Abstract
Background Understanding and continually assessing the achievability of global health targets is key to reducing disease burden and mortality. The Global Task Force on Cholera Control (GTFCC) Roadmap aims to reduce cholera deaths by 90% and eliminate the disease in twenty countries by 2030. The Roadmap has three axes focusing on reporting, response and coordination. Here, we assess the achievability of the GTFCC targets in Nigeria and identify where the three axes could be strengthened to reach and exceed these goals.
Methodology/Principal Findings First, historical data were analysed for cholera death rates and a range of well-established cholera risk factors (particularly those included in the best fit models), using correlation, regression, autocorrelation function and AutoRegressive Integrated Moving Averages. Next, future scenario datasets were created (based on varying degrees of socioeconomic development and emission reductions) and used to project future cholera outbreak occurrence and transmission to 2070. The two models used were a generalised linear model and a random forest model. The historical data highlighted the significant cholera burden in Nigeria and its correlation with sanitation access. The projections suggest that significant reductions in cholera cases could be achieved by 2030, particularly in the more developed southern states, but increases in cases remain a possibility. Meeting the 2030 target, nationally, currently looks unachievable and we propose a new 2050 target focusing on reducing regional inequities.
Conclusion/Significance The 2030 targets could potentially be reached by 2030 in some parts of Nigeria, but more effort is needed to reach these targets at a national level, particularly through access and incentives to cholera testing, sanitation expansion, poverty alleviation and urban planning. The results highlight the importance of and how modelling studies can be used to inform cholera policy and the potential for this to be applied in other contexts.
Author Summary By statistically analysing the historical cholera deaths and risk factors data and using future scenario cholera forecasting, we assessed the achievability of the Global Task Force for Cholera Control 2030 Roadmap in Nigeria, based on sustainable development and environmental protection. The results highlighted how regional inequities within Nigeria will likely prevent the 2030 targets being reached. The northern states in Nigeria are more rural, less developed and have greater levels of insecurity and conflict. Reaching the levels of peace and development achieved in the south at a national level, along with improving urban planning and access and incentives to cholera testing, will be fundamental in reaching the Roadmap targets in Nigeria. Cholera forecasting is understudied and here we present the most in-depth projections available for Nigeria. Projections provide a snapshot of potential future conditions and show what is needed going forward to meet and exceed health targets and reduce disease burden. We highlight how quantitative research can be used to inform policy and the relevance of doing so. Quantitative research should fundamentally aim to improve global health and by presenting our work with a strong policy focus, we highlight the relevance and importance of doing so.
Introduction
Global health and development targets are widely used strategies to address international challenges. In 1992, the Global Task Force on Cholera Control (GTFCC) was established as a global partnership of more than 50 institutions and in 2017, the GTFCC launched “Ending Cholera: A Global Roadmap to 2030”. The Roadmap encouraged partner organisations to sign the Declaration to End Cholera, which focused on three axes: 1. early detection and response, 2. interventions in cholera hotspots and 3. effective coordination at all levels. The GTFCC aims to stop country-wide uncontrolled cholera outbreaks by 2030 and eliminate the disease from twenty countries, resulting in a 90% reduction in cholera deaths (Fig 1) [1].
The GTFCC theory of change for cholera elimination (adapted from: [1])
Despite the commitment of governments and organisations to large-scale strategies, many argue that they fail to produce significant improvements in their goals [2,3]. There have been gains in cholera control at a local level in recent decades, although this has been minimal at the global level [4,5]. As the 2030 goal approaches, whether these targets can be achieved given the current pace of development and progress is highly uncertain [6]. Understanding the achievability of these targets will help countries and global partnerships to plan for 2030 and beyond, continuing to make improvements in cholera prevention and control. Goals should be ambitious and encourage partners to strive for the best outcome possible, but they also need to be clear and have significant commitment and motivation from governments, non-governmental organisations and the population.
There is an estimated 1.3 billion people at risk of cholera globally and approximately 2.86 million annual cases (1.3-4.0 million) [7,8], the majority of which are in sub-Saharan Africa and the Indian Subcontinent. Development and health are strongly connected, and several countries and regions with the lowest levels of development, in terms of poverty, education and health, also overlap with areas of high cholera burden [9]. Nigeria currently has one of the largest cholera burdens globally and a significant number of people living in poverty, making it a critical area of study. Nigeria is also one of the largest African economies, has the largest populations, and has made several gains in socioeconomic development in recent decades [10].
Here, we aim to understand if the GTFCC 2030 targets are achievable in Nigeria, based on the current Roadmap and progress. We first illustrate the current situation in terms of historical cholera and environmental and social risk factors. We then project cholera occurrence and transmission to 2070 both nationally and sub-nationally (administrative level 1) with varying degrees of global change using two models [11,12]. Relatively few studies have projected cholera burden, most of which have taken a climate change focus [13-15] and even fewer studies have evaluated the likelihood of meeting the 2030 GTFCC targets using forecasting tools [16,17]. The projection results will be used to inform the three axes outlined in the Roadmap, and we make suggestions of how to reach the 2030 targets in Nigeria, highlighting both successes and areas for improvement in the current strategy.
Methods
Datasets
The number of reported cholera deaths (from suspected or confirmed cholera cases) were used in the historical analysis, due to the GTFCC goals largely focusing on reducing mortality (90% reduction in deaths). However, cholera deaths were transformed from raw numbers to a death rate (deaths per 100,000 people) to make the results more comparable, as Nigeria is currently the most populous country in Africa (211 million, compared to the second largest population, Ethiopia at 118 million [18]). Cholera death data were taken from two open access sources, cholera data can vary among datasets and using multiple sources allows for greater understanding of over-reporting, under-reporting and reporting lags. The two widely used sources included: the WHO’s Global Health Observatory (1970-2016) [19] and the Global Health Data Exchange (GHDx) (1990-2016) [20]. The temporal and spatial scale of both cholera deaths datasets were annual and national. The historical environmental (meteorological drought, temperature, hydrological drought) and social (poverty, sanitation, conflict) data selected in the best fit models used here (detailed below) were taken from several data sources detailed in Table 1.
The environmental and social data fitted to the models and used in the historical analysis
WorldClim [26] was used for the projected temperature and precipitation data, at administrative level 1 and a monthly temporal granularity. WorldClim data is from Coupled Model Intercomparison Project 6 (CMIP6) downscaled future gridded temperature and precipitation projections (aggregated to administrative level 1), processed for nine global climate models. The data included minimum temperature and maximum temperature measured in degrees Celsius (°C) and precipitation (in mm). Projections were single values for 2050 and 2070 using three different Representative Concentration Pathways (RCPs) (RCP4.5, 6.0 and 8.5 (further explanation of the RCPs is available in S1 Text).
Historical Analysis
Time-series of the historical cholera deaths per 100,000 and environmental and social data were used to illustrate how cholera and its associated risk factors have changed through time. Pearson’s correlation coefficients of all the environmental and social variables data, based on the covariates included in the best fit models, were calculated to understand how they related to each other for the full datasets. One data source for each of the cholera and poverty data sources were selected for the correlations (WHO cholera data and the proportion of the population in extreme poverty) and were chosen based on the length of the timescale to increase data completeness. To visualise and analyse the trends, linear regression trend lines and loess curves, autocorrelation function (ACF) and AutoRegressive Integrated Moving Average (ARIMA) were used for the historical cholera death data. ACF is a measure of autocorrelation and describes how correlated the present value of the series is related with its past values. These approaches allow quantification of both the significance of time and the magnitude and direction of its effect, which can then be used to inform future conditions (via ARIMA).
Projections
The national projections were predicted using the best fit model from Charnley et al. [11]. In summary, national annual cholera data were taken from the WHO Global Health Observatory and transformed to a binary outcome variable of cholera outbreak occurrence (where 1 represented at least one reported cholera case). Nineteen environmental and social covariates were tested through a covariate selection process, which removed covariates which were not significantly associated with the outcome variable at a 10% confidence limit in a univariate model and then clustered the remaining covariates based on a threshold correlation coefficient of r > 0.75. The data were then fit to multivariate generalised linear models using maximum likelihood estimations and a log-log link function. The best fit model was identified based on Bayesian Information Criterion and area under the receiver operator characteristic curve. The selected covariates included Palmers Drought Severity Index (PDSI), average temperature (°C), poverty headcount at <$1.25/day and freshwater withdrawal per capita (m3/person/year). To test model robustness, autocorrelation diagnostics were run using AR1 (autoregressive model of order 1) and leave-one-out cross validation and found the effect of time and space negligible.
The sub-national projections were predicted using the best fit model from Charnley et al. [12]. In summary, sub-national cholera surveillance data were acquired from the Nigeria Centre for Disease Control. Data were available for 2018 and 2019 and the data were subset to only include cholera cases confirmed by rapid diagnostic tests or laboratory culture. Incidence was calculated from the confirmed cholera case data and used to model cholera time-varying reproduction number (R) for states (administrative level 1) which had sufficient cases (>40 cases), to ensure R was calculated within the time window (monthly sliding windows). Cholera R was then used as the outcome variable for fitting random forest models, with R values over 1 signifying increased transmission. Twenty-two covariates were considered for inclusion in the best fit model and were similarly clustered at the same correlation coefficient threshold (r > 0.75), to account for the potential of multi-collinearity. Possible combinations of covariates were then stepped through, with random forest variable importance, based on node impurity to guide the process. The best fit model was identified based on predictive power, evaluated through multiple metrics including correlation between actual v predicted values, R2 and root-mean-square error. Random forest models are easily overfit, to account for this, mean absolute error was used to test for the presence of overfitting in the model. The best fit model included PDSI, Multidimensional Poverty Index (MPI), percentage access to improved sanitation and monthly conflict event frequency.
The models were used to predict either cholera outbreak occurrence (national) or cholera R (sub-national) at decadal increments from 2020 to 2070 using the scenario data (described below). Bootstrap resampling (10,000 samples) was used to obtain 95% confidence intervals for all projections and all data sources used to fit the models are listed above in Table 1. The projections were applied to the 2030 Roadmap targets and a new 2050 target delineated. Twenty-fifty was chosen as the target year, due to it being halfway from the current target (2030) to the end of the projection period (2070), taking into consideration the full scope of the projections. Additionally, the two global initiatives the scenarios are based upon, the RCPs and SDGs, have 2050 targets [27,28].
Projection Scenarios
Five projection scenarios were created here to project cholera to 2070 using the two models. Despite 2030 being the target year for the Roadmap, cholera projections were made to the furthest point the data allowed (2070). The scenarios include several degrees of global change and are based upon the RCP scenarios and attainment of the Sustainable Development Goals (SDGs) [29]. By using the RCPs and SDGs, this covered a wide range of environmental and social scenarios, accounting for varying degrees of emissions reductions and socioeconomic development. Both the RCPs and SDGs are widely accepted methods and targets for environmental and socio-economic development in research and policy. The scenarios are briefly defined as follows, and more details are given below:
Scenario 1 (S1) - Best case scenario meeting RCP4.5 and the SDG
Scenario 2 (S2) - Intermediate progress scenario between S1 and S3
Scenario 3 (S3) - Minimal development and emissions reductions but progress is still made towards to SDGs and the Intergovernmental Panel on Climate Change targets
Scenario 4 (S4) - Some regression from the current levels of sustainable development and increased emissions
Scenario 5 (S5) - Worst case scenario with significant regression in development and emissions increases
National
The national projection scenarios were taken from Charnley et al. [11] for S1, S2 and S3. In summary, PDSI projections related to dataset averages and future conditions following the historical linear trend (PDSI [21] ∼ Year, coefficient −0.014961, p-value 0.000512). Temperature was based on the WorldClim projections (RCP4.5, 6.0 and 8.5 for 2050 and 2070) and water withdrawal on varying degrees of meeting SDG6.4 (increase sustainable water-use efficiency across all sectors to address water scarcity). In terms of water security, Nigeria is a high resource and low withdraw country (S1 Fig) and therefore has increased water use in its more optimistic scenarios, based on the assumption that more water use means greater access. Poverty scenarios were set at varying degrees of achieving SDG1.1 and 1.2, which state a 50% reduction in extreme (<1.25/day) poverty by 2030 and poverty eliminated by 2070.
For S4 and S5, RCP6.0 and 8.5 temperature values were reached 20 years earlier (2050 by 2030 and 2070 by 2050). For PDSI, previous literature was used to understand the full extent of drought changes across Nigeria, due to the limitations in projecting PDSI (S1 Text). Africa is not projected to be significantly drier and historical studies suggest relative stability [30-33]. However, projected PDSI changes are spatial heterogeneous and for Nigeria, drought projections suggest both stability [34] and drying [35,36]. To account for the full range of projected changes and uncertainty stated in the literature, the scenarios include both stability and significant drying to 2070. The poverty projections represent a 30% and 50% increase by 2070 for S4 and S5, respectively and for water withdrawal, instead of increasing from 2050 as seen in S1-S3, withdrawal would decrease by 10% (S4) and 20% (S5). A summary of the national scenarios is shown in Table 2.
National cholera projection scenarios for 2020-2070 at decadal intervals
Sub-national
As the sub-national projections could account for the known spatial heterogeneity in PDSI projections stated above. Sub-national monthly PDSI projections were calculated using the projected 2050 and 2070 environmental data. First, potential evapotranspiration (PET) (mm/day) was calculated with the temperature data and latitude using the Hargreaves method [37], where Ra is the mean extra-terrestrial radiation in mm/day, which is a function of latitude, and T represents daily air temperature in °C (package “SPEI” [38]).
PDSI is the output of a supply-and-demand model of soil moisture, which can be calculated using PET and precipitation. A common critique of PDSI is that the behaviour of the index varies by location, making spatial comparisons difficult. The self-calibrated (scPDSI) methods accounts for this by automatically calibrating the behaviour of the index at any location by replacing empirical constants in the index computation with dynamically calculated values [39]. Therefore, scPDSI was used to provide PDSI values for 2050 and 2070 for the three RCP scenarios. All calculations were completed in R Studio version 4.1.0 (package “scPDSI” [40]).
The scenarios for temperature and poverty followed the same pattern and targets as those in the national scenarios. Sanitation was based on SDG6.2 (achieve access to adequate and equitable sanitation and hygiene for all and end open defection by 2030) and conflict scenarios were guided by SDG16.1 (Significantly reduce all forms of violence and related death rates everywhere). The SDGs for both sanitation and conflict are particularly ambiguous, regardless of this difficulty, the sanitation and conflict targets were based on a similar pattern to MPI, achieving universal access to sanitation and conflict elimination by 2070 in S1 and a 50% decrease in sanitation access and 50% increase in conflict events by 2070 in S5. A summary of the sub-national scenarios is shown in Table 3.
Sub-national cholera projection scenarios for 2020-2070 at decadal intervals
Results
Historical Analysis
Fig 2 shows the time-series for the WHO and GHDx cholera deaths data for 1970-2016, compared to the mean values for Africa (average of 0.1 cases/100,000). The time-series highlights the historically high cholera burden in Nigeria, which is particularly evident in the GHDx data. Comparing the two datasets highlights many differences, showing the importance of considering multiple cholera data sources for analyses. The GHDx data illustrates a steep decline from 16.0 cases/100,000 in 1991 to 1.8 cases/100,000 in 2017, with some plateauing periods. Whereas for the WHO data, cholera appears relatively stable over time (average of 0.5 cases/100,000), instead witnessing large peaks, particularly in 1971 (5.1 cases/100,000), 1991 (7.8 cases/100,000) and 1999 (1.7 cases/100,000). Furthermore, the difference in death rate overall should be noted here, with the GHDx data reporting much higher deaths per 100,000.
The linear trends and loess curves for the cholera data illustrates a flat trend in the WHO data, whereas the GHDx data shows a steady decrease (S2 Fig). ACF for both datasets show a slow decay (gradual decrease) to within the confidence interval bands (S3 Fig). The ACF decay suggests that the effect of time is not particularly significant in either of the cholera datasets. The weak effect of time was also illustrated in the ARIMA analysis, which showed a flat forecast, predicting at the current rate of cholera deaths.
For the environmental and social risk factors analyses here (meteorological drought, temperature, water withdrawal, poverty, conflict, and sanitation), data incompleteness meant that calculating correlations were difficult (S4 Fig). Assumptions and averages of the data had to be taken when fitting the models used here (see S2 Text). Regardless of data limitations, cholera deaths had a strong negative correlation between sanitation access and a slight positive correlation with temperature from 1970 to 2016 (Fig 3).
Correlation matrix based on Pearson correlation coefficients for the six commonly selected covariates analysed here against the WHO cholera deaths data. Positive coefficients are blue suggesting a strong positive association between the corresponding row and column and negative coefficients are in red, suggesting a negative association. The ? represents a negligible value due to data incompleteness
Scenario Projections
National
Fig 4 shows the national cholera projections (in cholera outbreak occurrence, 0-1) to 2070 for the five scenarios, with 95% confidence intervals. National cholera occurrence for Nigeria decreased from S1 to S3, starting at 0.95 for all scenarios to 0.83 for S1 and 0.92 for S3 by 2070. These changes were relatively small and several of the confidence intervals overlap (the linear relationship, with standard error is presented in S5 Fig). The decreases to 2070 are minimal, especially when considering the levels of development in the S1 scenario. For S4 and S5, where socio-economic and environmental conditions regressed, there was an increase in cholera outbreak occurrence from 0.95 to 0.98 for S4 and 0.99 for S5.
National cholera projections for Nigeria, in cholera outbreak occurrence (0-1), for the five scenarios to 2070
Sub-national
For the sub-national projections, measured in terms of cholera R, there were several spatial heterogeneities, which may help explain some of the uncertainty seen in the national projections. Generally, R values decreased through the three time points shown (Fig 5) for S1, S2 and S3, with the number of states with R values over 1 decreasing, particularly for S1 to S3. For S4 and S5, the changes appear more complex, with some states faring better than others when faced with worsening social and environmental conditions. The south of the country had particularly high R values in these less optimistic scenarios, whereas the north saw little change and, in some cases, a slight improvement.
Sub-national cholera projections for Nigeria, in cholera reproduction number (R), for the five scenarios (S1 - orange, S2 - blue, S3 - green, S4 - red and S5 - purple) at 3 of the decadal time points (2030, 2050 & 2070)
Discussion
Using the historical evidence and future projections, this research has shed light on the future of cholera in Nigeria to 2030 and beyond. The historical data highlighted the historically high cholera burden in Nigeria, compared to the rest of Africa. The data also showed the wide range of cholera values that are reported and the importance of evaluating multiple data sources. The cholera projections provided a more detailed understanding of future trends and if socio-economic development and climate change mitigation could reduce cholera in Nigeria. Both the national and sub-national projections showed decreases in cholera burden with the more optimistic scenarios (S1-S3). Under S4 and S5, the national cholera burden worsened, despite the already high cholera outbreak occurrence, showing the need for continued development. Further explanation of the historical and future trends of cholera in Nigeria and how this could inform the GTFCC Roadmap are provided below.
Evidence from the Historical Data
The WHO data presented large outbreaks and peaks through the instrumental period, whereas the GHDx data steadily decreases but with higher cholera deaths overall. The historical cholera trends do not suggest a significant increase or decrease based on previous burden. The ARIMA forecasting and ACF suggested a weak relationship with time in the two datasets analysed. The results show no clear trend to suggest whether cholera is increasing or decreasing and whether the 2030 targets can be met from these data alone.
The lack of relationship between cholera and time found here further highlights the importance of understanding cholera risk factors and their temporal changes over time. Understanding these risk factors going forward could help to inform cholera burden and identify hotspots. For example, in 2020 more than half of the population were without access to safe sanitation [41], putting billions of people at risk of cholera. It is essential to continue monitoring changes in these risks, identify areas of prioritisation and collect accurate data.
Sanitation had the strongest correlation with the WHO dataset used here and was selected in several of the models used over other metrics of WASH, including access to clean water and hand-washing facilities. Progress has been made in the last 20 years (2000-2020) in terms of expanding access to WASH services in Nigeria, with the percentage of access to improved sanitation increasing from 52% to 62% and an expansion in access to improved drinking water by 31% [5]. However, if access to sanitation continues to increase by 10% every 20 years, this would result in over 87 million Nigerians without access by 2030 and therefore at a high risk of cholera (based on a 5% increase in access and a population projection of 264 million by 2030 [18].
In Nigeria, there is a divide in terms of WASH and development between the northern and southern regions of the country (S6 Fig) [42]. Northern states are generally more rural, which has potentially resulted in less development. For example, there is a 40% deficit in access to sanitation in the rural compared to the urban population. Nigeria’s rural population comprises 47% (99,895,289, 2021) of the total population, putting millions of people at risk of cholera in these potentially less developed areas [43]. However, in the last sixty years the rural population has decreased by 38% and with effective urban planning, this continued trend could significantly reduce cholera in Nigeria.
Evidence from the Scenario Projections
The national scenarios showed clear trends in terms of both improvements from S1 (lowest cholera occurrence) to S3 and regression in S4 and S5 (highest cholera occurrence). The projections showed that continued progress towards and beyond the SDGs (in particular SDG1 and 6) and emissions reductions (contributing to PDSI and temperature) would help to improve global health and particularly cholera. However, by 2030, in all the national projection, cholera was not close to the 2030 targets (90% reduction). By the end of the projection period (2070), with significant improvements in development and environmental protection, cholera was far from eliminated and eradication will likely take time, due to the pathogen circulating in the population and environmental reservoirs and via introductions [44,45].
The sub-national projections were more optimistic regarding the achievability of the GTFCC targets. Overall, for the S1 sub-national projections there was a decrease in R values to less than 1 by 2050 and for the southern states all S1 cholera projections were less than 1 by 2030 (compared to the northern states which were all above 1). An explanation for this is that more time will be needed in the northern states to reach the required development for significant transmission reductions. Additionally, in the northern states, development was already low and conflict already high, therefore worsening these conditions did not significantly impact cholera.
The projections suggest that southern Nigeria could potentially reach the 2030 targets and eradication may be possible in the future. Whereas the north of the country must be an area of prioritisation in terms of development, cholera response and conflict resolution. The sub-national projections additionally highlighted how vital it will be for the development and peace achieved in the south to continue or at a minimum, remain the same. As previously stated, the worsened conditions of S4 and S5 had a large impact on the southern states in terms of increasing cholera transmission. Decreasing levels of peace and development would potentially be catastrophic in the southern states and overall, to the cholera burden in the country.
An Update on the 2030 Targets and Roadmap
Based on the work presented here, the achievability of the GTFCC targets at the current pace of cholera control and development appears unlikely to be met by 2030 in Nigeria. Despite this, there has been significant progress in terms of health and development in Nigeria and the southern states appear far more likely to meet these goals. A new proposed 2050 target, building on the GTFCC Roadmap and expanding on the three axes will now be delineated. The new targets will aim at bringing the northern states to the same levels of development and peace achieved in the south.
Axis 1
Axis 1 largely focuses on surveillance and data, which are vital in target setting, allocation of resources and response. Improved surveillance may also help to reduce the cholera data inconsistencies found here. The axis focuses on early detection, but with cholera this is difficult, due to a large number of mild infections and several other diarrhoeal pathogens potentially causing disease (e.g., shigella, typhoid, dysentery), meaning a positive test does not always mean that cholera is the causative agent.
A method to help improve reporting in Nigeria, would be to offer incentives to test and report. Financial incentives have proved effective at improving health outcomes in Nigeria and are often cost-effective in the long-term as they prevent serious disease and morbidity [46,47]. To reduce nosocomial transmission and prevent testing hesitancy due to stigmatisation and restrictions [17], modification and improvements to the highly effective cholera rapid diagnostic test [48], allowing them to be used at home, may be helpful. The tests would need to be easy to use and report, inexpensive and widely available.
Emphasis is needed both at a government and academic level on improving data quantity and quality. Understanding reporting effort and the accuracy and precision of data are key areas of future research to fully understand how well the current data are representing cholera burden. At a global level, a metric of reporting effort would help when comparing disease data that has been collected across multiple countries and therefore with different methods and uncertainty.
Furthermore, risk factor data are needed to fully understand disease dynamics and plan for effective response and interventions. The results here suggest that improvements in tracking poverty and sanitation are good areas of prioritisation. In Nigeria, benefit could also be gained from testing environmental reservoirs, such as major lakes and rivers, which are known to be used for washing and drinking and can be fundamental in cholera transmission [49,50]. Better understanding on the environmental burden would be useful for both research and to understand local risk factors.
Axis 2
Axis 2 (cholera interventions) is arguably the most important area for reaching the GTFCC goals and several other health targets. The Roadmap highlights the need for long-term sustainable WASH implementation and strengthening of healthcare systems to anticipate cholera outbreaks (e.g., capacity building of staff, resources, diagnostics, education and societal engagement and emergency WASH intervention). However, the GTFCC Roadmap and previous research on cholera interventions heavily focuses on outbreak response [16], rather than development.
The Roadmap suggests that interventions should target states most at risk, with the analysis presented here suggesting northern states as a priority. Additionally, healthcare should be strengthened more generally, with greater resources and service availability, making healthcare an attractive career option to ensure sufficient human resources [51,52]. Development planning and targets must also consider that global crises can cause regression of progress, increasing the need to strive beyond health targets. For example, COVID-19 is estimated to have erased four years of progress against poverty and caused disaster-related deaths to rise sixfold [29].
Designating significant financial resources on outbreak response, is not a cost-effective way of reaching cholera targets, although fundamental to reducing mortality in outbreaks. More emphasis needs to be placed on improving peoples’ quality of life, lifting them out of poverty, providing them with basic services and empowering them to improve their own health through resources and education. In the absence of this development, outbreaks will continue to occur, spending financial resources in a reactionary way.
Axis 3
Axis 3 of the Roadmap involves commitment and coordination on a global level, across several sectors. Nigeria Centre for Disease Control (NCDC) currently works across multiple levels of the national systems and has a detailed response plan for diarrhoeal (including cholera) outbreaks titled, “Preparedness and Response to Acute Watery Diarrhoea Outbreaks” [53]. NCDC have a designated team working on cholera elimination as a priority within the country. Continued and increased funding to NCDC will be vital for them to continue their work toward cholera control. At a global level, a “One World - One Health” approach is needed to prevent pandemics and achieve the GTFCC targets at this level. Recent pandemics and global outbreaks (e.g., COVID-19 and monkeypox) have shown the catastrophic results of countries not working together in a joint effort to control disease [54].
Nigeria has made several gains in weakening the Boko Haram stronghold in the northeastern states. However, the conflict continues to threaten Nigeria’s security and several previous studies have suggested the negative impacts of conflict on health [54-57]. Bottom-up stabilisation efforts are working to address local level drivers of insecurity, including strengthening local conflict prevention, restoring governance and services and fostering social cohesion. While at a regional level, the Lake Chad Basin Commission and African Union Commission have highlighted short-, medium- and long-term stabilisation, resilience and recovery needs [58-60]. Fig 6 illustrates the current Roadmap and summarises the suggestion made here to improve cholera control beyond 2030 and achieve the GTFCC targets in Nigeria by 2050.
The 2030 GTFCC Roadmap for cholera elimination (black) with additional suggestions and areas of prioritisation (blue) for 2050 in Nigeria
Limitations
Data incompleteness and inconsistencies were issues when trying to evaluate the historical data. Incomplete data resulted in either the analysis not being completed (e.g., some of the correlations) or removing and averaging the data (both used in the data fitted to the models, S2 Text). As stated above, improving surveillance and a greater effort to collect data on cholera risk factors will be very important for target setting and resource allocation and prevent duplication of services [61]. Multiple data sources were used here to try and account for this issue, and in future cholera research, using sensitivity analysis and testing cholera assumptions across multiple data sources is one method to understand these differences.
All scenario projections have limitations, due to the uncertainty in trying to predict future conditions, along with the limitations of models. The wide range of future scenarios helps to account for some of this uncertainty but will still not be sufficient in capturing all potential future environments. For example, the scenarios here are uni-directional, either getting better or worse from current conditions. All social and environmental drivers either getting better or worse is unlikely, with some metrics improving and some worsening. To add further complexity, these changes could be spatially heterogeneous. Regardless of these limitations, this should not discourage scenario projection analysis, as it is still useful and valid in terms of understanding future changes and helping to inform cholera prevention and policy.
A further limitation is the different outcome variables and methodologies used in the projection models. It could be argued that two model types (generalised linear models and random forest) and two outcome variables (outbreak occurrence and R), make the results here difficult to compare and present different error and limitations (S2 Text). The GTFCC targets largely focus on reducing cholera deaths, which is why deaths (death rate) were chosen as the cholera metric for the historical data analyses. Although the discrepancies create difficulties in interpretation, to reach the 2030 targets and subsequently reduce deaths by 90%, global burden will have to substantially decrease, regardless of the metric used. Therefore, the projections are still useful in presenting the required decrease in burden.
Conclusions
In Nigeria, the GTFCC targets look difficult to achieve by 2030 at a national level, based on these results. However, the more urban and developed southern states may reach and go beyond these targets. There is a vital need for continued investment in long-term development, especially to reduce regional inequity in northern Nigeria. Despite the financial capital needed to improve healthcare, WASH and education, these interventions are cost-effective due to their wide-reaching impacts.
The results presented here highlight the importance of and how modelling studies can be used to inform cholera policy and the potential for this to be applied to other diseases and countries. The aim of health research should be to improve global health and quality of life and applying quantitative research to policy can help achieve this, while increasing the relevance of the research.
If the GTFCC targets are met in Nigeria, this will reduce the risk of cholera for hundreds of millions of people and greatly reduce the global burden of diarrhoeal disease and mortality in children under 5 years. Progress to go beyond the goals and targets set would be highly beneficial in combating the impact of global shocks and crisis. Continued progress in development (especially at an accelerated pace) and sustainable urban planning, could greatly improve Nigeria’s cholera burden and health status in the coming decades.
Data Availability
The cholera data used in the analysis are available from the WHO Global Health Observatory and the Global Health Data Exchange, while subnational data were shared by NCDC and can be requested. Details of social and environmental data are available, including a list of sources, in Table 1.
Supporting information
S1 Text. Additional information about the scenarios used for the projections
Projected temperature data were available through WorldClim, which gives projections for 2050 and 2070, under the four Representative Concentration Pathways (RCP) emissions scenarios. RCP2.6 requires emissions to start declining by 2020 and reach net zero by 2100, RCP4.5 requires emissions to peak in 2040, RCP6.0 peaks at around 2080 and RCP8.5 assumes emission will continue to rise throughout the century. These four scenarios are projected to have mean global warming by 2081-2100 of 1°C, 1.8°C, 2.2°C and 3.7°C, respectively (IPCC, 2014). RCP2.6 was not used in the projections, as this requires CO2 emissions to peak in 2020, a goal which has not been reached (NOAA, 2020). The data was presented as a monthly average at the admin 1 level, which was then transformed into a national yearly mean.
Projected PDSI data was difficult to obtain, and previous modelling studies have found spatial heterogeneity, making projecting drought across the continent challenging. There is also disagreement over drought changes in Africa under climate change and how populations will adapt to alterations in water scarcity (Ahmadalipour et al., 2019; Calow et al., 2010; Haile et al., 2020; Shanahan et al., 2009; Touchan et al., 2008; Verschuren et al., 2000). One of the main difficulties in calculating projected PDSI and discrepancies in results is the algorithm used to calculate PET (Tian-Jun and Tao, 2013). PET algorithms do not account for changes in vegetation cover expected due to elevated CO2, making the “warming leads to drying” narrative flawed and over-simplistic, causing overestimations (Yang et al., 2020). Several studies which have taken a more critical approach of the methodology found that drought did not significantly change over long periods and changes are likely to be seen on finer spatial scales (Sheffield et al., 2012; Yang et al., 2020). These findings are consistent with tree-ring (Touchan et al., 2008), lake sediment records (Verschuren et al., 2000) and the PDSI dataset used here. This suggests that droughts are neither more severe nor longer now than historically. Despite the large number of projected drought studies, extracting these results and using them here is challenging, as they are on different spatial scales, using different methodology and data. There are also issues with the projected indices used, as these often vary and are potentially not comparable. To account for this, three scenarios were created from historical PDSI data using univariate linear regression models for drought and year. This method used the coefficients for each country to project the future drought data, with scenario 3 continually plotting the coefficient increase until 2070 (or it reached −4 or 4, the extremes of the PDSI scale), scenario 1 accepting the above hypotheses that PDSI is an overestimation and drought will not change over the projected period and scenario 2 as a median value.
Projected population data was based on the United Nation’s World Population Prospectus (2019). The projections are based on available data on population size, levels of fertility, mortality and international migration. Data is from censuses, registration of births and deaths, demographic and health surveys, official statistics and population registers. More recently, the data has considered refugee statistics, prevalence of HIV and antiretroviral coverage, infant and under five morality and migration flows. The projections use the cohort component method using a variety of demographic assumptions concerning fertility, mortality and migration. This considers the past experiences of the country and reflects uncertainty and other countries in similar conditions. The medium variant projection corresponds to the median of several trajectories of each demographic component derived using the probabilistic model of the historical variability over time. Prediction intervals represent the spread in the distribution of outcomes across the projected trajectories and thus provide and assessment of the uncertainty inherent in the medium variant projection. Therefore, only the medium variant projection was included in the model and not as an offset.
Projected poverty headcount at <$1.90 a day was based on Sustainable Development Goal (SDG) 1 (UN, 2015), which states that by 2030, extreme poverty (<1.25/day) will be eliminated and the population living in poverty will be reduced by >50%. This goal is very ambitious and will require significant human and economic resources. Several of the terms within the SDGs are also ambiguous, meaning the aims and roadmap to achieve them are not clear. With regards to poverty, the setting used in the SDGs is slightly lower ($1.25 compared to $1.90) and it is difficult to distinguish the level of poverty within the data; therefore, the projected scenario will mainly align with the second part of the goal, to halve the population in poverty by 2030. Despite their limitations, the SDGs provide a globally recognised pathway to a sustainable future and what institutions and governments should be aiming to achieve, making their use important in scientific research.
Creating projected freshwater withdrawal per capita data for the scenarios also presented challenges, as this is largely down to human behaviour and therefore hard to predict. While SDG6.4 states that by 2030, water use efficiency will be sustainably increased across all sectors. Projected data is not freely available and may not consider climate change and alterations in societal behaviour, all of which will alter water stress in the future. To maintain sustainable levels of water resources, rates of withdrawal need to be lower than replenishment. Renewable resources come from internal river flows and groundwater from rainfall. To understand the national historical water security, data was plotted for both freshwater withdrawal and freshwater resources (Ritchie, 2017). The figure helps to illustrate that most countries in Africa have both low resources and use, making the likely use of freshwater sustainable in most countries. Exceptions to this are Gabon, Republic of Congo and Liberia which have comparatively high resources and low withdrawal as well as Madagascar, Libya, Sudan, Mauritania and Morocco which have comparatively low resources and high use. Given the relationship between cholera outbreaks and water withdrawal, most countries in Africa could increase their freshwater withdrawal, except for Madagascar, Libya, Sudan, Mauritania and Morocco which should reduce their use, to improve sustainability.
Both are taken as national per capita averages for the full dataset (1985-2014) in m3 and transformed on a logarithmic scale. The dashed line represents the cutoff for countries to be categorised as high withdraw (375m3) [23]
For A, the WHO data [19] and B, the GHDx data [20]. The dashed blue line represents the confidence interval, with ACF measured as the correlation coefficient of the residuals. The lag is set to 10log10(N/m), where N is the number of observations and m the number of series. ACF plots are bar charts of coefficients of correlation between a time series and it lagged values, with a confidence band
Correlation coefficient (r) represents the correlation between the covariate and the WHO cholera deaths data [19] and the p-values. Most are not found to be significant, potentially due to a lack of complete data
S2 Text. Limitations of the models used in the scenario projections. Including methods of how data incompleteness was handled before model fitting
Data Incompleteness
Climate, drought and socio-economic data were missing for several countries and years, meaning that data had to be averaged or omitted, potentially introducing error.
Cholera data has issues with underreporting, issues may have arisen from assigning the outcome variable to zero for missing years, as this could have led to the underrepresentation of cholera outbreaks (a sensitivity analysis for this assumption showed this to be the best possible outcome).
Climate data were missing for Co^te d’Ivoire and drought data were missing for Rwanda, The Gambia, Guinea-Bissau, Djibouti, Burundi, Benin, Cabo Verde, Sa Tome’ and Principe, Comoros, Mauritius and Seychelles. Environmental data for these countries were derived by taking the mean of their neighbouring countries, whereas islands were excluded.
Where a country’s socio-economic data were missing for some years, a national average was taken from the available data points and used for all years. If national data were missing for the full instrumental period, these countries were removed from the analysis.
After examining data completeness across the full dataset, we designated the instrumental period for analysis to be 2000–2016 to limit omitting missing data and interpolation.
Methodological Limitations
Seasonal changes in cholera and the presence of waterbodies within countries facilitating transmission were not considered in this model.
Generalised linear models assume a monotonic relationship and therefore non-linear effects of several covariates might not be captured and evaluating these non-linear effects are a potential area of future work.
Data Incompleteness
The data were subset to only include cases that were confirmed either by rapid diagnostic tests or by laboratory culture, this removed many cases. A sensitivity analysis proved that removing suspected cases did not bias the results.
The drought indices were used to measure relative dryness/wetness, not long-term drought changes, due to the short timescale of the cholera surveillance dataset (2 years).
The relatively short timescale (2 years) and and spatial scale (6 states meeting the case threshold) of the cholera data means the results are more accurate at presenting cholera at the current time in Nigeria, rather than historically and caution is needed when making generalisable conclusions.
During times of crisis, cholera may also be over-reported or more accurately represent the cholera burden in the area. This is due to the presence of cholera treatment centres, increased awareness among the population and healthcare workers and external assistance from non-governmental organisation, detecting cases that may have been missed previously.
Methodological Limitations
Both long-term and short-term changes to the population may take time before changes in cholera transmission are evident. While some influences may be considered slow-onset or rapid-onset and therefore defining their beginning is subjective. However, calculating R on monthly sliding windows and using monthly covariate data helped to reduce potential lagged effects on the R values.
Average values for the full dataset by state, for A, percentage access to sanitation and B, Multidimensional Poverty Index (MPI). The sources and timescales of the data are shown in Table 1
Acknowledgements
We would like to thank and acknowledgement all those who collected and curated the datasets used here, including the Nigeria Centre for Disease Control for their direct involvement. This work was supported by the Natural Environmental Research Council [NE/S007415/1], as part of the Grantham Institute for Climate Change and the Environment’s (Imperial College London) Science and Solutions for a Changing Planet Doctoral Training Partnership. We also acknowledge joint Centre funding from the UK Medical Research Council and Department for International Development [MR/R0156600/1].
References
Sources
- 12.
