The sampling frame for the household survey consisted of all households in rural Artibonite Department as listed by the Institut Haitien de Statistique et d'Informatique (IHSI). The 2011 IHSI figures are based on a 2009 census and adjusted for population growth and movement after the January of 2010 earthquake. The IHSI list consists of enumeration areas (EAs), with population generally between 200 and 1,500 persons. EAs are categorized further into either rural or urban entities. According to IHSI statistics, Artibonite Department consists of 2,059 EAs and 1,478,515 total persons. The population defined as rural consists of 1,621 EAs and 1,005,417 persons (68% of the population).

The household survey consisted of 40 clusters (EAs) of 12 households, each for a sample size of 480 households. This survey was based on an expected prevalence of 50%, with a 5% margin of error for access to drinking water treatment products and access to an improved water source. We oversampled by 20% to account for difficulties in reaching selected clusters and completing all households within a cluster. The 40 clusters were randomly selected from the list of rural EAs using probability proportional to size (PPS) sampling, and therefore, all households had an equal chance of being selected. The 40 rural EAs selected from the IHSI listing were plotted and enumerated in Google Earth (Google Inc., Mountain View, CA). Household enumeration consisted of counting and marking visible roofs on the Google Earth image; all of the satellite images were from 2010 or newer. On arrival at the cluster, team leads used a combination of global positioning system (GPS) orientation (Garmin GPSMAP 76CSx; Garmin International, Inc., Olathe, KS) and Google Earth satellite images (showing enumerated households) to locate households within the EA. The shape files for each of the selected EAs were downloaded to each GPS unit so that they could be located precisely. Within each cluster, 12 households were selected using systematic random sampling, such that the entire EA was covered. Households were replaced with an adjacent household if the respondent was unavailable or the house was abandoned. The preferred respondent in the household survey was the female head of the household and/or the person responsible for water collection if at least 16 years of age (the legal age of consent in Haiti). Oral consent was obtained before proceeding with the interview.

The survey instrument consisted of questions regarding household demographics, preferred methods of communication for health messaging, access to modes of communication (e.g., radios and cell phones), cholera knowledge, and water collection, use, and treatment. Respondents were also asked about current water treatment practices, and a water test for free chlorine residual was conducted using portable test kits (CN-66 Color Wheel; Hach Company, Loveland, CO). The questionnaire was written in English, translated into Haitian Creole, and reviewed by local enumerators for errors. The questionnaire was piloted in an EA not selected for the survey and subsequently modified.

An Epi Info version 7 (CDC, Atlanta, GA) database was developed in French for data entry. All analyses, including χ² tests, were conducted in SAS version 9.3 (SAS Institute, Cary, NC). Data were weighted to account for unequal probabilities of selection. A two-stage weighting was applied; the first stage was based on the inverse probability of selecting a household within the selected EA using the Google Earth enumeration, and the second stage was based on the ratio of the planned and the actual number of households surveyed per EA.

Water source sampling took place concurrently with the household survey. One water sampler, trained in sterile sampling techniques, worked with each survey team to collect water used for drinking by 12 selected households in each cluster. Therefore, the water sampler attempted to collect a random sample of drinking water sources used by the surveyed households. In some instances, selected sources could not be sampled, because there was insufficient time to walk to the source.

Water samplers were also trained to fill out a water source description form to allow consistent classification of the source type and time of sample collection. At each source, a 100-mL sample was collected using sterile techniques in Whirlpak bags containing sodium thiosulfate. The samples were stored on ice in coolers before microbiological analysis. Samples were delivered for analysis within 8 hours of collection. Sample blanks were carried in each cooler to test at least one time for potential contamination of samples during storage and transport.

Two laboratories were assembled to conduct the microbiological sample analysis at the end of each day. Water samples were analyzed for total coliforms and Escherichia coli using the Colilert/Quanti-Tray method (IDEXX Laboratories, Inc., Westbrook, ME). The Quanti-Tray 2000 provides bacterial counts as low as < 1 most probable number (MPN) per 100-mL sample and up to 2,419.6 MPN/100-mL sample. Samples were incubated in Boekel Digital Incubators (Boekel Scientific, Feasterville, PA) at 35°C (±0.5°C) for 24 hours. Laboratory blanks were performed each day in each laboratory (15% of total) for quality assurance and control (QA/QC) of analytical procedure; field blanks were analyzed for QA/QC of each water samplers' collection techniques and accounted for 5% of total samples.

WHO has developed health risk levels for categorizing concentrations of E. coli found in water sources. Analyses were performed with respect to these WHO categories of health risk level.10

The assessment protocol, including the household questionnaire, was approved by the National Bioethics Committee of Haiti's Ministry of Public Health and Sanitation. The protocol was also reviewed by the CDC and determined to be a non-research public health program activity.

Of 433 respondents for household surveys, 88.4% were female. Respondent and household characteristics are summarized in Table 1. Approximately 60% of respondents reported having contact with a health worker in the last 3 months, of which 47% indicated that the health worker visited their household. The most commonly reported preferred mode for receiving health messages was radio (32.8%), with the majority (61.5%) of these responses from those individuals who owned a radio. Community health workers/brigadiers were the second most common response (26.7%) followed by doctors/nurses (11.5%), community meetings (10.3%), and church (7.5%). Trucks with megaphones and short message service (SMS) were each preferred by less than 5% of respondents.

Cholera knowledge was high, with a majority of respondents describing the roles of contaminated water in transmission and water treatment in cholera prevention. Almost all respondents (88.8%) agreed that cholera was preventable, and three-quarters of respondents could name the two most common prevention messages: treat drinking water (75.2%) and wash hands with soap and water (75.9%). The most common response for a way of contracting cholera was contaminated water (81.2%) followed by contaminated or undercooked food (70.9%). Other ways were answered much less frequently than these two responses; the next most common response was contact with someone sick with or who died of cholera at 7.0%. Slightly less than 5% (4.8%) of respondents did not know any ways that cholera was contracted.

Respondents were asked about their current primary source and secondary sources for drinking water along with primary water source for washing dishes, cooking, and bathing. Surface water (defined as unprotected springs, river/canal, and unclassified surface water) accounted for 45.3% of respondents' primary drinking water source (Table 2). Public taps/fountains/kiosks and boreholes with handpumps were the most commonly reported protected source types. Sixteen (3.4%) respondents reported using private kiosks for drinking water; water vended at these kiosks is provided from private companies, which advertise treatment with reverse osmosis membrane filtration. These sources were classified as improved, because these households also reported use of improved water sources for cooking and bathing.1 When all primary drinking water sources were classified according to the JMP definitions, 42.3% of households reported the use of improved sources.

Respondents were asked about consumption of river or canal water when away from the home (e.g., when working). Among respondents who did not report surface water as either their primary or secondary drinking water source (n = 236), 21.8% reported drinking river or canal water periodically. Among all respondents, 31.7% indicated periodic consumption of surface water when outside of the home.

As with drinking water source types, sources of water used for other purposes (washing dishes, cooking, and bathing) varied widely, and surface water was most commonly reported; approximately 50% of respondents listed using surface water for dishes/cooking, and 60% used surface water for bathing. Many respondents listed multiple source types for these other purposes.

Of 108 water samples taken, 55 (50.9%) samples were from improved sources, and 53 (49.1%) samples were from unimproved sources. Two-thirds of all samples (66.7%) were contaminated with E. coli, and 46.3% of all samples had an E. coli concentration greater than 10 MPN/100 mL (Table 3).

Both the improved and unimproved water sources showed a range of levels of fecal contamination, although a higher proportion of improved water sources had concentrations classified at lower risk levels. Approximately one-half (50.9%) of improved water sources sampled were positive for E. coli, and 23.6% had E. coli concentrations greater than 10 MPN/100 mL ( = 2.5 MPN/100 mL). For unimproved sources, 83.0% were positive for E. coli, with 69.8% of samples greater than 10 MPN/100 mL ( = 54.2 MPN/100 mL). A Pearson χ² test was conducted and confirmed a significant difference between the improved/unimproved source categories and presence of E. coli (P = 0.0004).

The source types tested with the best water quality were boreholes and private kiosks; 3 of 18 borehole samples were positive for E. coli ( = 0.7 MPN/100 mL). All six private kiosks were < 1 MPN/100 mL. All surface water (n = 6) and dug well samples (n = 14) tested positive for E. coli: = 1,681.4 and 119.6 MPN/100 mL, respectively. The public taps/fountains/kiosks had a wide variability in water quality, with results in all five health risk categories; 78.6% of samples from this source type (n = 28) were positive for E. coli ( = 6.5 MPN/100 mL). The geometric mean E. coli concentration for each source type is shown in Table 4.

Respondents were asked about their perception of the safety of their drinking water; 64.1% of respondents perceived their water source to be “safe as is.” The perceived safety of different water source types used ranged from 56.3% to 68.2%, with the exception of private kiosk users, of whom 95.3% thought it was safe. Of 154 (34.2%) respondents who did not believe their primary drinking water source type to be “safe as is,” the primary reasons were the water was not treated (86.2%), the source was not protected (16.6%), and the water did not taste good (9.3%).

Adult females bore the brunt of water collection (81.1%), but both male and female children had water collection responsibilities (20.3% and 38.1%, respectively). An adult male was involved in water collection in 16.3% of households. Most respondents (62.4%) reported collecting water two or more times a day. Of those respondents who could estimate the time required for water collection, almost one-half (44.1%) reported being 5 minutes or less from the source, and 71.4% reported being less than 30 minutes from the source. One-quarter of respondents did not know the time to source (25.7%).

Exposure to household water treatment products was high. The majority (80.5%) of respondents reported having treated their water in the past 3 months; however, 37.9% of those households (31.6% overall) had a water treatment product present at the time of the visit. Reasons for not treating water included an inability to afford the product/filter (45.8%) and a lack of access to the product (16.5%).

Respondents were asked about use of specific water treatment products and methods in the past 3 months. The most common household water treatment products used by households were disinfection products. In the last 3 months, 27.5% reported using two or more types of disinfection products. Sodium dichloroisocyanurate (NaDCC) tablets were the most commonly used product overall (Table 5). The majority (86.3%) of those individuals having treated water in the past 3 months did so in the form of Aquatabs (Medentech, Ltd., Wexford, Ireland; 69% overall). The 67-mg tablets were the most commonly used, but a range of sizes were reported (8.5, 17, 33, and 67 mg). PYAM tablets (Laboratorio Pyam S.A., Buenos Aires, Argentina) were used by 11.9% of households having treated in the past 3 months (8% overall). The other two most commonly used disinfection products in the last 3 months were liquid bleach (sodium hypochlorite, ranges from 5–15% active chlorine; locally branded as Jif or Clorox; 23.9% and 19% overall) and granular chlorine (calcium hypochlorite, ranges from 65–70% active chlorine; 9.4% and 8% overall). This concentrated form of granular high-test hypochlorite (HTH) was not distributed for household water treatment as part of response efforts, but it has been adapted for this use and is locally sold in some markets in small quantities. The use of locally marketed dilute sodium hypochlorite products specifically designed for household water treatment (locally branded as Dlo Lavi, Gadyen Dlo, and Klorfasil) was reported by 11 respondents (3.0% and 2.5% overall).

Use of a combined flocculant–disinfectant powdered mixture (branded as PUR; Proctor and Gamble Company, Cincinnati, OH) was also reported by nine respondents. Filters were not commonly used among rural households in the last 3 months (12.3% and 10% overall); among filter users, biosand filters were the most common. Boiling was not asked as a potential water treatment method used in the past 3 months on the questionnaire.

Table 5 also contains the current water treatment products and methods reported by households at the time of the survey. The proportion of those households currently using Aquatabs, liquid bleach, and PYAM was lower than the last 3 months. Boiling was mentioned as a current method of water treatment by one respondent.

Approximately one-half of the respondents treating water in the last 3 months reported paying for a product (50.9%). Of those respondents that paid for a product, liquid bleach and granular chlorine were the primary products purchased (70.2% and 79.8%, respectively), but 42.3% of respondents indicated that they had paid for Aquatabs. Markets and non-governmental organizations (NGOs) were the most common sources for Aquatabs, which was closely followed by health facilities. Most respondents using liquid bleach or granular chlorine reported obtaining the product in the market (60.5% and 77.4%, respectively).

Most respondents (85.1%) cited disinfection as their preferred method of treatment; this method was followed by filtration (9.9%) and boiling (3.8%). These preferences matched previous reported use (Table 5), with the exception of boiling. The most important reported reason for a preference, regardless of method indicated, was ease of use. The market was the preferred place to obtain household water treatment products (42.6%) followed by NGOs (21.9%), health facilities (18.0%), and community health workers (17.3%).

Respondents were asked to describe their water treatment practices. The results were analyzed as amount of product per 20-L bucket of water, given that most respondents stored drinking water in this vessel type (Table 1). Based on the manufacturer websites or specific dosage recommendations disseminated during the cholera response, the recommended quantity for each product was 67 mg of Aquatabs (e.g., one tablet of 67-mg size, two tablets of 33-mg size, etc.), 25 drops of liquid bleach (Jif or Clorox), two tablets (19 mg each) of PYAM, and two sachets of PUR. It was not possible to analyze dosage results for granular chlorine, because there was no standardized unit of measurement reported by households.

Reported dosing with the specific products used varied significantly, and the majority of respondents underdosed. Although just over two-thirds of all respondents reported use of Aquatabs in the previous 3 months, only 47% of these respondents knew what type of tablet they had used and/or had it in the home at the time of the survey. Approximately one-half (49.0%) of 143 respondents reported applying an acceptable dose (i.e., at least 67 mg); 25.0% of respondents underdosed, and 26.0% of respondents said they did not know how many tablets to use. Most of these respondents thought that one should put one tablet in one 20-L bucket, regardless of the tablet size. Of respondents reporting the use of liquid bleach in the last 3 months (n = 65), 15.9% reported the correct dose of at least 25 drops. Most of these respondents were underdosing (79.9%); three-quarters reported using five or fewer drops (75.1%). All respondents using PYAM or PUR also indicated underdosing.

Among households with water available to test (n = 328), 12.7% of households had a detectable free chlorine residual result (Table 6). Among households reporting having treated the current water in the house (38.3%), 30.4% had detectable free chlorine residual. Free chlorine residual was also analyzed with respect to time of treatment. Of 27.1% of respondents who said that they had treated water on the day of the survey, 48.1% were positive for free chlorine residual.