Introduction

Cholera is a centuries-old acute watery diarrheal disease caused by the bacterium Vibrio cholerae. Globally, it still kills between 21,000 and 143,000 people annually [1], with mortality rates remaining persistently high in some endemic regions [2]. Death from cholera is mainly due to hypovolemic shock, which results from rapid, severe fluid loss, without timely and adequate treatment [3]. Despite being both preventable and treatable, it remains a significant public health threat with the potential for high morbidity and mortality, particularly in settings with limited access to clean water, sanitation, and healthcare [4]. In countries such as Haiti, these conditions – worsened by humanitarian crises (e.g., conflict, natural disasters) – have contributed to large, recurring outbreaks and substantial deaths [5].

Existing research demonstrates that certain population groups are at higher risk of severe disease or death from cholera than others, including undernourished children [6], fetuses or neonates of women infected during pregnancy [7–9], and individuals immunologically naïve to the disease [3]. Some studies have further identified comorbidities (e.g., HIV infection, hypertension), advanced age (e.g., ≥ 50 years), late presentation for clinical care, and male sex as contributing factors, but these findings have not been consistently observed across settings [10,11]. Consequently, there remain gaps in our understanding of additional high-risk groups and specific predictors of mortality.

Over the past 15 years, Haiti has experienced several waves of devastating cholera outbreaks. After an initial epidemic that began in 2010, no cases were reported from March 2019 until a resurgence in October 2022, which caused 88,087 suspected cases and 1,324 deaths through December 2024 [12]. Given that there is little data on the most recent cholera epidemic in Haiti, we aimed to (1) describe the characteristics of suspected cholera cases treated at 15 health facilities in Haiti’s Artibonite and Centre regions, (2) estimate the case fatality ratio (CFR) by demographic strata and clinical symptoms at presentation, and (3) explore the predictors of cholera-related mortality.

Methods

Results

Of 19,870 registry records, we excluded 1,958 (35 cases treated outside our study period, 768 duplicate and repeated visit records, and 1,155 with missing CTC site identifier) (see Fig 1 for the selection of eligible cases). We analyzed 17,912 cases, including 134 decedents, who were treated in 15 CTCs in the Artibonite and Centre regions between October 10, 2022 and December 28, 2024. Nearly 88% (15,225/17,400) of individuals were classified as meeting the MoH’s definition of a suspected cholera case. Fig 2 shows the distribution of cases and deaths over the two-year period, with three distinct waves: an initial peak between November 2022 and February 2023 immediately after the start of the new outbreak in October 2022, followed by some smaller waves in 2023 in the context of an overall declining trend.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Selection of eligible cases for the full analysis of suspected cholera cases^a in 15 health facilities in the Artibonite and Centre, Haiti, during a confirmed cholera outbreak, 2022-2024.

^a Suspected cholera case defined by: Acute watery diarrhea and attendance for medical care at a cholera treatment center.

https://doi.org/10.1371/journal.pntd.0014626.g001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Distribution of 17,912 suspected cholera cases^a and case fatality ratios (CFR) reported by month in 15 health facilities in the Artibonite and Centre, Haiti, during a confirmed cholera outbreak, 2022-2024.

^a Suspected cholera case defined by: Acute watery diarrhea and attendance for medical care at a cholera treatment center.

https://doi.org/10.1371/journal.pntd.0014626.g002

The demographic and clinical characteristics of cases are summarized in Table 1, along with the associated CFRs. Overall, 76% of cases in the registry were treated in the Centre region. Nearly 51.0% (9,056/17,822) of cases were males. Children under five years old accounted for almost a third of all cases (5,688/17,659), the highest proportion of all age groups. Over half (9,551/15,283) of patients had symptoms lasting one day or more from symptom onset to presentation, but only 20.2% (3,577/17,690) took ORS prior to presentation. About 75.0% of all cases suffered from moderate or severe dehydration at presentation, and nearly two-thirds (8,951/13,759) reported vomiting. Less than three percent of individuals self-reported previously receiving an oral cholera vaccine (OCV), and there were no deaths among vaccinees. However, because the overall distribution was sparse and uneven across health facilities, with 92.9% of all vaccinees seen at just three CTCs, we did not include OCV receipt in additional analyses. A rapid diagnostic test for cholera was performed for only 4,964 individuals (27.7% of cases), primarily from four sites. Of those tested, 98.1% were positive for V. cholerae (See supplementary file, S1 Table).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of 17,912 suspected cholera cases^a in the Artibonite and Centre, Haiti, during a confirmed cholera outbreak, 2022-2024.

https://doi.org/10.1371/journal.pntd.0014626.t001

Cases treated at the CTCs in the Artibonite had a higher CFR than those in the Centre (1.4% vs. 0.6%). Older individuals had the highest age-specific CFRs, ranging from 1.1% among those aged 40–59 years to 3.1% among those ≥60 years. CFR was also high among cases with severe dehydration (2.1%) or vomiting (1.1%).

In univariable analysis assessing predictors of mortality and accounting for clustering at the facility-level, odds of death were significantly higher among those aged 40–59 years old (odds ratio [OR] 2.35, 95%CI 1.27 - 4.34, p = 0.007), and ≥ 60 years old (OR 7.25, 95%CI 4.28 - 12.28, p < 0.001), compared with children under 5 years old (see Table 2). Similarly, severe dehydration (OR 6.36, 95%CI 3.51 - 11.52, p < 0.001) and vomiting (OR 3.97, 95%CI 1.83 - 8.61, p = 0.001) were associated with greater odds of death in contrast to those with no or mild dehydration and no vomiting, respectively. Factors associated with significantly lower odds of death included presenting to CTCs in 2023 (OR 0.40, 95%CI 0.27 - 0.61, p < 0.001) or 2024 (OR 0.25, 95%CI 0.07 - 0.83, p = 0.02), compared with 2022.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Predictors of death from suspected cholera^a in the Artibonite and Centre, Haiti, during a confirmed cholera outbreak, 2022-2024. (N = 17,912).

https://doi.org/10.1371/journal.pntd.0014626.t002

In the mixed-effects multivariable logistic analysis, after adjusting for CTC site, year of presentation, older age, severe dehydration, and vomiting remained significant predictors of death from AWD. Cases aged ≥ 60 years had six-fold odds of death (adjusted OR [aOR] 6.08, 95%CI 3.56 - 10.38, p < 0.001). Severe dehydration (aOR 5.24, 95%CI 2.85 - 9.64, p < 0.001) and vomiting (aOR 2.96, 95%CI 1.33 - 6.56, p = 0.008) were associated with significantly higher odds of death in this population, compared to those with no or mild dehydration and no vomiting, respectively. By contrast, presenting 24 hours or more after symptom onset was inversely associated with death (aOR 0.59, 95% CI 0.37 - 0.94, p = 0.03).

We evaluated multicollinearity among the predictors (including vomiting and severe dehydration) using the variance inflation factor (VIF). VIF values were low across all imputed datasets, with a maximum of 1.61 for severe dehydration. The average residual intraclass correlation coefficient across the imputed datasets was 0.18, suggesting meaningful clustering by facility (i.e., cases within the same CTC had more similar odds of deaths than patients from different facilities). Results were generally consistent in sensitivity analyses restricting the analysis to the complete case data or to only the subset of AWD cases that met the Haitian MoH’s definition of suspected cholera case (See S2-S4 Tables) and using a missing-indicator method to manage missing data in lieu of the MICE approach (S5 Table).

Discussion

This study on the clinical characteristics and predictors of mortality in AWD patients medically attended at 15 CTCs during a cholera epidemic in Haiti from 2022 to 2024 has some key strengths. It includes a large cohort of cases, uses data from 15 CTCs across two major regions (Artibonite and Centre) of the country, and covers multiple years. It also reflects real programmatic conditions, while reporting a case-fatality ratio below 1%, which is an important public health indicator for quality of care [21].

The overall observed CFR of 0.7% is similar to the CFRs for the study regions (each 0.8%) and nationwide (1.2%) based on the Haitian MoH situation reports over the same period [12]. However, for the first three months of the epidemic, our observed CFR (1.4%) is lower than the nationally reported 14-day CFR (3.0%) [22]. Although death from diarrhea is rare when adequate interventions and treatment are provided, progress on global case fatality rates has generally now stalled, and, in some outbreaks, mortality has increased [4]. CFR in our cohort is considered ‘acceptable’ on aggregate based on general guidance of a minimum CFR threshold of <1% in treatment facilities [21,23] and is consistent with our prior reports from the 2013–2017 epidemic period in Haiti [13]. This illustrates that case fatality ratios below the ‘acceptable’ standard are indeed achievable even in health systems with limited resources. Most deaths from cholera should be preventable with appropriate staff, systems, supplies, and support. We have previously described Zanmi Lasante’s approach to early identification, training, and case management of cholera within a comprehensive multi-sectoral public health response including water, sanitation, hygiene and OCV [13].

We observed that CFR decreased over the three calendar years of this outbreak – a phenomenon that is often documented in outbreak settings. The reasons for this require further study as current routine data collection in Haiti – and most humanitarian settings – does not permit a fuller understanding of causal factors, either system- or patient-related. The improvement in CFR could, for example, reflect a gradual strengthening of the outbreak response over time. Since Haiti had not reported a confirmed cholera case for the three years prior to the resurgence in 2022, health facilities may have required some time to establish or re-establish human resources and supply chain to respond to and provide care to a large volume of patients with diarrhea. There may have also been shifts in patient-related health-seeking factors, but these have not been fully studied. One would expect that with increased public health messaging and community awareness, patients would progressively seek care earlier in the course of their illness. In our analysis, time from symptom onset to presentation was inversely associated with death. A similar finding has been reported for Mozambique [24], and could indicate a survivor bias, whereby more severe cases die before reaching a health facility while those presenting later may have milder illness.

Overall, there was a higher burden of cases among children under 5 and those aged 5–19 years compared to older adults, which is consistent with the epidemiology of cholera in other countries, as limited prior natural exposure increases susceptibility [25]. However, older adults were more likely to die than their younger counterparts in our study, with age ≥ 60 years old being a significant predictor of death. Previous studies of patients with medically-attended cholera in Haiti during the 2010–2011 epidemic and in several African countries including Congo, Guinea-Bissau, Nigeria, and Zambia have reported a similar association, and suggest that immunosenescence and pre-existing comorbidities may complicate cholera treatment in older patients [26–29]. We were, however, unable to assess the underlying prevalence of comorbidities including malnutrition in children, or hypertension for example, as such data are not routinely documented in the national registry, and medical records in this humanitarian setting were sparse.

A better understanding of the association between age and cholera mortality is important for improved triaging and may offer pathways to reduce mortality: individualized care for adults with suspected cholera, including attention to their comorbidities, is not typically included in cholera case management protocols [30]. For example, the prevalence of hypertension in Haiti among adults over 40 years of age is estimated to be 24–70% [31,32] and of heart failure to be 3.8% (higher than in high-income countries) [33]. Both of these diseases may complicate the management of severe dehydration, and comorbidities such as diabetes and HIV have been found to be important predictors of cholera mortality in other contexts [10]. Routine collection of data on existing comorbidities in AWD registries during cholera outbreaks may therefore be important for better identifying groups at higher risk of death.

Our study showed a consistent association between severe dehydration on presentation and risk of death, which is unsurprising as hypovolemic shock is the classic cause of death in cholera, and clinical assessment for and management of dehydration the core of case management. In line with our findings, studies of cholera in Nigeria and South Sudan have reported a four- to five-fold increased odds of death with severe dehydration [34,35]. In addition, vomiting in our cohort was associated with increased risk of mortality, and patients presenting with vomiting had nearly three-fold odds of dying. Vomiting has previously been linked with fetal death among pregnant women with cholera in Haiti [7,8], but not separately from dehydration in the general patient population, as the low collinearity between these variables suggests in our study. While vomiting may be a marker of severe cholera illness [35], it could also be a proxy for being unable to tolerate ORS, or for another pathophysiological issue. For example, patients with vomiting may have more severe electrolyte derangements, or kidney injury.

We further found that, after accounting for the patient-level factors, 18% of the variation in log-odds of death was attributable to differences between CTCs (as indicated by a residual intraclass correlation of 0.18). This may be due to clustering of comorbidities or other markers of increased mortality rate for which we could not account. Additionally, facility-level factors such as quality of care, case management protocols, and staffing can impact cholera outcomes. For example, being treated in a health facility with less than 0.5 nurses per two patients have been associated with higher odds of cholera-related death in Cameroon [36], while being treated in government-operated health facilities in Kenya lowered risk of mortality [37]. Our study did not encompass a health-facility level assessment of the CTCs to further evaluate the likely factors behind this variance.

Less than half of all cases sought medical care for their diarrheal episode within 24 hours of symptom onset, and only 20% took home-based ORS before presenting to a CTC. Although these factors were not significantly associated with mortality in the multivariable analysis, they are noteworthy observations in a country that already had nearly 10 years of experience with the management and prevention of cholera prior to the resurgence in 2022 and signify important opportunities for community management. Our research over the past decade in these regions consistently shows high levels of knowledge about cholera among residents of these areas, and low access to the preventive and treatment modalities in the community [38]. A protracted sociopolitical crisis in Haiti since 2018 may have further restricted access to basic resources and increased vulnerability to the effects of cholera [39].

Our study had limitations. We included multiple CTCs, but they may not be representative of other CTCs in the two regions or across Haiti. However, the cohort is the largest analysis of predictors of mortality from the country, and the included regions are generally underserved, lacking in many basic infrastructural supports, with a high prevalence of extreme poverty and, in those ways, resemble many regions of Haiti and other cholera-afflicted regions of the world. Institutional deaths are only one component of the impact of cholera outbreaks, and the surveillance registry we analyzed does not permit estimation of community deaths, which in previous studies have been high [40,41]. By analyzing individuals who presented to a CTC, we may have systematically excluded people who died quickly before seeking care, biasing the estimate for time from symptom onset to presentation.

We used the WHO’s broad case definition of suspected cholera, considering all medically attended AWD during the outbreak as suspected cholera [18]. However, because the signs and symptoms of mild to moderate cholera can overlap with those of other diarrheal diseases, AWD caused by other enteropathogens may have been attributed to cholera, leading to misclassification and an overestimation of cholera cases [42]. Rapid diagnostic testing for V. cholerae was performed without any particular regularity or protocol, and ultimately for about a quarter of cases. Therefore, we did not use RDT positivity as a primary inclusion criterion for our analysis. Yet, more than 95% of those tested were in fact positive for V. cholerae, albeit other studies have reported lower RDT positivity [43]. A sensitivity analysis using the Haitian MoH’s more limited definition of suspected cholera case did not change our findings, including the aggregate case fatality ratio (see S4 Table). Lastly, some key variables had a large proportion of missing data, including OCV receipt, which was excluded from the analysis partly due to low self-reported vaccination across CTCs. Although a large OCV campaign took place in 2022, and smaller ones in 2023 and 2024 [17,44], our study was not designed to evaluate the effectiveness of OCV on CFR. Nevertheless, our results remained robust across the selected strategies for handling missingness.

Conclusion

This study analyzes a cohort of 17,912 patients with medically-attended acute watery diarrhea in two regions of Haiti during the 2022–2024 cholera epidemic. The overall case fatality ratio was 0.7% but varied by epidemic year. Older age, severe dehydration, and vomiting emerged as significant predictors of mortality. Our findings add to the evidence on predictors of cholera mortality in the Haitian context during a resurgence of cholera while highlighting serious data gaps that limit our ability to identify intervenable variables for improved case management and outbreak response. Although data collection is challenging in fragile health systems and during humanitarian crises, we recommend strengthening the quality of surveillance data systems to help advance our understanding of the associations between patient characteristics and poor outcomes, and to inform strategies for further reducing cholera-related deaths.

Supporting information

Acknowledgments

The authors sincerely thank the Zanmi Lasante health workers and data collectors for their dedication and contributions to cholera surveillance throughout the outbreak.