Modeling the Effect of Water, Sanitation, and Hygiene and Oral Cholera Vaccine Implementation in Haiti

Abstract

Introduction

In October 2010, cholera was introduced to earthquake-stricken Haiti.¹ Within days of its introduction, a National Cholera Surveillance System was implemented.¹ Through June 30, 2013, Haiti had reported 663,134 cases of cholera (Figure 1)²; of these, 366,995 (55.3%) were hospitalized and 8160 (1.2%) died. As we approach the three-year mark, cholera will likely be considered endemic to Haiti.

Open in a separate window

Figure 1.

Total (n = 663,134) cholera cases by week, Haiti, October 20, 2010–June 30, 2013.

Water, sanitation, and hygiene (WASH) interventions, such as latrines, point-of-use chlorination and piped water, have long been recognized as effective prevention measures against cholera and other diarrheal diseases.^3–12 In 2008, 63% of the Haitian population had access to improved water and 17% to improved sanitation.¹³ In 2010, after the earthquake, the Haitian Directorate for Potable Water and Sanitation reported that 26% of the rural population received improved water and 10% improved sanitation; in the Port-au-Prince metropolitan area, the coverage was 35% and 20%, respectively (Haitian Directorate for Potable Water and Sanitation five-year plan). Many public health scientists believe that sustained improvements in access to safe water and sanitation can eliminate transmission of cholera in Haiti, citing interventions used throughout South and Central America in the 1990s.^14,15 The WASH interventions, including hand washing, have the additional benefit of reducing the incidence of other diarrheal and respiratory diseases.^3,5,16,17 Although improving water and sanitation infrastructure is the ultimate goal of the Haitian Government and the international community, it will take considerable time.¹⁸

Oral cholera vaccine (OCV) has been proposed as an effective adjunct for cholera control in endemic and epidemic settings.^19,20 Two whole-cell, killed, World Health Organization–prequalified OCVs are available: Dukoral^® (Crucell, Stockholm, Sweden) and Shanchol™ (Shantha Biotechnics, Hyderabad, India). Both vaccines require two doses given two weeks apart, with protective immunity developing approximately one week after the second dose.^21,22 The Haitian government sanctioned two pilot studies²³ to assess the acceptability and feasibility of Shanchol™ vaccine, one in urban Haiti and one in rural Haiti.²⁴ Based on these pilot study findings and findings from previous OCV studies, the Pan American Health Organization has recommended targeted or mass OCV campaigns that use Shanchol™ as an intermediate bridge to reduce cholera transmission in Haiti while improvements in water and sanitation infrastructure are implemented.²⁴ We present results of a model that illustrates the potential impact of WASH and OCV interventions independently and in combination. These results can aid public health decision makers in allocating resources to prevent cholera transmission in Haiti.

Methods

We used Excel 2010 (Microsoft, Redmond, WA) to develop a spreadsheet-based, static mathematical model in which we allowed a degree of indirect protection (or herd immunity) for OCV and WASH interventions by including non-linear relationships between percentage of population covered and percentage of population effectively protected (i.e., for a given percentage vaccinated or who received WASH interventions, an additional percentage was also indirectly protected). For WASH interventions, we included latrines, point-of-use chlorination, and community piped water (standpipes). We divided Haiti's population into urban and rural elements. For both urban and rural populations, and for each intervention, we constructed three scenarios that illustrated potential rates-of-growth of coverage over 20 years. We also constructed scenarios in which we allowed a combination of WASH and OCV in rural and urban areas. In these combined scenarios, we conservatively assumed that persons who received OCV would not be covered by WASH interventions and vice versa. Thus, coverage for either WASH or for OCV interventions would never exceed 50%. We modeled 16 scenarios: six WASH, six OCV, and four that combined WASH and OCV interventions. For further details, see Online Supplemental Materials.

Intervention effectiveness.

For each intervention, we included non-linear relationships between coverage and effectiveness that take into account indirect protective effects²⁷ (Figure 2). For OCV (Shanchol™), we fitted an exponential curve to the OCV (Dukarol^®) coverage-effectiveness modeling data from Longini and others.²⁸ (Figure 2; Supplemental Table 2). The randomized control trial data for Shanchol™ administered as a two-dose regimen showed a direct efficacy of 67% after two years, which is nearly identical to that for Dukarol^®21 and 66% after three years.^29,30 Therefore, we assumed that our use of the Dukarol^® coverage-effectiveness curve as a proxy for the Shanchol™ coverage-effectiveness curve was reasonable. We did not examine partial vaccination effect (i.e., receiving only one dose). The OCV coverage in our model implies effective coverage with two doses of Shanchol™ vaccine, and in addition, assumes that all two dose recipients will receive one booster dose every three years thereafter (Supplemental Table 4). For latrines and point-of-use chlorination, we estimated non-linear curves based on data from two reviews of interventions^3,5 (Figure 2; Supplemental Table 2). For piped water, we assumed a non-linear curve with a protective effect of 90% at 100% coverage (Figure 2; Supplemental Table 2). Because there are little data on the synergistic effect of one or more WASH interventions, we used a conservative approach and assumed no additive effect across the various combinations of possible WASH interventions.⁷ Therefore, we used a stepwise introduction of WASH interventions over time, and the intervention with the stronger protective effect supplanted the other (i.e., piped water > chlorinated water > latrines). For further detail, see the Online Supplemental Material.

Open in a separate window

Figure 2.

Coverage-effectiveness curves for various interventions. Black line indicates oral cholera vaccine; dark gray dotted line indicates piped water; gray dotted-dashed line indicates point-of-use chlorination; light gray dashed line indicates latrines.

Results

We developed eight urban scenarios (three WASH, three OCV, and two WASH/OCV combined) and eight rural scenarios (three WASH, three OCV, and two WASH/OCV combined). WASH scenario 1 (WASH/R1 + WASH/U1) averted 78,567 cases of cholera. WASH scenario 2 (WASH/R2 + WASH/U2) averted 71,106 cases of cholera. WASH scenario 3 (WASH/R3 + WASH/U3) averted 57,949 cases of cholera (Tables 1 and ​and4,4, Figure 3).

Open in a separate window

Figure 3.

Cumulative cases of cholera averted by water, sanitation and hygiene (WASH) interventions, oral cholera vaccine interventions (OCV) or a combination of both (Combi) over a 20-year period in Haiti, and assuming a baseline national cholera incidence rate from Malawi (1990–2010) applied to urban and rural Haiti.

OCV scenario 1 (OCV/R1 + OCV/U1) averted 77,636 cases of cholera. OCV scenario 2 (OCV/R2 + OCV/U2) averted 57,668 cases of cholera. OCV scenario 3 (OCV/R3 + OCV/U3) averted 38,569 cases of cholera (Tables 2 and ​and4,4, Figure 3).

The rate of intervention coverage extension had the largest effect on cases of cholera averted (the difference between scenarios 1, 2 and 3 for either WASH or OCV).

Combined scenario 1 (Combined/R1 + Combined/U1) averted 88,974 cholera cases. Combined scenario 2 (Combined/R2 + Combined/U2) averted 71,586 cholera cases (Tables 3 and ​and4,4, Figure 3).

In our sensitivity analyses, we found that although the absolute number of cases of cholera averted is sensitive to the expected number of cholera cases given different baseline annual incidence of cholera, the relative effect of each intervention scenario is the same (Figure 4 ; Supplemental Figure 1). Our sensitivity analysis of combined interventions (Figure 5 ) demonstrated decreasing returns on investment (marginal increase of the number of cholera cases averted) when OCV coverage at year 5 and piped water coverage at year 20 are high. The OCV coverage of 30% at year 5 achieved similar outcomes with that of 50% coverage at year 5, regardless of piped water coverage of 10–50% at year 20 (Online Supplemental Material).

Open in a separate window

Figure 4.

Cumulative cases of urban (U) cholera cases averted by water, sanitation and hygiene (WASH/U1), oral cholera vaccine (OCV/U1) and a combination of WASH and OCV (Combined/U 1) scenarios when 20-year baseline annual incidence data from Malawi (1990–2010), Mozambique (1990–2010) and India (1961–1981) are applied to Haiti demographic data.

Open in a separate window

Figure 5.

Cumulative cholera cases averted over 20 years in Haiti by combined water, sanitation and hygiene (WASH) and oral cholera vaccine (OCV) intervention scenarios. The x-axis refers to OCV coverage at year 5, and the different lines refer to the proportion of total piped water coverage (10%, 20%, 30%, 40%, and 50%) at year 20. We assumed that 1) persons who received OCV would not be covered by WASH interventions and vice versa; 2) WASH and OCV interventions never exceed 50%, respectively; 3) point-of-use chlorination increases from 20% (urban), or 26% (rural) and will remain the same until piped water takes over (provided assumption 1 met); 4) latrine coverage remains 10% until it is taken over by point-of-use chlorination or piped water; 5) piped water baseline = 0% in rural areas and 10% in urban areas, and piped water coverage starts increasing at a constant rate from year 6 onwards; and 6) OCV coverage increases at a constant rate from 1% baseline at year 0, peaks at year 5 and decreases thereafter at a constant rate to reach 5% at year 20.

In our final sensitivity analysis, we explored scenarios to assess the impact on cases averted after a more rapid scale up of OCV coverage by year 5, as well as scenarios with sustained OCV coverage to year 20 (Table 5) . In our four combined scenarios described in Table 3, OCV coverage reached either 20% or 10% at year 5, and decreased to 5% or 0% by year 20, respectively (Combined/U1, Combined/R1, Combined/U2, Combined/R2). For Combined/U1 + R1, effective OCV coverage increased from 1% at year 0 at a constant rate, reached 20% at year 5, then decreased at a constant rate to 5% at year 20 (Table 3), thereby averting 88,974 cases over 20 years (Table 4). If in this scenario, effective OCV coverage was allowed to reach 50% at year 5, and then decrease at a constant rate to 5% at year 20, an additional 6,738 cases (95,712 cases) would be averted (Table 5). For Combined/U2 + R2, effective OCV coverage increased from 1% at year 0 at a constant rate, reached 10% at year 5, then decreased at a constant rate at 0% at year 20 (Table 3), thereby averting 71,586 cases over 20 years (Table 4). If in this scenario, effective OCV coverage were allowed to reach 50% at year 5, and then decrease at a constant rate to 5% at year 20, an additional 23,933 (95,519 cases) would be averted (Table 5). However, we estimated very small further increases in cases averted when we allowed for effective OCV coverage to reach 50% by year 5 and remain at that level to year 20 (Table 5). For example, in modified scenario Combined/U1 + R1, sustaining 50% coverage from year 5 through year 20 resulted in 95,777 cases averted (i.e., an additional 65 cases averted). Similar modest increases in case averted (95,703 cases averted; i.e., an additional 184 cases averted) were estimated for Combined/U2 + R2 (Table 5). For results of the other sensitivity/uncertainty analyses, see Online Supplemental Material.

Discussion

Our results demonstrate that the rate of expanding coverage of WASH and OCV interventions affects the cumulative number of cases of cholera averted. The scenarios demonstrate that the modeled WASH and OCV interventions averted similar numbers of cholera cases. The assumptions of coverage for this model took into consideration the theoretical implementation of WASH and OCV interventions. Our goal was to demonstrate the scope of results given different rates of implementation and levels of coverage attained through a variety of scenarios, as well as with the sensitivity and uncertainty analyses. Scenarios that combined WASH and OCV interventions were most effective, which supports current efforts to implement both interventions when feasible.²⁴

The WASH infrastructure provides a long-term, sustainable solution for prevention of cholera.¹² Evidence from Europe and North America over the past two centuries, and more recently from Latin America, demonstrate that as water and sanitation coverage improves, the risk of epidemic or endemic cholera transmission is greatly reduced.^12,14,15 WASH also prevents the transmission of many other diarrheal diseases, which in Haiti, as in many developing countries, is a leading killer of children less than five years of age.^32,33 The overall benefit of expanding WASH coverage extends far beyond its effect on cholera alone.

The OCVs should help reduce the burden of cholera while WASH coverage is expanded, given the considerable amount of time required to improve WASH infrastructure (e.g., piped water and sewers). However, an OCV program should not be considered as a long-term alternative substitute for WASH. Implementation of OCVs will present its own challenges. Currently available OCVs are not 100% efficacious, induced immunity wanes over time thereby requiring periodic booster dosing, and today's globally available OCV supply is not sufficient to vaccinate the entire Haitian population with the required two-dose regimen. In addition, evidence from the routine childhood expanded program for immunizations and recent nationwide vaccine campaigns in Haiti has demonstrated varying ranges of coverage.^34–37 Although rapid expansion of effective OCV coverage to 50% of Haitian population (10 million doses of administered vaccine or more) by year 5 may avert an additional 6,000–24,000 cases (Table 5), such rapid expansion is likely beyond the country's current capacity. Therefore, we highlight coverage scenarios (Table 3) in our model that we believe could be realistically achieved based on Haiti's recent experience with routine expanded program for immunizations and vaccine campaigns.

Our study has several limitations. First, we chose a static model while simultaneously incorporating an indirect effect by applying non-linear coverage-effective curves to WASH and OCV interventions. Thus, the model takes into account the current effect of an intervention (direct and indirect protection) and is an improvement over a classical static model. Unlike a model that simulates the transmission dynamics of cholera over time (e.g., ordinary differential equation models),³⁸ a static model does not account for the future effect of the current intervention because the baseline incidence does not take into account the intervention applied in the previous year(s). However, our static model, like others,^39,40 avoids having to estimate uncertain and unknown parameters required for dynamic models that explore the impact of multiple interventions introduced at various stages over time.⁴¹ More data will be needed to reduce the parameter uncertainty of existing dynamic models of cholera for Haiti.^41,42 Second, although we accounted for population growth, we did not account for the likely migration of the Haitian population from rural to urban areas over the next 20 years. Third, we recognize that the baseline 20-year annual cholera incidence data from Malawi, Mozambique, and India that we used as illustrations for medium, high, and low incidence, respectively, may have been subject to under-reporting. However, our findings were robust across all three baseline country scenarios. However, it is clear that every country's experience with endemic cholera is unique. Only time will tell what Haiti's experience will be. Fourth, apart from modeling urban and rural Haiti separately, we did not study the impact of geographic variation on cholera incidence and intervention implementation (e.g., targeted immunization). Fifth, we acknowledge the uncertainty associated with the coverage-effectiveness curve used for each intervention. However, because data are sparse for OCV and WASH intervention coverage-effectiveness curves, we used modeling outputs of Longini and others²⁸ to fit our exponential curves for OCV, and we also applied exponential curves to the WASH coverage-effectiveness relationship.²⁷

Our study emphasizes that intervention coverage affects variation in estimated number of cumulative cholera cases averted over an extended period, and demonstrates the probable synergistic effects of WASH and OCV when used in combination. Our study should not be interpreted as an exact prediction for the number of cholera cases that could be averted in Haiti under the scenarios outlined, but it serves to demonstrate that WASH and OCV interventions can play an important role in decreasing the burden of cholera, and that maximizing intervention coverage is the central variable to their success. Transmission and intervention dynamics need to be understood so that informed decisions can be made about how to allocate limited resources. The Haitian Government recently released its National Plan for the Elimination of Cholera.¹⁸ This plan outlines a combination of public health interventions that include the use of OCV while expanding access to clean water and sanitation. Our study suggests that this combined strategy will be effective.

Supplementary Material

ACKNOWLEDGMENTS

Notes

Disclaimer: The findings and conclusions expressed in this report do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Footnotes

References