Climate resilience of public health preventive and adaptive measures against diarrhea in Northern Ghana; a case study of the Tamale Metropolitan Area

Abstract

Background

Diarrhea remains one of the biggest public health threats in Ghana, and it is the most common cause of both morbidity and mortality among children in sub-Sahara Africa. Several preventive and adaptative public health measures such as Rota Virus vaccination and improving access to potable water are being implemented. There is little research in Ghana examining the climatic resilience of these preventative and adaptive interventions. The study aimed at determining whether the preventative and adaptive interventions are climate-resilient by using morbidity data.

Methods

In this research, time series data of monthly all-cause diarrhea morbidity, rainfall, temperature, and relative humidity which span the period of January 2014 to December 2020 were analyzed. The study used the Auto Regression Distributed Lag cointegration approach to model the impact of the climatic variables on all-cause diarrheal morbidity.

Results

Public health preventative and adaptation strategies are climate robust and unresponsive to changes in climatic variables. The residuals in the models estimated are white noise and do not exhibit serial autocorrelation or conditional heteroscedasticity.

Conclusion

In the Tamale Metropolitan Area, climate change has no effect on diarrhea morbidity, supporting the claim that climate resilience is an important component of public health preventive interventions against diarrhea. The study findings as a whole emphasize the significance of climate-sensitive, evidence-based public health strategies.

Introduction

Globally, diarrhea is a serious public health issue and it is still the main cause of morbidity and mortality worldwide. There is evidence that diarrhea accounts for one billion illness episodes and 1.8 million fatalities annually [1]. The majority of these fatalities take place in developing nations, where it is estimated that diarrheal illness is directly responsible for 25% of under-five mortality [2, 3]. Infectious organisms such as viruses, bacteria, protozoa, and helminths that are passed from one person’s excrement to another’s mouth are what causes diarrhea. Many likely still need to be discovered as some are well known, others have only lately been discovered, or are emerging novel agents. They differ in the load of organisms required to induce infection and illness as well as the path from the feces to the mouth [4].

The prevalence of acute diarrheal illnesses has not decreased significantly despite the widespread use of oral rehydration therapy, even though the adoption of oral rehydration therapy is responsible for a significant drop in yearly diarrheal mortality from 4.5 million in the 1980 s to the current level of roughly 2 million children [5]. This is large since the majority of the world’s population, particularly in developing nations lack access to clean water and is still plagued by destitution, filthy conditions, and a lack of cleanliness [6]. In Ghana, diarrhea is a major contributor to morbidity and mortality, with an incidence of 20% in children under five and an estimated 7% of diarrhea-specific mortality.

There are numerous and intricate transmission routes via which climatic conditions affect diarrheal illnesses. Flooding, frequently brought on by an excessive amount of rain, is associated with a rise in the prevalence of diarrheal disease [7, 8]. Flushing pathogens from environmental reservoirs or feces into freshwater supplies is one way that rainfall runoff and flooding can expose people to pathogens [9, 10]. Conversely, water scarcity caused by long periods of drought may force people to drink contaminated water and practice unhygienic behavior, which increases the risk of diarrheal disease [11]. Several studies in sub-Sahara Africa discovered an increased in incidence related to both the dry and wet seasons [12–14], lower-than-average rainfall [11] downpours [15] and intermittent rainfall [16]. Increased rates of diarrheal incidence have been linked to rising temperatures, considering that increased pathogen growth in food and water sources may be a result of warming temperatures [17, 18]. Empirical evidence from [19] and [11] revealed that temperature was a significant predictor of the incidence of diarrheal disease with a One Degree Celsius rise in temperature leading to a 7% increase in all-cause diarrheal disease.

There is a myriad of intervention strategies against diarrheal disease in Ghana, In April 2012, Ghana became one of the first countries in Africa to incorporate the monovalent rotavirus vaccine into its routine childhood immunization program [20]. The vaccine had a 56% efficacy in double-blind randomized controlled studies. However, this efficacy value is less than the 70% to 92% achieved in high-income countries. About 53% of Ghana’s population had access to portable water as of 2008 with urban populations enjoying a disproportionally higher access. With financial assistance from the International Monetary Fund, Ghana’s water supply has undergone a transition process that has involved the adoption of fiscal, regulatory, and institutional reforms to improve access [21]. Piped sources and mechanized boreholes are the primary sources of portable water in the Tamale metropolitan area [22]. There are still significant gaps in our knowledge of the sensitivity of climate change predictors and associated risks to match current diarrheal morbidity in less developed nations and environments with inadequate health infrastructure, the regular gathering of long-term disease data necessary to carry out such assessments is resource-intensive and difficult, as such there isn’t enough quality empirical evidence on the nexus between climatic variables and diarrheal disease in Ghana. A few studies in different geographical locations in Ghana have however established the sensitivity of all causes of diarrhea to climatic factors [23–26]. This study seeks to quantify the climatic resilience of public health initiatives to combat diarrheal disease in northern Ghana. The results of this research will contribute to enhancing the level of local and global preparedness as well as understanding of climatic conditions and the health threats they pose.

Materials and methods

Study site

The Tamale Metropolitan Area was selected for the study as a result of the cultural and socioeconomic diversity in the Metropolis and the availability of reliable data on the study variables, Tamale Metropolitan Area is located in the northern region of Ghana. Geographically, Tamale is located on latitude 9˚ 23’N and longitude 0˚ 50’W. It is the administrative capital of the Northern Region of Ghana. The Metropolitan Area occupies approximately 750 sq. km, equivalent to 13% of the total land area of the Northern Region. Figure 1 below shows the geographical map of the Tamale Metropolitan Area.

Fig. 1.

Geographical map of Tamale Metropolitan Area. (source: (Ghana Statistical Service, 2013)

Data types and sources

Monthly time series data on all-cause diarrhea spanning the period of January 2014 to December 2020 were obtained from the District Health Information Management System managed by the Ministry of Health. Data on temperature, rainfall, and relative humidity spanning the period of January 2014 to December 2020 were obtained from the Northern Regional Meteorological Agency.

Trend analysis

Trend analysis depicts the series’ elevation or dips over time. This study estimated 3 trends, thus, linear, log-linear, and quadratic trends. To determine which of the three trends best describes the series, the three distinct models were evaluated.

A time trend in a time series is a linear function of time t. The model is given by;

1

where is the actual value at time is the error term and, , are the regression coefficients of the actual values on time. A linear trend denotes a steady increase or decrease in a series.

The empirical specification will be as follows if there is a linear trend in all-cause diarrhea disease;

2

The empirical specification will be as follows if climate variables (temperature, rainfall, and relative humidity) empirically show a linear trend.

3

4

5

If the data display a quadratic trend, the model is stated as follows;

6

In this equation, is the actual value at time is the error term and, , are the regression coefficients of the actual values on time. A series that exhibits a quadratic trend implies that the growth or decline rate of a series is not constant.

If all-cause of diarrhea disease exhibits a quadratic trend, the empirical specification will be as follows;

7

Similarly, if climatic variables (Relative Humidity, Temperature, Rainfall) exhibit a quadratic trend, the empirical specification will be as follows;

8

9

10

The model’s log-linear representation is shown as;

11

from the equation above, is the actual value at time is the error term and, , are the regression coefficients of the actual values on time.

If all-cause of diarrhea disease exhibits a log-linear trend, the empirical specification will be as follows;

12

Empirically, if climatic variables (Relative humidity, Temperature, Rainfall) exhibit a log-linear trend, the empirical specification will be as follows;

13

14

15

Unit root test

To analyze the unit root properties of the series, the Augmented Dickey-Fuller Test (ADF) test and the Kwiatkowski-PhillipsSchmidt-Shin (KPSS) Test were used in the study.

Cointegration test

Cointegration is used to investigate the long-run correlation between two-time series observations. The concept of co-integration was advanced through the ground-breaking works of Engle and Granger, (1987) and also in subsequent works of Johansen, (1988) and Soren Johansen and Juselius, (1990). The cointegration approach has an inherent superiority over the linear regression approach concerning the test of association as the likelihood of spurious correlations as a result of mere coincidence or confounders is non-existent.

The traditional Engle and Granger, (1987) cointegration method do not apply in instances where the variables are integrated of different orders, in such instances the Autoregressive Distribution Lag (ARDL) cointegration method is a more appropriate method in determining the long-run relationship between time variables irrespective of the order of integration. The results of the ARDL approach are more efficient and realistic given those spontaneous and long-run associations between variables obtained from the parametrized results. If at least one cointegration vector is found, the reparametrized version of the ARDL method of cointegration is called the Equilibrium Correction Model (ECM).

The ARDL model is specified below:

16

, assume the lag order q is equal for all variables. The variables are then allowed to be only integrated of order zero, I (0), and integrated purely of order one, I [1], or a combination of different orders of integrations.

By minimizing a specific model selection criterion, such as Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), or Hannan-Quinn Information Criteria (HQIC), one can arrive at the ideal lag orders p and q.

When error correction is specified, reparameterization is performed as follows:

17

Ethical consideration

Ethical approval for this study was granted by the ethical review committee of the Committee on Human Research, Publication, and Ethics (CHAPRE) of the Kwame Nkrumah University of Science and Technology.

Results

Summary statistics of variables

As indicated in Table 1 for the climatic variables, relative humidity had an average value of 58.8, with maximum and minimum values of 82.5 and 20 diarrhea cases stood at 1043 with maximum and minimum values of 2065 and 212 cases, respectively. The average temperature of the study period stood at 29.48 °C with maximum and minimum temperature values of 34.2 °C and 26.4 °C, respectively. Rainfall had an average value of 84.29 mm with maximum and minimum values of 343.6 mm and 0 mm, respectively. Rainfall had the highest variability followed by relative humidity and temperature, with relative humidity being negatively skewed whereas temperature and rainfall were positively skewed and all climatic variables being platykurtic. Average diarrhea was positively skewed and platykurtic.

Table 1.

Descriptive statistics of study variables

Variable	Relative Humidity (%)	Temperature (°C)	Rainfall (mm)	Diarrhea
Mean	58.86	29.48	84.29	1043
Median	64.50	29.00	51.65	1046
Minimum	20.50	26.40	0.00	212
Maximum	82.50	34.20	343.6	2065
STD	19.57	1.95	88.75	429.4
CV	0.332	0.066	1.053	0.411
Skewness	−0.477	0.49	0.918	0.201
Ex. Kurtosis	−1.169	−0.75	−0.164	−0.405

Number of Data points = 84

Trend analysis

Results on the trend analysis as shown in Table 2 reveal that the log linear trend best fits both climatic variables and diarrhea given the lowest AIC and BIC values.

Table 2.

Trend analysis of variables

Model	R-squared	Adj R-squared	AIC	BIC
Relative Humidity
Linear	0.001	−0.011	740.9	745.8
Quadratic	0.002	−0.023	742.9	750.2
Log-Linear	0.001	−0.024	88.55	95.84
Temperature
Linear	0.009	−0.002	352.5	357.3
Quadratic	0.017	−0.008	353.9	361.2
Log-Linear	0.017	−0.008	−216.7	−209.4
Rainfall
Linear	0.007	−0.005	994.4	999.3
Quadratic	0.007	−0.018	996.4	1004
Log-Linear	0.053	0.021	206.7	213.2
Diarrhea
Linear	0.087	0.075	1254	1257
Quadratic	0.449	0.435	1212	1219
Log-Linear	0.423	0.409	77.98	85.27

As shown in Fig. 2 below, the series plot for relative humidity for the study period indicate a seasonal pattern with no consistent trend and no outliers as well as sudden shifts in the series.

Fig. 2.

Trend of Relative Humidity

As indicated in Fig. 3 below, temperature exhibits a fluctuating trend that slowly wanders upwards. The series plot indicates seasonality with no obvious outliers and no sudden shift in trend with the peak temperature recorded around may of 2017.

Fig. 3.

Trend of Temperature

As indicated in Fig. 4 below, rainfall exhibits no consistent trend. The series plot for the time period indicates seasonality with no outliers and sudden shifts.

Fig. 4.

Trend of Rainfall

As indicated in Fig. 5, diarrhea cases exhibits a fluctuating pattern with no consistent trend. The series plot indicates no obvious outliers no sudden shift in trend no seasonal pattern.

Fig. 5.

Trend of Diarrhea

Unit root test

The unit root test is conducted to determine the stationarity properties of a series. This study employed the Kwaitkowski-Philips-Schmidt-Shin test (KPSS) test whose null hypothesis posits that the series is stationary whereas the alternative hypothesis posits that the series is not stationary. If the test statistic exceeds the critical value during a KPSS test, the null hypothesis is to be rejected. If the critical value is smaller than the critical value, the null hypothesis is not rejected. The KPSS test was fitted for constant as well as constant and trend for all of the series. The KPSS test results, which are shown in Table 3, reveal that for the model fitted with constant as well as constant and time trend, a significant test statistic is higher than the critical value at 5% significance level for just diarrhea. This indicates that the null hypothesis which indicates that the series is stationary is not rejected for all the series except diarrhea. Thus, relative humidity, temperature, and diarrhea are integrated of order zero, I (0).

Table 3.

KPSS test of the series in levels

Variable	Test with Constant		Constant + Trend
	Test Statistic	Critical Value	Test Statistic	Critical Value
Relative Humidity	0.019	0.467	0.016	0.148
Temperature	0.083	0.467	0.032	0.148
Rainfall	0.048	0.467	0.019	0.148
Diarrhea	0.624	0.467	0.393	0.148

Given the fact that the series for diarrhea was not integrated of order zero I (0), the study went further to examine the stationarity of the series after the first differencing. As indicated in Table 4, the findings of the KPSS test of the first difference of the series show a significant test less than the critical value for both the constant model as well as the constant and time model. This indicates that all the series are integrated of order one I [1].

Table 4.

KPSS test of the first difference of the series

Variable	Test with Constant		Constant + Trend
	Test Statistic	Critical Value	Test Statistic	Critical Value
Relative Humidity	0.035	0.467	0.02	0.148
Temperature	0.018	0.467	0.012	0.148
Rainfall	0.033	0.467	0.021	0.148
Diarrhea	0.084	0.467	0.029	0.148

Auto-regressive distributed lag (ARDL) cointegration

After examining the unit root properties of the series, the study performed the Autoregressive Distributed Lag cointegration test. The conventional Johansen and Juselius, (1990) cointegration procedures are not relevant in this case because the study variables are integrated of different orders, hence the ARDL cointegration approach is used in the study. Using the Bounds test of Cointegration, the dynamic correlation between diarrhea and climatic variables is examined. The null hypothesis is rejected if the computed F- statistic value exceeds the 5% upper bound. We are however unable to reject the null hypothesis if the computed F-statistic value is below the upper bound critical value at the 95% confidence level.

The findings in Table 5 show that diarrhea and relative humidity do not cointegrate, given that the computed F-statistic in modeling the association between diarrhea and relative humidity is 4.107, which is lower than the upper bounds at 90%, 95%, and 99% significant levels. Similarly, the computed F-statistic in modeling the relationship between temperature and diarrhea of 4.329 is lower than the upper bounds at 90%, 95%, and 99% significant levels which is indicative of the fact that the series are not cointegrated. The findings further reveal that rainfall and diarrhea are not cointegrated, given that the computed F-statistic of 4.256 is lower than the upper bounds at 90%, 95%, and 99% significant levels.

Table 5.

ARDL cointegration test between diarrhea and climate variables

Model	90% CL	95% CL	99% CL	p-value
	I (0) I (1)	I (0) I (1)	I (0) I (1)	I (0) I (1)
	4.095 4.860	5.032 5.875	7.163 8.148	0.015 0.037
	Computed F-statistics		Decision
Diarrhea and Relative humidity	4.107		No cointegration	Significant
Diarrhea and Temperature	4.329		No cointegration	Significant
Diarrhea and Rainfall	4.256		No cointegration	Significant

Model diagnostics

The study then conducts the model diagnostic test after the cointegration test. As indicated in Table 6, given that the p-values for each of the models used to examine the association between climatic factors and diarrhea in the study were insignificant, it can be concluded that all of the residuals in the model demonstrate that they are free from serial autocorrelation and heteroscedasticity. As a result, the error term is stated to be homoscedastic and the model’s residuals are free from serial autocorrelation. This suggests that the model’s error terms can be interpreted as white noise.

Table 6.

Ljung-Box test and ARCH-LM test of model between diarrhea and Climatic variables

Model	Lags	Ljung Box Test		ARCH-LM Test
		Test Statistic	P-Value	Test Statistics	P-Value
Diarrhea and Relative Humidity	5	14.21	0.014	7.609	0.179
	10	15.64	0.111	12.89	0.229
	15	15.63	0.155	19.39	0.196
Diarrhea and Temperature	5	14.43	0.013	7.253	0.203
	10	21.49	0.179	11.71	0.305
	15	21.75	0.264	19.51	0.192
Diarrhea and Rainfall	5	4.224	0.518	4.402	0.493
	10	11.06	0.354	8.959	0.536
	15	11.21	0.427	12.22	0.662

Discussion

Overall, the results suggest that climate resilience is an important component in public health prevention measures against diarrhea in Ghana. We found in this study that there was a consistent seasonal trend in the relative humidity contrasting previous findings [27], who noted that relative humidity in Manga, Ghana’s Upper East, exhibited an erratic pattern with no discernible trend, a plausible explanation of this variation could be the range of the data set used for this study. Evidence from [28] seems to point to minimal seasonal variation in relative humidity patterns. Instead, over an extended period, the patterns shift gradually. Similarly, temperature demonstrated a steady increase over time without any sudden changes in trend, which is consistent with past observations. The steady rise in average temperature has also been noted by [29, 30]. reported on seasonal temperature variations as well, but no trends were discovered. Rainfall also displayed seasonality without any abrupt changes or clear trends, which is in line with a previous study [31]. However, it is important to note that some studies have reported varying or conflicting results regarding these variables’ trends and patterns. For instance, some studies have suggested that rainfall trends in Ghana may be declining over time probably as a result of natural climate variability [32, 33].

Age-related trends and seasonal patterns may play a role in the transmission of intestinal parasites, as temperature and relative humidity affect parasite survival in the environment. Results from present study revealed that all-cause diarrhea morbidity exhibited a fluctuating pattern with an overall downward trend. The observation is consistent with findings from [34] which found that diarrhea rates typically dropped from November to December and rose from January onward but showed an overall decreasing trend over the five-year period of the study.

These observations were contrasted with those of other investigations carried out in other parts of the world. Over the course of these studies, there have been conflicting reports as to which climatic variables are associated with diarrhea. For instance [35], in an investigation to quantify the correlation between climate variability and diarrhea-related morbidity in subtropical Taiwan, found that the diarrhea-related morbidity was greatly impacted by relative humidity. Similarly, in their study to assess how weather conditions affect infectious diarrhea [36], found that relative humidity was associated with an increased risk for infectious diarrhea in Guangzhou [37]. In their study to evaluate the climate-diarrhea nexus in Mekong Delta also found a significantly positive relationship between increased diarrhea and relative humidity however; our study did not find a significant association. A null association was observed between temperature and diarrhea which is similar to previous findings [38–40]. Our study further found that rainfall did not significantly influence diarrhea similar to findings from [40], whereas other studies have found it to be a significant predictor [11, 35, 41]. It is important to note that the effects of climatic factors on diarrhea may be influenced by several other factors, such as population characteristics [42], and socioeconomic factors [43]. Therefore, further studies with comprehensive data collection methods are needed to fully understand the complex relationship between climatic factors and diarrhea incidence.

While it is important to acknowledge the efforts being made to combat diarrhea in Ghana, some may argue that there are limitations to the effectiveness of these preventive and adaptive interventions. For instance, while access to potable water may improve as a result of these interventions, climate change can lead to water scarcity and contamination, thereby undermining the impact of these interventions. Additionally, while vaccinations like the Pneumococcal and Rota Virus vaccine can help prevent diarrhea, these vaccines may not be accessible to all, particularly in rural areas or low-income households. Furthermore, studies have shown that climate change can increase the risk of diarrhea by affecting rainfall patterns and temperatures [39]. A previous study examined relationships between climatic factors and the prevalence of age-specific diarrheal illnesses in Taiwan and found diurnal temperature and high precipitation to be associated with all cause infectious diarrhea, with the highest risk observed for bacterial and viral diarrheal among population under five years. Similarly [11], analyzed the effects of changes in temperature and precipitation on the prevalence of diarrhea in children under three years old in sub-Saharan Africa and found that in the sub-Saharan Africa, an absence of rain during the dry season increases the prevalence of diarrhea. When rainfall is ordinary or below normal compared to the monthly long-term average, such shortages take place in many places. Another study found that while an increased in monthly minimum temperature decreased diarrheal disease, an increased in monthly average maximum temperature increased the prevalence of diarrhea. Further [41], in a prospective cohort study examined if increased diarrhea from all causes and water contamination were related to high temperatures and heavy rain, their findings suggested that pathogens that build up in dry conditions may be spread to children by heavy rain ultimately leading to diarrhea their findings also suggested that Higher temperatures were positively associated with diarrhea 1–3 weeks later. Moreover, the success of these interventions could also depend on factors beyond climate resilience such as education and awareness campaigns about safe hygiene practices. It is essential to recognize that these preventive and adaptive interventions are not a panacea for diarrhea and may require complementary strategies to address the complex challenges associated with this issue. Overall, the study underscores the importance of evidence-based public health policies that are climate sensitive. The findings of this study will improve knowledge of climatic conditions and the risks they pose to public health as well as the state of local and global preparedness. There is also the need for future research into other household and external factors that could potentially impact diarrheal morbidity.

Conclusion

This study set out to investigate the climate resilience of the public health prevention and adaptation measures against diarrhea in the Northern Region of Ghana using time series data. The results of the trend analysis for the climatic variables reveal a trend that does not differ much from other studies. Diarrhea showed a declining trend while being erratic. The pre-modeling analysis of the data reveals that some of the series are integrated of a different order. The ARDL estimation finds that at 95% and 99% confidence level, there is no cointegration between diarrhea morbidity and climate variables, which is indicative of the fact that climate change have neither short nor long-run impact on diarrhea morbidity in the Tamale Metropolitan area, demonstrating the assertion that climate resilience is a feature of public health preventive measures against diarrhea. As this study only uses data from passive surveillance systems, there is a potential bias of under- or over-reporting the number of cases of diarrhea that may result from data collection and processing. This is because some facilities may not have well-trained data experts, which reduces the validity and reliability of the data collected. Monthly data on individual, household and external factors that could potentially impact diarrheal morbidity are not available at the metropolitan and regional level. Better systematic data collection and surveillance systems need to be put in place to identify unreported diarrheal cases, offer precise case definitions, and shed light on any potential confounding or moderating factors. The findings of the study may improve both local and global preparedness as well as knowledge of climate conditions and the health risks they entail.

Authors’ contributions

AGZ and SZ jointly conceived and designed the study, and AGZ and YA performed data analysis and interpretation. AGZ drafted the manuscript under the supervision of SZ. AGZ, SZ, and YA jointly undertook critical revision of the manuscript. All authors have approved the manuscript for publication.

Funding

The research was self-sponsored.

Data availability

The datasets used for this study are available from the corresponding author on request.

Declarations

Ethics approval and consent to participate

Ethical approval for this study was granted by the ethical review committee of the Committee on Human Research, Publication, and Ethics (CHAPRE) of the Kwame Nkrumah University of Science and Technology with reference number CHAPRE/AP/442/21.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.