity and specificity of a serologic assay, which are not available for Haiti (14). Last, the survey was cross-sectional and did not account for temporal waning of serologic markers.

In summary, in 2017, the seroprevalence of V. cholerae vibriocidal antibodies was 12.4% in Cerca-la-Source and 9.54% in Mirebalais in Haiti, suggesting a high rate of recent infection even at a time when case incidence was declining. Although commune-level incidence data were not available for direct comparison, in 2017, the reported annual incidence for the Centre Department, where Cerca-la-Source and Mirebalais are located, was 4.3 cases/1,000 inhabitants, which offers a general frame of reference (13,15). Those findings inform our understanding of cholera epidemic dynamics in Haiti, which is now experiencing a resurgence of cholera after nearly 3 years without a confirmed case. Our results demonstrate a higher-than-expected disease prevalence and suggest the need for improved surveillance to monitor cholera and guide public health interventions, especially during the waning phase of outbreaks.