Notes from the Field: Ongoing Cholera Epidemic - Tanzania, 2015-2016

Spatial distribution of suspected and confirmed cholera cases in Mwanza City, Northern Tanzania

Cholera, which is caused by Vibrio cholerae, persists as a devastating acute diarrheal disease. Despite availability of information on socio-cultural, agent and hosts risk factors, the disease continues to claim lives of people in Tanzania. The present study explores spatial patterns of cholera cases during a 2015-16 outbreak in Mwanza, Tanzania using a geographical information system (GIS) to identify concentrations of cholera cases. This cross-sectional study was conducted in Ilemela and Nyamagana Districts, Mwanza City. The two-phase data collection included: 1) retrospectively reviewing and capturing 852 suspected cholera cases from clinical files during the outbreak between August, 2015, and April, 2016, and 2) mapping of residence of suspected and confirmed cholera cases using global positioning systems (GPS). A majority of cholera patients were from Ilemela District (546, 64.1%), were males (506, 59.4%) and their median age was 27 (19-36) years. Of the 452 (55.1%) laboratory tests, 352 (77.9%) were confirmed to have Vibrio cholerae infection. Seven patients (0.80%) died. Cholera cases clustered in certain areas of Mwanza City. Sangabuye, Bugogwa and Igoma Wards had the largest number of confirmed cholera cases, while Luchelele Ward had no reported cholera cases. Concentrations may reflect health-seeking behavior as much as disease distribution. Topographical terrain, untreated water, physical and built environment, and health-seeking behaviors play a role in cholera epidemic in Mwanza City. The spatial analysis suggests patterns of health-seeking behavior more than patterns of disease. Maps similar to those generated in this study would be an important future resource for identifying an impending cholera outbreak in real-time to coordinate community members, community leaders and health personnel for guiding targeted education, outreach, and interventions.

Refugee Settlements and Cholera Risks in Uganda, 2016-2019

During 2016 to 2019, cholera outbreaks were reported commonly to the Ministry of Health from refugee settlements. To further understand the risks cholera posed to refugees, a review of surveillance data on cholera in Uganda for the period 2016-2019 was carried out. During this 4-year period, there were seven such outbreaks with 1,495 cases and 30 deaths in five refugee settlements and one refugee reception center. Most deaths occurred early in the outbreak, often in the settlements or before arrival at a treatment center rather than after arrival at a treatment center. During the different years, these outbreaks occurred during different times of the year but simultaneously in settlements that were geographically separated and affected all ages and genders. Some outbreaks spread to the local populations within Uganda. Cholera control prevention measures are currently being implemented; however, additional measures are needed to reduce the risk of cholera among refugees including oral cholera vaccination and a water, sanitation and hygiene package during the refugee registration process. A standardized protocol is needed to quickly conduct case-control studies to generate information to guide future cholera outbreak prevention in refugees and the host population.

Early detection of cholera epidemics to support control in fragile states: estimation of delays and potential epidemic sizes

BACKGROUND

Cholera epidemics continue to challenge disease control, particularly in fragile and conflict-affected states. Rapid detection and response to small cholera clusters is key for efficient control before an epidemic propagates. To understand the capacity for early response in fragile states, we investigated delays in outbreak detection, investigation, response, and laboratory confirmation, and we estimated epidemic sizes. We assessed predictors of delays, and annual changes in response time.

METHODS

We compiled a list of cholera outbreaks in fragile and conflict-affected states from 2008 to 2019. We searched for peer-reviewed articles and epidemiological reports. We evaluated delays from the dates of symptom onset of the primary case, and the earliest dates of outbreak detection, investigation, response, and confirmation. Information on how the outbreak was alerted was summarized. A branching process model was used to estimate epidemic size at each delay. Regression models were used to investigate the association between predictors and delays to response.

RESULTS

Seventy-six outbreaks from 34 countries were included. Median delays spanned 1-2 weeks: from symptom onset of the primary case to presentation at the health facility (5 days, IQR 5-5), detection (5 days, IQR 5-6), investigation (7 days, IQR 5.8-13.3), response (10 days, IQR 7-18), and confirmation (11 days, IQR 7-16). In the model simulation, the median delay to response (10 days) with 3 seed cases led to a median epidemic size of 12 cases (upper range, 47) and 8% of outbreaks ≥ 20 cases (increasing to 32% with a 30-day delay to response). Increased outbreak size at detection (10 seed cases) and a 10-day median delay to response resulted in an epidemic size of 34 cases (upper range 67 cases) and < 1% of outbreaks < 20 cases. We estimated an annual global decrease in delay to response of 5.2% (95% CI 0.5-9.6, p = 0.03). Outbreaks signaled by immediate alerts were associated with a reduction in delay to response of 39.3% (95% CI 5.7-61.0, p = 0.03).

CONCLUSIONS

From 2008 to 2019, median delays from symptom onset of the primary case to case presentation and to response were 5 days and 10 days, respectively. Our model simulations suggest that depending on the outbreak size (3 versus 10 seed cases), in 8 to 99% of scenarios, a 10-day delay to response would result in large clusters that would be difficult to contain. Improving the delay to response involves rethinking the integration at local levels of event-based detection, rapid diagnostic testing for cluster validation, and integrated alert, investigation, and response.

Geographical distribution and antibiotics susceptibility patterns of toxigenic Vibrio cholerae isolates from Kisumu County, Kenya

Background

Multiple drug resistance has become a major threat to the treatment of cholera. Recent studies in Kenya have described the epidemiology, especially the risk factors, of cholera; however, there is little information on the phenotypic and drug susceptibility patterns of Vibrio cholerae (V. cholerae) in outbreaks that in the recent past have occurred in western Kenya.

Aim

To characterise and determine the antibiotics’ susceptibility profiling of toxigenic V. cholerae isolates from Kisumu County.

Setting

The project was conducted in Kisumu County, Kenya.

Methods

A total of 119 V. cholerae O1, biotype El Tor, isolates collected during 2017 cholera outbreak in Kisumu County were used for this study. The samples were cultured on thiosulphate-citrate-bile salts sucrose (TCBS) agar and biochemical tests were carried out using standard procedures. Susceptibility tests were conducted by using various conventional antibiotics against standard procedures.

Results

Of the 119 isolates, 101 were confirmed to be V. cholerae belonging to serotypes Inaba and Ogawa, with Inaba being the predominant serotype (73.95%). The isolates were susceptible to ciprofloxacin (100%), ofloxacin (100%), gentamycin (100%), doxycycline (99%), ceftriaxone (99%) and streptomycin (96.04%) antimicrobials, and resistant to erythromycin (53.47%), amoxicillin (64.4%), nalidixic acid (83.2%) and ampicillin (89.11%), with high resistance to cotrimoxazole (99%) and tetracycline (97%).

Conclusion

Vibrio cholerae was resistant to multiple antibiotics, including those commonly used in the management of cholera. Taken together, there is a need to carry out regular surveillance on antimicrobial drug resistance during outbreaks.

Highly targeted spatiotemporal interventions against cholera epidemics, 2000-19: a scoping review

Globally, cholera epidemics continue to challenge disease control. Although mass campaigns covering large populations are commonly used to control cholera, spatial targeting of case households and their radius is emerging as a potentially efficient strategy. We did a Scoping Review to investigate the effectiveness of interventions delivered through case-area targeted intervention, its optimal spatiotemporal scale, and its effectiveness in reducing transmission. 53 articles were retrieved. We found that antibiotic chemoprophylaxis, point-of-use water treatment, and hygiene promotion can rapidly reduce household transmission, and single-dose vaccination can extend the duration of protection within the radius of households. Evidence supports a high-risk spatiotemporal zone of 100 m around case households, for 7 days. Two evaluations separately showed reductions in household transmission when targeting case households, and in size and duration of case clusters when targeting radii. Although case-area targeted intervention shows promise for outbreak control, it is critically dependent on early detection capacity and requires prospective evaluation of intervention packages.

Evaluation of an Emergency Bulk Chlorination Project Targeting Drinking Water Vendors in Cholera-Affected Wards of Dar es Salaam and Morogoro, Tanzania

In August 2015, an outbreak of cholera was reported in Tanzania. In cholera-affected areas of urban Dar es Salaam and Morogoro, many households obtained drinking water from vendors, who sold water from tanks ranging in volume from 1,000 to 20,000 L. Water supplied by vendors was not adequately chlorinated. The Tanzanian Ministry of Health, Community Development, Gender, Elderly and Children and the U.N. Children’s Fund, Tanzania, collaborated to enroll and train vendors to treat their water with 8.68-g sodium dichloroisocyanurate tablets (Medentech, Ireland). The Centers for Disease Control and Prevention (CDC) provided monitoring and evaluation support. Vendors were provided a 3-month supply of chlorine tablets. A baseline assessment and routine monitoring were conducted by ward environmental health officers. Approximately 3 months after chlorine tablet distribution, an evaluation of the program was conducted. The evaluation included a full enumeration of all vendors, an in-depth survey with half of the vendors enumerated, and focus group discussions. In total, 797 (88.9%) vendors were included in the full enumeration and 392 in the in-depth survey. Free residual chlorine (FRC) was detected in 12.0% of tanks at baseline and 69.6% of tanks during the evaluation; however, only 17.4% of these tanks had FRC ≥ 0.5 mg/L. The results suggest high acceptability and use of the chlorine tablets by water vendors. However, given variation in the water source used and longer storage times, dosing could be increased in future programming. Bulk chlorination using chlorine tablets offers an efficient community-level approach to treating water closer to the point of use.

Machine Learning Model for Imbalanced Cholera Dataset in Tanzania

Cholera epidemic remains a public threat throughout history, affecting vulnerable population living with unreliable water and substandard sanitary conditions. Various studies have observed that the occurrence of cholera has strong linkage with environmental factors such as climate change and geographical location. Climate change has been strongly linked to the seasonal occurrence and widespread of cholera through the creation of weather patterns that favor the disease's transmission, infection, and the growth of Vibrio cholerae, which cause the disease. Over the past decades, there have been great achievements in developing epidemic models for the proper prediction of cholera. However, the integration of weather variables and use of machine learning techniques have not been explicitly deployed in modeling cholera epidemics in Tanzania due to the challenges that come with its datasets such as imbalanced data and missing information. This paper explores the use of machine learning techniques to model cholera epidemics with linkage to seasonal weather changes while overcoming the data imbalance problem. Adaptive Synthetic Sampling Approach (ADASYN) and Principal Component Analysis (PCA) were used to the restore sampling balance and dimensional of the dataset. In addition, sensitivity, specificity, and balanced-accuracy metrics were used to evaluate the performance of the seven models. Based on the results of the Wilcoxon sign-rank test and features of the models, XGBoost classifier was selected to be the best model for the study. Overall results improved our understanding of the significant roles of machine learning strategies in health-care data. However, the study could not be treated as a time series problem due to the data collection bias. The study recommends a review of health-care systems in order to facilitate quality data collection and deployment of machine learning techniques.

Understanding the Impacts of Short-Term Climate Variability on Drinking Water Source Quality: Observations From Three Distinct Climatic Regions in Tanzania

Climate change is expected to increase waterborne diseases especially in developing countries. However, we lack understanding of how different types of water sources (both improved and unimproved) are affected by climate change, and thus, where to prioritize future investments and improvements to maximize health outcomes. This is due to limited knowledge of the relationships between source water quality and the observed variability in climate conditions. To address this gap, a 20-month observational study was conducted in Tanzania, aiming to understand how water quality changes at various types of sources due to short-term climate variability. Nine rounds of microbiological water quality sampling were conducted for Escherichia coli and total coliforms, at three study sites within different climatic regions. Each round included approximately 233 samples from water sources and 632 samples from households. To identify relationships between water quality and short-term climate variability, Bayesian hierarchical modeling was adopted, allowing these relationships to vary with source types and sampling regions to account for potentially different physical processes. Across water sources, increases in E. coli/total coliform levels were most closely related to increases in recent heavy rainfall. Our key recommendations to future longitudinal studies are (a) demonstrated value of high sampling frequency and temporal coverage (a minimum of 3 years) especially during wet seasons; (b) utility of the Bayesian hierarchical models to pool data from multiple sites while allowing for variations across space and water sources; and (c) importance of a multidisciplinary team approach with consistent commitment and sharing of knowledge.