During March 22–April 6, 2011, we conducted a cross-sectional seroepidemiologic survey in Grande Saline Commune. All consenting residents ≥ 2 years old from selected households were eligible for enrollment. After accounting for non-participation and design effect, our target sample was 600 households, based on assumptions of four consenting persons per household (for 2,400 participants) and a 75% V. cholerae infection rate. The true infection rate was unknown; 75% was chosen for these calculations because the high reported cholera attack rate in the commune suggested widespread exposure. This sample size would estimate the proportion of infected persons who develop severe disease to within 4%, assuming 5% of seropositive participants developed severe disease (given previous estimates of 2–11%).5–7 We used a type I error rate of 5% and power of 80% for sample size calculations.

The sampling frame included all households within the commune located in the commune's 13 villages (Figure 1). Local authorities reported that nearly all commune residents resided within these villages, which are all located along the commune's main road. We selected households within each village in proportion to population estimates from local authorities because current census data were not available at the municipal level. Local community health workers identified subsections of each village. Study teams enrolled households within all subsections by starting at local landmarks within each section and moving sequentially from household to household. We used satellite maps of villages to minimize exclusion of household groups.

Study teams interviewed participants by using a Visual CE (SYWARE, Cambridge, MA) electronic questionnaire on a personal digital assistant. The survey included questions about demographic and household characteristics, reported cholera diagnosis by a health care provider, diarrheal symptoms and treatment, antacid use, and deaths caused by cholera within the household since the start of the outbreak in October 2010. Within each household, we asked the first participant encountered (or parent or guardian for children less than 12 years old) about household characteristics (e.g., number of persons in household and primary water source) and individual characteristics, and asked subsequent participants only about individual characteristics. We used a household identification number to link records within a household. All information was self-reported (or for children < 12 years old, reported by a parent or guardian), including clinical information; detailed clinical records were not available. We collected a single blood specimen from consenting participants as described below.

Institutional Review Boards from the Haitian National Ethics Committee (Protocol #2011-FWD-CHOLERA-01) and the U.S. Centers for Disease Control and Prevention (CDC Protocol #6038) granted approval for this study, including for testing of control serum specimens. Haitian interviewers provided participants with oral and written explanations of the study. Participants marked checkboxes to indicate informed consent (adults ≥ 18 years old), parent or guardian permission (children < 18 years old), and child assent (children 7–17 years old). No personal identifiers were collected.

Blood specimens were collected in volumes consistent with participants' age and weight (maximum = 10 mL) in Vacutainer (Tiger Red/Black Top; Becton-Dickinson Laboratories, Franklin Lakes, NJ) serum separator tubes and transported on cold packs to the local health center in Grand Saline Commune. At a temporary serum processing laboratory in the health center, serum was separated after centrifuging the blood specimens, labeled, and stored on cold packs until the end of each workday. Serum specimens were transferred to a refrigerator at the Hôpital Albert Schweitzer, Haiti, and were transported twice a week on cold packs to the National Public Health Laboratory in Port-au-Prince. They were then held refrigerated until shipment to CDC (Atlanta, GA). At CDC, specimens were inventoried and accessioned, and each specimen was divided into 0.5-mL aliquots and frozen at −80°C. Fifty unlinked control serum specimens collected in Haiti before the cholera outbreak were included to determine pre-2010 antibody levels in this population.

Two V. cholerae strains from Haiti, 2010EL-1786 (serotype Ogawa) and 2012EL-1407 (serotype Inaba), were used as target antigens in the vibriocidal assays. The complete genome sequence of 2010EL-1786 has been published.27 The Inaba strain 2012EL-1407 emerged in Haiti in March of 2012.28

Preliminary bacterial growth curves aided in standardizing the inoculum, time of incubation, and growth medium necessary to achieve a target inoculum of 3.0×10⁷ colony-forming units/mL of live bacteria in the early log phase of growth. The Ogawa isolate was grown to early log phase in BactoTM Heart Infusion Broth (Becton Dickinson and Company, Sparks, MD), divided into aliquots, and frozen at −80°C. From pilot study results, we concluded that 3 hours of incubation at 37°C of a standard inoculum from the frozen stock yielded 3.0×10⁷ colony-forming units/mL. Thereafter, on each assay occasion we determined the final concentration of live bacteria. If the counts did not fall within 1 SD of the average of all accepted runs, we rejected the previous day's assay. We repeated this process for the Inaba serotype.

Three lots of commercial guinea pig complement (Innovative Research, Novi, MI) were tested in pilot studies. The best performing lot was selected, and all serotype-specific assays were run using the same lot. Pilot studies of complement activity aided in standardizing the assay to assure addition of excess complement. Controls were run to ensure that commercial complement did not contain naturally occurring vibriocidal activity.

Known positive serum specimens obtained from patients with confirmed cholera infections and known negative specimens were run on each test occurrence for each assay. Controls were unlinked from any patient identifiers.

Both serotype-specific assays were validated with known positive and negative serum specimens and a positive cutoff was established by receiver operating characteristic analysis. The assays in a 96-well microtiter format were set up accordingly: two-fold dilutions of heat-inactivated control and test serum samplers were titrated by using a BioMek NX^P Laboratory Automation Workstation (Beckman Coulter, Brea, CA). After an initial 1:10 dilution of each serum, 11 two-fold dilutions were made to a final concentration of 1:40,560. Every serum specimen was run in duplicate. All specimens were tested with serotype Ogawa and repeated with serotype Inaba.

For each assay, the appropriate V. cholerae strain was cultured at 37°C to early log phase and washed twice. Cultures and complement were mixed, added to each serum dilution, incubated for 1 hour at 37°C, then overlayed with warmed BactoTM Heart Infusion Broth. After 2.5 hours at 37°C, wells were read visually for turbidity indicating growth of bacteria. The titer was expressed as the reciprocal of the highest dilution of serum with > 50% reduction in turbidity. If positive controls results did not fall within one dilution ± 1:640, or if the negative control was positive at any dilution, the assay was rejected and re-tested. In addition, if plate counts did not fall within 1 SD of the average of previously validated specimens, the corresponding assay was rejected and re-tested.

We tested serum specimens in an indirect enzyme-linked immunosorbent assay (ELISA) format using Immulon 2HB (Thermo Scientific, Rochester, NY) microtiter plates coated with commercially available CT (Sigma, St. Louis, MO). Individual serum specimens were diluted 1:50 in phosphate-buffered saline with 1% bovine serum albumin and diluted two-fold to 1:1600, followed by the addition of mouse anti-human IgG monoclonal antibody HP604329 or commercial goat anti-human IgA (Invitrogen, Frederick, MD) conjugated to alkaline phosphatase. After incubation for 1 hour at 37°C, we added disodium p-nitrophenyl phosphate (Sigma), stopped the reaction with 3M NaOH, and read optical densities (ODs) at 405 nm with a Spectramax Plus 384 device (Molecular Devices, Sunnydale, CA).

We optimized antigen, serum, and conjugates by checkerboard titrations before testing and validated ELISA results by using positive and negative serum from patients with and without cholera, respectively. Statistical thresholds were calculated by using receiver operating characteristic analysis for each assay and antibody class.

Since red cells were not available, we determined ABO blood group for a subset of participants by using the reverse-typing test tube method with A1 and B Referencells (Immucor, Norcross, GA) as specified by the manufacturer. Anti-A and Anti-B murine monoclonal antibodies (Immucor) were tested as controls. A limitation of this approach is that it does not directly detect ABO blood groups but rather depends on the presence of anti-A and anti-B antibodies in serum.

Survey data were downloaded from personal digital assistants to a Microsoft (Redmond, WA) Access 2003 database and analyzed by using SAS version 9.3 (SAS Institute, Cary, NC).

Seropositivity was determined by measuring vibriocidal antibody responses to Ogawa and Inaba serotypes from isolates obtained from Haiti. Secondary analyses examined anti-CT IgA and anti-CT IgG results. We defined clinical severity of disease by level of treatment required, similar to definitions from previous bacteriological surveys.5,6 All clinical data were self-reported. Participants who reported a cholera diagnosis but not watery diarrhea (i.e., logically inconsistent responses) were excluded from analyses describing cholera and disease severity.

Severe disease was defined as diarrhea with overnight hospitalization at a health facility and receipt of IV fluids, moderate disease as diarrhea and a visit to a health facility but not meeting criteria for severe disease, and mild disease as diarrhea without a health facility visit.

We calculated basic frequencies, as well as odds ratios (ORs) and 95% confidence intervals (CIs) by using simple logistic regression adjusted for household design effect. We used proportional-odds cumulative logistic regression models to assess influence of potential risk factors on disease severity (classified on a four-level ordinal scale: severe, moderate, mild, or asymptomatic), adjusting for age quartile, and antacid use. Predictors were blood group and socioeconomic characteristics (literacy, speaking French, and household possession of a cell phone, radio, or electricity). For each model, we calculated the score test to assess whether the proportional odds assumption was contradicted.

During the study period, we enrolled 2,622 Grande Saline residents ≥ 2 years old from 1,240 households. There were an average of 4.4 (SD = 2.5) eligible persons per household of whom we enrolled a mean of 2.1 participants (range = 1–10). Enrollment targets were met in each of the commune's 13 villages. We collected serum from 2,527 (96%) participants.

The median age of participants was 23 years (range = 2–90 years), and 59% were female (Table 1). The median age for males (18 years) was less than the median age for females (26 years); among adults ≥ 18 years old, 1,013 (65%) were women, and 544 (35%) were men. Among all participants, 1,394 (53%) were literate, 300 (11%) spoke French, and 257 (10%) had attended at least some secondary school. Among 1,557 adults ≥ 18 years old, 729 (47%) were literate, 220 (14%) spoke French, and 219 (14%) attended at least some secondary school. The most common occupations among adults were rice farming (565, 36%), farming other crops (343, 22%), and vending or trading (307, 20%).

Among the 1,240 households in the study, 603 (49%) owned a cellular phone, 434 (35%) owned a radio, and 52 (4%) had electricity (including from generators). Regarding primary drinking water source, 472 (38%) households bought water from vendors, 398 (32%) households used a canal or the Artibonite River, 180 (15%) used a public or private tap, and the remaining 190 (15%) used other sources, including wells. Of 962 (78%) households that reported having treated their drinking water since the outbreak began, 953 (99%) used at least one of three treatment methods: 513 (53%) used chlorine tablets, 246 (26%) used filters, 91 (9%) used chlorine solution, and 112 (12%) used a combination of these methods. No additional information was collected on alternate sources of drinking water or the regularity of drinking water treatment.

Eleven (< 1%) participants reported having been diagnosed with HIV infection, 77 (7%) females 10–55 years old reported being pregnant, and 418 (16%) participants reported taking antacids in the previous six months.

We tested a subset of 866 participants for ABO blood groups by reverse typing; 386 (45%) had blood group O, 242 (28%) had group A, 194 (22%) had group B, and 44 (5%) had group AB.

Among all participants, 541 (21%) reported an episode of watery diarrhea since the start of the cholera outbreak in October 2010, including 466 (18%) who reported a cholera diagnosis from a health care worker (Table 1). Seventy-nine (3%) participants reported a cholera diagnosis but no watery diarrhea and were excluded from cholera and disease severity analyses. Among the 541 participants reporting watery diarrhea, 83% reported feces that resembled water used to cook rice (rice water stool), 71% reported painful leg cramps, 43% reported vomiting more than once, and 11% reported bloody stool (Table 2). Participants with watery diarrhea who reported a cholera diagnosis were more than twice as likely as participants with watery diarrhea without a cholera diagnosis to report each of these symptoms except for bloody stool, for which the proportions were similar (11% versus 9%, respectively). Participants with a cholera diagnosis reported a median of seven stools on the worst day of illness compared with three for those not reporting a cholera diagnosis. Most participants reported a date of diarrhea onset during October 15–November 31, 2010 (Figure 2A). This distribution of illness over time was similar to the distribution of cholera cases reported to Haiti's national cholera surveillance system by local health care providers in Grande Saline (Figure 2B Figure 2.

A, Self-reported dates of watery diarrhea onset for participants reporting a cholera diagnosis, cholera serosurvey, Grande Saline, Haiti, 2010–2011 (n = 354). B, Grande Saline cholera cases reported to Ministry of Health and Population (MSPP) surveillance, Haiti, 2010–2011.

A higher proportion of participants in the oldest quartile (> 43 years old) reported a cholera diagnosis (23%) than participants in younger age quartiles (16–17%) (Table 3). A cholera diagnosis was reported in similar proportions by females and males (OR = 1.05, 95% CI = 0.85–1.29). A cholera diagnosis was more commonly reported by participants who were illiterate (OR = 1.73, 95% CI = 1.40–2.14), did not speak French (OR = 1.86, 95% CI = 1.23–2.81), and who reported using antacids (OR = 1.73, 95% CI = 1.32–2.27). A cholera diagnosis was also more commonly reported by participants in households lacking a cell phone (OR = 1.35, 95% CI = 1.07–1.69), radio (OR = 1.39, 95% CI = 1.09–1.78), and electricity (OR = 2.22, 95% CI = 1.07–4.64). Equal proportions of participants with blood group O and those with other blood groups reported a cholera diagnosis (23%).

Based on validation results, a titer of 320 was set as the threshold for seropositivity for the Ogawa and Inaba vibriocidal assays. Secondary analyses examined survey results using lower, more lenient, vibriocidal titers of 80 and 160. Specific OD threshold values were calculated for both anti-CT ELISAs. Among the 2,527 (96%) participants from whom we collected serum, 980 (39%) had a positive Ogawa or Inaba vibriocidal result (titer ≥ 320) (Table 4), 1,331 (53%) participants had a vibriocidal titer ≥ 160, and 1,626 (64%) had a titer ≥ 80; 732 (29%) participants had a positive anti-CT test result for IgA or IgG. None of the 50 pre-epidemic serum specimens had a vibriocidal Ogawa titer > 80; two had a vibriocidal Inaba titer of 320.

Participants 13–23 years old had the highest prevalence of vibriocidal seropositivity (46%), and adults > 43 years old had the lowest (30%) (Table 5). Whereas a cholera diagnosis was most commonly reported by participants in the oldest age quartile, vibriocidal seropositivity was less common among participants in this quartile compared with the youngest quartile (OR = 0.59, 95% CI = 0.47–0.74). The prevalence of seropositivity in each age quartile varied according to the antibody test (Figure 3 Figure 3.

Proportion of participants within each age group seropositive for Vibrio cholerae antibody tests, cholera serosurvey, Grande Saline, Haiti, 2011. Anti-CT: antibody against cholera toxin.

Among all participants, regardless of antibody status, 6.2% had severe disease, 9.1% had moderate disease, 6.0% had mild disease and 78.7% were asymptomatic (Table 6). Among the 948 participants with vibriocidal titers ≥ 320 and severity data, 9.0% had severe disease, 9.7% had moderate disease, 6.9% had mild disease, and 74.5% were asymptomatic. A higher proportion of the 717 participants with positive anti-CT results had severe disease (12.3%) than did the vibriocidal group (9.0%), and a smaller proportion were asymptomatic (68.9% versus 74.5%).

Among 77 seropositive participants with blood grouping results who reported watery diarrhea, those with blood group O were not significantly more likely than those with other blood groups to report overnight hospitalization (OR = 2.11, 95% CI = 0.81–5.49) or receipt of IV fluids (OR = 1.91, 95% CI = 0.75–4.84). Among seropositive participants, antacid use was significantly associated with increased disease severity (Table 7). After adjusting for age and antacid use, lack of a cell phone (adjusted OR [aOR] = 1.46, 95% CI = 1.06– 2.01), radio (aOR = 1.63, 95% CI = 1.15–2.32), or electricity (aOR = 2.96, 95% CI = 1.04–8.45) in the household were associated with increased disease severity among seropositive participants. Illiteracy, inability to speak French, primarily using water from the river or canal, and treating drinking water were not significantly associated with increased disease severity after adjusting for age and antacid use among the subset of seropositive participants.

Among a subset of 589 households with information about all household members (including participants and non-participants), there were 2,396 residents, of whom 311 were reported to have had cholera (including participants). Within these households, there were 13 cholera deaths reported by study participants; estimated case-fatality rate = 4%. Based on these limited data, approximately 1 in 185 household residents died of cholera.