Listen

Consumer satisfaction and drinking water availability in evolving urban settings are major issues and their understanding is a key to improving water service quality and managing water utility. This study aims to examine factors influencing subscribers' satisfaction with drinking water availability in the Municipality of Bujumbura, Burundi. To achieve this, a survey was carried out from 3 to 21 May 2021 on a random sample of 391 subscribers. The outcome variable was the logarithm of the odds of being satisfied with drinking water availability and explanatory variables included sociodemographic factors (commune, gender, age, education level, marital status, and religion), water distribution factors and co-production-related factors. The chi-square test (Fisher's exact test if conditions are not met) and binary fixed-effects logistic regression models were applied to these data using R software, version 4.1.2. Findings showed that the time of drinking water supply and frequency of water-related outages significantly influenced satisfaction with drinking water availability after adjustment for age. This study could help decision-makers who are in charge of drinking water distribution identify other factors associated with water availability satisfaction in Bujumbura Municipality and other urban settings.

Listen

  • Drinking water availability satisfaction rates were computed.

  • Logistic regression model was performed.

  • Predicting probabilities of being satisfied with drinking water availability was done.

  • The contribution of each of the selected variables in the explanation of the satisfaction with drinking water availability helped prioritize them.

Bujumbura, chi-square test, logistic regression, satisfaction, water availability
Listen
BIC

Bayesian Information Criterion

OR

odds ratio

aOR

adjusted odds ratio

ROC

receiver operating characteristic

AUC

area under curve

CI

confidence interval

UN

United Nations

SDG

sustainable development goal

REGIDESO

Régie de Production et de Distribution d'Eau et d’Électricité (Burundi Water and Power Supply Company)

CV

contribution of a variable

DRU

drinking water utility

df

degrees of freedom

Listen

Worldwide, water service relies on water supply and helps people to satisfy their daily water needs. Water consumption is an indicator of water needs, as well as water availability which relies on the following key pillar parameters: land use and land cover, rainfall, lithology, lineaments, drainage density, geomorphology and topography (Mohammed & Sahabo 2015; Mandy et al. 2020). Specifically, water public service in most Sub-Saharan African countries including Burundi is facing a lot of challenges due essentially to population boom, galloping urbanization rate, rehabilitation and extension of water facilities. This part of Africa is expected to be the next hotspot of water scarcity according to findings from a study conducted elsewhere (Baggio et al. 2021). Two-thirds of the population is estimated to suffer from water shortage worldwide by 2025 (Kama et al. 2023). In cases of water shortage or lack of drinking water, people are likely to be exposed to waterborne illnesses and make long distances to fetch water, an indication of bad socioeconomic conditions (Chard et al. 2019; Ahmad et al. 2022; Salehi 2022; Zheng et al. 2022). Children and mothers are then given the task of collecting water. As wished by the United Nations (UN), ensuring water availability and sustainable management of water and sanitation contributes to achieve the sixth of the Sustainable Development Goals (SDGs), especially the first target which consists of achieving access to safe drinking water for all at a low cost by 2030, a challenging objective for several African countries including Burundi (United Nations 2015).

Statistical methods have been applied to data related to drinking water services. In fact, chi-square tests and multilevel logistic regression models were used to analyze the impact of intermittent water supply on satisfaction with water supply in two Chinese provinces (Shandong and Hubei) (Li et al. 2020). This study showed, among other findings, that households did not have sufficient water. A study conducted in Kisumu city (Kenya) found that the gender of the head of the household did not influence satisfaction with water delivery, but socioeconomic factors did (Ocholla et al. 2022). Besides, a study conducted in Southern Sri Lanka used logistic regression models and found that socioeconomic parameters significantly influenced the satisfaction of water consumers (Ellawala & Priyankara 2016). Using ordinary logistic regression, a study conducted in Ethiopia showed that subscribers were ready to pay more if water service quality was improved (Tessema 2020).

In Burundi, the urban water supply as well as the one in rural settings is most of the time intermittent, which means that both of them often undergo interruptions. Water used in urban households is provided by REGIDESO (Burundi Water and Power Supply Company). Rapid population growth is not in accordance with fresh water and food production (Okello et al. 2015; Megerle & Niragira 2020). For Bujumbura Municipality settlements, drinking water is provided mainly by surface water (around 90% from Lake Tanganyika), springs (7%), and groundwater (3%). In other second cities such as Gitega and Ngozi, springs and boreholes are the main sources of drinking water. The largest and biggest water project was planned to meet water needs by 2005. Because of the lack of return-investment due to imprioritization of water projects against security matters in 2000 years (period of civil war), water production became and is still insufficient for the entire urban population, leading to rationing practices in order to achieve fairness in water supply. Our study area (Bujumbura Municipality) was chosen because it is the largest and the oldest area, and possesses a large number of drinking water subscribers and other private investors. These private companies are known for their brands. The most famous are namely Kinju, Kandi, Sangwe, Aquavie, Eagle, Jibu, Life and Saana.

Currently, there is no REGIDESO competitor in supplying drinking water because water costs are extremely low in the country. As a matter of fact, REGIDESO is a state-owned and highly subsidized company. To the best of our knowledge, there is no study which combined the analysis of determinants of satisfaction with drinking water availability in Bujumbura Municipality and probabilities predictions. Our study aims to determine factors influencing satisfaction with water availability in Bujumbura Municipality using a fixed-effects logistic regression model and predict probabilities of being satisfied with drinking water availability.

Listen

Study area

Listen
Figure 1 displays the map of selected households in Bujumbura Municipality. Created in 1897 by Germans, Bujumbura Municipality is the capital city of Burundi. Bounded by Mutimbuzi commune to the north, Isare and Kanyosha communes to the east, Kabezi commune to the south and Lake Tanganyika (altitude 778 m2) to the west, Bujumbura Municipality is the main city and the economic capital of Burundi, Gitega being the political capital of Burundi. All those communes are parts of Bujumbura province. Bujumbura Municipality is located in the east of the Democratic Republic of Congo, between 3°30′ and 3°51′ south latitude and between 29°31′ and 29°42′ east longitude. Its population size increased from 497,166 to 743,514 inhabitants from 2008 to 2020 and has continued to increase rapidly till now (Institute of Statistics & Economic Studies of Burundi 2017). With an area of 114 km2, Bujumbura Municipality is densely populated. It is subdivided into three communes: Muha contains three zones (Kanyosha, Kinindo, Musaga), Mukaza contains four zones (Rohero, Buyenzi, Bwiza, Nyakabiga) and Ntahangwa contains six zones (Ngagara, Gihosha, Buterere, Cibitoke, Kamenge, Kinama).
Figure 1
Study area.
View largeDownload slide

Study area.

Data collection

Listen

A survey was conducted on co-production and water consumer satisfaction from 3 to 21 May 2021 in Bujumbura Municipality, the capital city of Burundi. Co-production involves filling out forms for public drinking water utilities, submitting a water consumption index or other information, being a member of a decision-making process regarding water improvement service and using new technologies to pay bills. Subscribers of Burundi Water and Power Supply Company (REGIDESO) were interviewed using a questionnaire. The number of subscribers was distributed as follows in the three communes of Bujumbura: 32,707 in Ntahangwa, 11,875 in Mukaza and 20,780 in Muha.

The sample size n was calculated using Krejcie and Morgan's formula (Uakarn et al. 2021):
(1)
where denotes the chi-squared statistics (⁠⁠) with one degree of freedom for a large sample, N the number of subscribers (N = 65,362), p the proportion of subscribers who are satisfied with drinking water availability (⁠⁠) and d the acceptable error margin (⁠⁠). This yields a basic sample size of 382 subscribers. About 5% of contingencies were added to this sample size, leading to a total sample size of 396. These subscribers were selected with a probability proportional to the number of subscribed households in each commune. The non-response rate was 1.3% (5/396).

Outcome and independent variables

Listen

The satisfaction with drinking water availability was coded as: 1 = not at all satisfied, 2 = not satisfied, 3 = does not know, 4 = a little bit satisfied, 5 = satisfied, 6 = very satisfied, 7 = extremely satisfied. Due to low counts in the modalities, the first three categories of this variable and the last four ones were grouped to form a binary outcome (0 = not satisfied, 1 = satisfied). Hence, the outcome variable was the logarithm of the odds of being satisfied with water availability. Explanatory variables were commune (1 = Muha, 2 = Mukaza, 3 = Ntahangwa), gender (1 = male, female), age class (in years) (1 = 18–31, 2 = 32–40, 3 = 41–51, 4 = 52 and more), education level (1 = None, 2 = Primary, 3 = Secondary, 4 = Higher), marital status (1 = Single, 2 = Married, 3 = Divorced/widowed), religion (1 = Catholic, 2 = Protestant, 3 = Muslim), occupation (1 = None, 2 = Civil servant, 3 = Trades people or businessman, 4 = Other), mode of supply (1 = Private tap, 2 = Public tap), Time of water supply (1 = Daytime, 2 = Night, 3 = Both), frequency of water-related outages (1 = Rarely, 2 = Often/sometimes/everyday), water access issues (0 = No, 1 = Yes), housing (1 = House, 2 = Villa, 3 = City), authorization to complete forms for the drinking water utility (1 = Disagree, 2 = Neutral, 3 = Agree), authorization to complete or submit water consumption index or other information about the drinking water utility (1 = Disagree, 2 = Neutral, 3 = Agree), having been a member of a decision-making process concerning the improvement of the public drinking water service (1 = Disagree, 2 = Neutral, 3 = Agree), carrying out a task in a formal or informal way falls within respondent's responsibilities (1 = Disagree, 2 = Neutral, 3 = Agree), authorization to use new technologies (1 = Disagree, 2 = Neutral, 3 = Agree) and waiting time (in min) to receive drinking water service (1 = 0–9, 2 = 10–19, 3 = 20–34, 35–180).

Statistical analysis

Listen
The rate of satisfaction with water availability was computed globally and according to individual characteristics. Additionally, the chi-square test was used to check independence between the dependent variable and each explanatory variable. The null hypothesis of independence was rejected if the p-value was lower than ⁠. In cases where the expected number of subscribed households in cells was lower than 5 or the marginal numbers were too unbalanced, Fisher's exact test was used. The Cramer's V coefficient was used to measure the strength of the relationship between two categorical variables. Taking its values from 0 to 1, this coefficient is given by the square root of the ratio of the observed and the maximum chi-square statistics. Besides, the maximum chi-square statistics is given by the sample size times the minimum of the number of rows minus one and the number of columns minus one for a contingency table. The relationship between the dependent variable and each explanatory variable was considered null or very weak for V < 0.10, weak for 0.10 ≤ V < 0.20, middle for 0.20 ≤ V < 0.30, and strong for 0.30 ≤ V < 1.00. A fixed-effects logistic regression model was used to examine factors predicting satisfaction with water availability. Model parameters were estimated using the maximum likelihood method. To assess the significance of these parameters, Wald's test was used. Variables were considered significant in univariate models if the p-value was lower than 20%. This p-value represents the probability of obtaining a statistic greater than or equal to what was observed under the null hypothesis. The crude odds ratio (OR), defined as the ratio of the odds of satisfaction with water availability for two given events, was obtained by applying the exponential function to the model parameters. The 95% confidence interval (CI) was also derived to assess the practical significance of the odds ratios. Significant variables from univariate logistic regression models were used in the multivariable (full) logistic regression model (Cox 1958):
(2)
where p stands for the probability of being satisfied with water availability, the intercept, model parameters, explanatory variables and an error term. Taking into account other factors, an adjusted odds ratio (aOR) and its 95% CI were computed for each parameter, except for the intercept. At a significance level of 5%, a backward elimination based on the parsimony principle was used to select models. The saturated model was compared to the empty model (without explanatory variables) and the full model (containing significant variables from univariate models). Model selection was based on the Bayesian Information Criterion (BIC):
(3)
where L denotes the likelihood of the model, k the number of explanatory variables in the model and n the number of observations. The lower the BIC, the better the model. Taking into account other explanatory variables in the model, McFadden's adjusted R2 was computed as the complement of the ratio of the likelihood of the saturated model minus the number of explanatory variables and the likelihood of the empty model. The model adequacy was checked by the global model significance test using Wald's test and the model was validated using the link test. The null hypothesis of the link test states that the additional linear predicted value squared used to rebuild the model is equal to zero. Contrary to the linear predicted value, this additional squared term should be not statistically significant. The Receiver Operating Characteristic (ROC) Curve, which is a graphical tool for evaluating and comparing the performance of the model, plots the true positive rate (sensitivity) on the y axis and the false positive rate (1-specificity) on the x axis in a scatter plot and was used to assess the discriminatory performance of the saturated model. Besides, it helped to compare the performance and predictive power of the selected model. The area under curve (AUC), a measure of the predictive power of an explanatory variable, was estimated. There was no discrimination between satisfied and unsatisfied subscribers for AUC = 0.5, lower discrimination for ⁠, acceptable discrimination for ⁠, excellent discrimination for ⁠, exceptional discrimination for and perfect discrimination for (the gold standard) (Swets 1979; Hajian-Tilaki 2013). In case of ⁠, predicted probabilities of drinking water availability satisfaction can be computed.

To prioritize factors associated with water availability satisfaction, the contribution of a variable (CV) to explaining drinking water availability satisfaction was calculated as the difference between the chi-square of the final model with and without the variable divided by the chi-square of the final model with the variable. Points were considered outliers if the Cook's distance was greater than 4/n and the studentized residuals outside the interval [−2, 2] where n denotes the number of observations. The Welsh-Kuh's distance was computed for each of the 391 observations as a difference in fits (DFFITS) and was used to check influential points. The threshold or cut-off value was given by two times the square root of (p + 2) divided by n, where p is the number of parameters. We also used the Hoaglin and Welsh criterion based on leverage points. For this criterion, a point was considered to be of high leverage if its leverage (diagonal element of the projection matrix, i.e. the hat matrix) was higher than two times (p + 1) divided by n. Data were analyzed using R software, version 4.1.2.

Listen

Sample size (n), number of customers satisfied with water availability globally and according to sociodemographic characteristics (n+), proportions (%) and their 95% confidence intervals (CI) are summarized in Table 1.

Table 1

Sociodemographic characteristics

CharacteristicCategoriesnn +%95% CI
Commune Ntahangwa 197 135 68.5 62.0–75.0 
Mukaza 64 45 70.3 59.0–81.6 
Muha 130 80 61.5 53.1–70.0 
Gender Male 182 130 71.4 64.8–78.0 
Female 209 130 62.2 55.6–68.8 
Age (years) 18–31 92 52 56.5 46.3–66.7 
32–40 103 69 67.0 57.8–76.1 
41–51 90 60 66.7 56.8–76.5 
52 + 106 79 74.5 66.2–82.9 
Marital status Single 75 46 61.3 50.2–72.5 
Married 284 190 66.9 61.4–72.4 
Divorced/widowed 32 24 75.0 59.7–90.3 
Education level None 53 29 54.7 41.1–68.3 
Primary 66 47 71.2 60.2–82.3 
Secondary 129 84 65.1 56.8–73.4 
Higher 143 100 69.9 62.4–77.5 
Religion Catholic 237 160 67.5 61.5–73.5 
Protestant 105 63 60 50.6–69.4 
Muslim 49 37 75.5 63.3–87.7 
Occupation None 67 44 65.7 54.2–77.2 
Civil servant 107 74 69.2 60.3–78.0 
Businessman 160 107 66.9 59.5–74.2 
Other 57 35 61.4 48.6–74.2 
Overall  391 260 66.5 61.8–71.2 
CharacteristicCategoriesnn +%95% CI
Commune Ntahangwa 197 135 68.5 62.0–75.0 
Mukaza 64 45 70.3 59.0–81.6 
Muha 130 80 61.5 53.1–70.0 
Gender Male 182 130 71.4 64.8–78.0 
Female 209 130 62.2 55.6–68.8 
Age (years) 18–31 92 52 56.5 46.3–66.7 
32–40 103 69 67.0 57.8–76.1 
41–51 90 60 66.7 56.8–76.5 
52 + 106 79 74.5 66.2–82.9 
Marital status Single 75 46 61.3 50.2–72.5 
Married 284 190 66.9 61.4–72.4 
Divorced/widowed 32 24 75.0 59.7–90.3 
Education level None 53 29 54.7 41.1–68.3 
Primary 66 47 71.2 60.2–82.3 
Secondary 129 84 65.1 56.8–73.4 
Higher 143 100 69.9 62.4–77.5 
Religion Catholic 237 160 67.5 61.5–73.5 
Protestant 105 63 60 50.6–69.4 
Muslim 49 37 75.5 63.3–87.7 
Occupation None 67 44 65.7 54.2–77.2 
Civil servant 107 74 69.2 60.3–78.0 
Businessman 160 107 66.9 59.5–74.2 
Other 57 35 61.4 48.6–74.2 
Overall  391 260 66.5 61.8–71.2 
View Large

Globally, the rate of satisfaction with water availability is 66.5% (95% CI: [61.8, 71.2]). Divorced or widowed subscribers are more likely to be satisfied with water availability. Mukaza subscribers are highly satisfied with water availability (70.3; CI = [59.0, 81.6]) compared to Muha and Ntahangwa subscribers, but the differences in proportions are not statistically significant (61.5; CI = [53.1, 70.0] versus 68.5; CI = [62.0, 75.0]).

Sample size, number of customers satisfied with water availability globally and according to water-related characteristics, proportions and their 95% confidence intervals are summarized in Table 2.

Table 2

Water-related characteristics

CharacteristicCategoriesnn +%95% CI
Mode of water supply Private tap 337 230 68.2 63.3–73.2 
Public tap 54 30 55.6 42.1–69.0 
Time of water supply Daytime 27 11 40.7 21.8–59.7 
Night 125 48 38.4 29.8–47.0 
Both 239 201 84.1 79.4–88.8 
Frequency of water-related outages Rarely 276 194 70.3 64.9–75.7 
Often/sometimes/everyday 115 66 57.4 48.3–66.5 
Water access issues No 317 225 71.0 66.0–76.0 
Yes 74 35 47.3 35.8–58.8 
Housing Ordinary housing 214 151 70.6 64.4–76.7 
Villa 102 62 60.8 51.2–70.3 
City 75 47 62.7 51.6–73.7 
Overall  391 260 66.5 61.8–71.2 
CharacteristicCategoriesnn +%95% CI
Mode of water supply Private tap 337 230 68.2 63.3–73.2 
Public tap 54 30 55.6 42.1–69.0 
Time of water supply Daytime 27 11 40.7 21.8–59.7 
Night 125 48 38.4 29.8–47.0 
Both 239 201 84.1 79.4–88.8 
Frequency of water-related outages Rarely 276 194 70.3 64.9–75.7 
Often/sometimes/everyday 115 66 57.4 48.3–66.5 
Water access issues No 317 225 71.0 66.0–76.0 
Yes 74 35 47.3 35.8–58.8 
Housing Ordinary housing 214 151 70.6 64.4–76.7 
Villa 102 62 60.8 51.2–70.3 
City 75 47 62.7 51.6–73.7 
Overall  391 260 66.5 61.8–71.2 

n: number of subscribers; n + : number of satisfied subscribers; %: satisfaction rate; CI: confidence interval.

View Large

This table shows that subscribers who live in ordinary housing are more satisfied with the availability of water than those who live in a villa or a city (70.6%; CI = [64.4, 76.7]), but the differences in proportions are not significant. Besides, subscribers who are more likely to be satisfied with the availability of water are those who draw water from a private tap (68.2%; CI = [63.3, 73.2]), who receive water day and night in their households (84.1%; CI = [79.4, 88.8]), who rarely experience water-related outages (70.3%; CI = [64.9, 75.7]), who do not have water access issues (71.0%; CI = [66.0, 76.0]), who have internal and external shower and water closet layouts (75.1%, CI = [68.8, 81.5]); 74.9%, CI = [68.5, 81.2] respectively) and who have ordinary housing (70.6%; CI = [64.4, 76.7]).

Sample size, number of customers satisfied with water availability globally and according to co-production- and quality-related characteristics, proportions and their 95% confidence intervals are summarized in Table 3.

Table 3

Co-production and quality-related characteristics

CharacteristicCategoriesnn +%95% CI
Authorization to complete forms for the drinking water utility (DRU) Disagree 192 127 66.1 59.4–72.9 
Neutral 98 67 68.4 59.1–77.7 
Agree 101 66 65.3 56.0–74.7 
Authorization to submit water consumption index to the DRU Disagree 197 130 66.0 59.3–72.6 
Neutral 82 49 59.8 49.0–70.5 
Agree 112 81 72.3 64.0–80.7 
Having been a member of a decision-making concerning DRU improvement process Disagree 201 130 64.7 58.0–71.3 
Neutral 114 76 66.7 57.9–75.4 
Agree 76 54 71.1 60.8–81.3 
Carrying out a task in a formal or informal way falls within one's responsibilities Disagree 160 107 66.9 59.5–74.2 
Neutral 120 79 65.8 57.3–74.4 
Agree 111 74 66.7 57.8–75.5 
Authorization to use new technologies Disagree 198 130 65.7 59.0–72.3 
Neutral 87 55 63.2 53.0–73.4 
Agree 106 75 70.8 62.0–79.5 
Waiting time (in min) to receive drinking water-based complaints 0–4 101 74 73.3 64.6–82.0 
5–9 106 62 58.5 49.0–67.9 
10–14 76 52 68.4 57.9–79.0 
15 + 108 72 66.7 57.7–75.6 
Waiting time (in min) to receive a service 0–14 132 86 65.2 57.0–73.3 
15–29 117 87 74.4 66.4–82.3 
30 + 142 87 61.3 53.2–69.3 
Overall  391 260 66.5 61.8–71.2 
CharacteristicCategoriesnn +%95% CI
Authorization to complete forms for the drinking water utility (DRU) Disagree 192 127 66.1 59.4–72.9 
Neutral 98 67 68.4 59.1–77.7 
Agree 101 66 65.3 56.0–74.7 
Authorization to submit water consumption index to the DRU Disagree 197 130 66.0 59.3–72.6 
Neutral 82 49 59.8 49.0–70.5 
Agree 112 81 72.3 64.0–80.7 
Having been a member of a decision-making concerning DRU improvement process Disagree 201 130 64.7 58.0–71.3 
Neutral 114 76 66.7 57.9–75.4 
Agree 76 54 71.1 60.8–81.3 
Carrying out a task in a formal or informal way falls within one's responsibilities Disagree 160 107 66.9 59.5–74.2 
Neutral 120 79 65.8 57.3–74.4 
Agree 111 74 66.7 57.8–75.5 
Authorization to use new technologies Disagree 198 130 65.7 59.0–72.3 
Neutral 87 55 63.2 53.0–73.4 
Agree 106 75 70.8 62.0–79.5 
Waiting time (in min) to receive drinking water-based complaints 0–4 101 74 73.3 64.6–82.0 
5–9 106 62 58.5 49.0–67.9 
10–14 76 52 68.4 57.9–79.0 
15 + 108 72 66.7 57.7–75.6 
Waiting time (in min) to receive a service 0–14 132 86 65.2 57.0–73.3 
15–29 117 87 74.4 66.4–82.3 
30 + 142 87 61.3 53.2–69.3 
Overall  391 260 66.5 61.8–71.2 

n: number of subscribers; n + : number of satisfied subscribers; %: satisfaction rate; CI: confidence interval.

View Large

The largest proportions of customers satisfied with water availability are observed among subscribers who waited an average of between 15 and 29 min to receive service from REGIDESO (74.4%; 95% CI = [66.4, 82.3]), and those who waited at least 4 min before their complaints are received at reception (73.3%; 95% CI = [64.6, 82.0]). All satisfaction rates are above 50% and there are no significant differences in water availability satisfaction across categories of variables.

The chi-square test or Fisher exact test rejected the null hypothesis of independence between satisfaction with water availability and time of water supply, frequency of water-related outages, water access issues, shower layout and water closet layout. The relationship between water availability satisfaction and occupation (Cramer's V= 0.793) and the relationship between water availability satisfaction and authorization to complete forms for the drinking water utility (DRU) (V= 0.984) were very strong. The relationship was strong between water availability satisfaction and time of water supply (V= 0.468) and chances to find a value equal to 0.468 at random were very weak (p-value < 0.001).

Significant factors associated with water availability satisfaction at a level of 20% in univariate logistic regression models were gender (p-value = 0.054), age class (p-value = 0.070), educational level (p-value = 0.191), religion (p-value = 0.147), mode of water supply (p-value = 0.069), time of water supply (p-value < 0.001), frequency of water-related outages (p-value = 0.014), housing (p-value = 0.169), waiting time (in min) to receive a service (p-value = 0.081), and waiting time (in min) to receive drinking water-based complaints (p-value = 0.158). These variables were considered in the multivariable logistic regression model (Table 4).

Table 4

Multivariable logistic regression model

CharacteristicCategoriesaOR95% CIp-value
Gender    0.219 
 Male 1.00   
 Female 0.71 0.42–1.22 0.219 
Age class (years)    0.017 
 18–31 1.00   
 32–40 1.85 0.90–3.80 0.095 
 41–51 2.07 0.98–4.39 0.056 
 ≥ 52 3.57 1.62–7.91 0.002 
Education level    0.171 
 None 1.00   
 Primary 2.88 1.06–7.77 0.037 
 Secondary 2.03 0.84–4.91 0.116 
 Higher 2.46 0.99–6.13 0.054 
Religion    0.872 
 Catholic 1.00   
 Protestant 1.09 0.60–1.98 0.775 
 Muslim 1.25 0.51–3.05 0.621 
Mode of water supply    0.597 
 Private tap 1.00   
 Public tap 0.81 0.36–1.79 0.597 
Time of receiving water    <0.001 
 Daytime 1.00   
 Night 0.75 0.29–1.93 0.548 
 Both 8.43 3.21–22.09 <0.001 
Frequency of water-related outages    0.004 
 Rarely 1.00   
 Often/sometimes/everyday 0.43 0.24–0.77 0.004 
Housing    0.108 
 Ordinary housing 1.00   
 Villa 0.54 0.30–0.99 0.047 
 City 1.08 0.52–2.26 0.836 
Waiting time (in min) to receive a service    0.090 
 0–14 1.00   
 15–29 1.85 0.94–3.66 0.076 
 30 + 0.93 0.49–1.79 0.836 
Waiting time (in min) to receive drinking water-based complaints    0.723 
 0–4 1.00   
 5–9 0.81 0.39–1.69 0.572 
 10–14 1.15 0.51–2.6 0.734 
 15 + 1.19 0.55–2.55 0.663 
CharacteristicCategoriesaOR95% CIp-value
Gender    0.219 
 Male 1.00   
 Female 0.71 0.42–1.22 0.219 
Age class (years)    0.017 
 18–31 1.00   
 32–40 1.85 0.90–3.80 0.095 
 41–51 2.07 0.98–4.39 0.056 
 ≥ 52 3.57 1.62–7.91 0.002 
Education level    0.171 
 None 1.00   
 Primary 2.88 1.06–7.77 0.037 
 Secondary 2.03 0.84–4.91 0.116 
 Higher 2.46 0.99–6.13 0.054 
Religion    0.872 
 Catholic 1.00   
 Protestant 1.09 0.60–1.98 0.775 
 Muslim 1.25 0.51–3.05 0.621 
Mode of water supply    0.597 
 Private tap 1.00   
 Public tap 0.81 0.36–1.79 0.597 
Time of receiving water    <0.001 
 Daytime 1.00   
 Night 0.75 0.29–1.93 0.548 
 Both 8.43 3.21–22.09 <0.001 
Frequency of water-related outages    0.004 
 Rarely 1.00   
 Often/sometimes/everyday 0.43 0.24–0.77 0.004 
Housing    0.108 
 Ordinary housing 1.00   
 Villa 0.54 0.30–0.99 0.047 
 City 1.08 0.52–2.26 0.836 
Waiting time (in min) to receive a service    0.090 
 0–14 1.00   
 15–29 1.85 0.94–3.66 0.076 
 30 + 0.93 0.49–1.79 0.836 
Waiting time (in min) to receive drinking water-based complaints    0.723 
 0–4 1.00   
 5–9 0.81 0.39–1.69 0.572 
 10–14 1.15 0.51–2.6 0.734 
 15 + 1.19 0.55–2.55 0.663 

aOR, adjusted odds ratio; CI, confidence interval.

View Large

The BIC of this full model was 498.72. The following variables were gradually removed from the full model using a backward selection: religion (p-value = 0.872), waiting time (in min) to receive drinking water-based complaints (p-value = 0.740), mode of water supply (p-value = 0.534), gender (p-value = 0.204), housing (p-value = 0.151), educational level (p-value = 0.136) and waiting time (in min) to receive a service (p-value = 0.107). The BIC of the saturated model (438.31) was lower than the BIC of the empty model (504.64) and of the full model (498.72). The time of drinking water supply and frequency of water-related outages significantly influenced satisfaction with drinking water availability after adjustment for age (Table 5).

Table 5

Saturated logistic regression model

CharacteristicCategoriesaOR95% CIp-value
Age class (years)    0.010 
 18–31 1.00   
 32–40 2.09 1.05–4.16 0.035 
 41–51 2.04 1.01–4.13 0.047 
 ≥ 52 3.31 1.63–6.71 0.001 
Time of water supply    <0.001 
 Daytime 1.00   
 Night 0.87 0.36–2.08 0.757 
 Both 9.02 3.77–21.56 <0.001 
Frequency of water-related outages    0.001 
 Rarely 1.00   
 Often/sometimes/everyday 0.40 0.23–0.68 0.001 
CharacteristicCategoriesaOR95% CIp-value
Age class (years)    0.010 
 18–31 1.00   
 32–40 2.09 1.05–4.16 0.035 
 41–51 2.04 1.01–4.13 0.047 
 ≥ 52 3.31 1.63–6.71 0.001 
Time of water supply    <0.001 
 Daytime 1.00   
 Night 0.87 0.36–2.08 0.757 
 Both 9.02 3.77–21.56 <0.001 
Frequency of water-related outages    0.001 
 Rarely 1.00   
 Often/sometimes/everyday 0.40 0.23–0.68 0.001 

aOR, adjusted odds ratio; CI, confidence interval.

View Large

McFadden pseudo-R2 is 0.22. This value lies between 0.2 and 0.4, indicating that the correlation is lower. At a significance level of 5%, the Wald test rejected the null hypothesis that all parameters were null (χ2 = 82.07, df = 6, p-value < 0.001). Then, the explanatory variables in the final model were collectively significant. This meant that at least one variable explained the model. The link test did not reject the null hypothesis that the fit was good or else that the observed rates and the predicted rates of water availability satisfaction in subgroups of subscribers were close or else that the squared additional term was equal to zero (z= − 0.61, p-value = 0.541, 95% CI = [−0.27, 0.14]). Then, the model was good, an indication that the model fitted well the data. The correctly classified subscribers' rate was 75.96%. AUC was estimated at 0.80, showing the optimal model had a strong predictive power with excellent discrimination (Figure 2). Hence, probabilities of water availability satisfaction could be made.
Figure 2
ROC and AUC.
View largeDownload slide

ROC and AUC.

Table 6 displays the six higher and the six lower predicted probabilities of being satisfied with water availability according to selected variables.

Table 6

Predicted probabilities

Time of water supplyAge classFrequency of water-related outagesProbabilities
Day and night 52–81 Rarely 0.930 
Day and night 32–40 Rarely 0.893 
Day and night 41–51 Rarely 0.891 
Day and night 52–81 Often/sometimes/everyday 0.840 
Day and night 18–31 Rarely 0.800 
Day and night 32–40 Often/sometimes/everyday 0.768 
Daytime 18–31 Rarely 0.307 
Night 18–31 Rarely 0.279 
Daytime 32–40 Often/sometimes/everyday 0.269 
Daytime 41–51 Often/sometimes/everyday 0.264 
Night 32–40 Often/sometimes/everyday 0.243 
Night 18–31 Often/sometimes/everyday 0.133 
Time of water supplyAge classFrequency of water-related outagesProbabilities
Day and night 52–81 Rarely 0.930 
Day and night 32–40 Rarely 0.893 
Day and night 41–51 Rarely 0.891 
Day and night 52–81 Often/sometimes/everyday 0.840 
Day and night 18–31 Rarely 0.800 
Day and night 32–40 Often/sometimes/everyday 0.768 
Daytime 18–31 Rarely 0.307 
Night 18–31 Rarely 0.279 
Daytime 32–40 Often/sometimes/everyday 0.269 
Daytime 41–51 Often/sometimes/everyday 0.264 
Night 32–40 Often/sometimes/everyday 0.243 
Night 18–31 Often/sometimes/everyday 0.133 
View Large

A subscriber who received water day and night in his household, who was aged 52–81 years and who rarely experienced water-related outages had 0.930 as a probability of being satisfied with water availability. However, a subscriber who received water during the day, who was aged 18–31 years and who often or sometimes or every day experienced water-related outages had 0.133 as a probability of being satisfied with water availability.

Figures 3–6 plot Cook's distance, Welsh-Kuh's distance, Studentized residuals, and leverage, respectively.
Figure 3
Cook's distance.
View largeDownload slide

Cook's distance.

Figure 3
Cook's distance.
View largeDownload slide

Cook's distance.

Close modal
Figure 4
Welsh-Kuh's distance.
View largeDownload slide

Welsh-Kuh's distance.

Figure 4
Welsh-Kuh's distance.
View largeDownload slide

Welsh-Kuh's distance.

Close modal
Figure 5
Studentized residuals.
View largeDownload slide

Studentized residuals.

Figure 5
Studentized residuals.
View largeDownload slide

Studentized residuals.

Close modal
Figure 6
Leverage.
View largeDownload slide

Leverage.

Only two observations had an influence on the overall regression model according to Cook's distance (Figure 3). Besides, the Welsh-Kuh's distance threshold was 0.30. Hence, as confirmed by Cook's distance, approximately 1% of all observations (4 out of 391) were considered as influential points based on the Welsh-Kuh's distance (Figure 4)). Fifteen (out of 391) observations had a studentized residual outside of the interval [−2,2], indicating that less than 4% (3.8%) of observations were considered as outliers (Figure 5). This is a sign that most of the studentized residuals (more than 95%) were concentrated in the interval [−2,2]. Given the relationship between Cook's distance and leverage, and according to Hoaglin and Welsh's criterion, there were only 27 observations (6.9%) that had a leverage greater than 0.04, an indication that few observations were considered as outliers with respect to explanatory variables (Figure 6).

Listen

The aim of our study was to examine factors influencing water availability satisfaction among water subscribers and to predict probabilities of being satisfied with water availability in Bujumbura Municipality, Burundi. A study conducted in Johannesburg, South Africa, underlined that to achieve household satisfaction with water supply, it is enough to make a compromise between water quality and the quality of service provided (Mahlasela et al. 2020). Besides, a study conducted in Chile on factors associated with water service quality satisfaction showed that service quality is influenced by the perception of water quality and the payment system (Denantes & Donoso 2021a). In our study, less than two-thirds (66.5%) of interviewed water subscribers were satisfied with water availability. This result corroborates that found by Budiyono and colleagues (55%) in the Coastal of Semarang City, Indonesia (Budiyono et al. 2020). This high rate of satisfaction observed among water subscribers in Bujumbura Municipality is justified by the fact that REGISEDO makes cuts so that the entire urban population has tap water. In Bujumbura Municipality, drinking water demand is very strong, especially in the dry season. Water availability varies depending on seasons (rainy season, dry season), changes in water storage and many other factors such as regulations, quantity of water available and water demands as found elsewhere (Barlow et al. 2004). As the single company which distributes drinking water in Bujumbura Municipality, REGIDESO is making more efforts to produce water even for households located at high altitudes. This study showed also that the time of water supply and frequency of water-related outages significantly influence water availability satisfaction after adjustment for age. During the day, there are several public and private utilities that use drinking water intensively such as restaurants, government departments, businesses and other organizations. That is why subscribers who receive water during the day and night are more likely to be satisfied with water availability. In addition, customers who experience water-related outages frequently, sometimes or daily are less likely to be satisfied with water availability than those who do not. Factors influencing satisfaction with drinking water availability can also be found at an individual level. In fact, a study conducted in Saskatchewan (Canada) showed that subscribers who do not boil tap water, who do not experience tap water odor and who live far away from urban settings are more likely to be satisfied with tap water (Bermedo-Carrasco et al. 2018). Besides, a study conducted in Chile showed that trust in the Water Supply Board has an influence on subscribers' satisfaction with drinking water availability (Denantes & Donoso 2021b).

Subscribers perceive or judge the quality of drinking water based on several factors such as the color of the water, its transparency, its taste, its odor, trust in the drinking water supply and risks due to chemicals which themselves influence the perception of and satisfaction with drinking water quality (Doria 2010; Villar-Navascués & Fragkou 2021). This perception, in turn, influences subscribers' satisfaction with the quality of the drinking water service. Further studies should focus on comparing satisfaction rates with drinking water availability among urban and rural subscribers.

Listen

The aim of our study was to examine factors influencing satisfaction with water availability in Bujumbura Municipality. Findings showed that the satisfaction rate with drinking water availability among water subscribers was high. Furthermore, the time of drinking water supply and frequency of water-related outages significantly influenced the satisfaction with drinking water availability after adjustment for age. Subscribers with a higher level of education were more likely to be satisfied with water availability than their counterparts who had no education level. Subscribers who drew water only in the daytime were less likely to be satisfied than those who got water both daytime and night. Subscribers living in Mukaza or Ntahangwa communes were not more likely to be satisfied with water availability than those living in Muha commune. The highest predicted probability of being satisfied with drinking water availability was observed among subscribers aged between 52 and 81 years, who received water both day and night in their households and rarely experienced water-related outages. To tackle the problem of unavailability of drinking water in some neighborhoods of Bujumbura Municipality, one solution would be to drill.

Listen

The authors are grateful to drinking water subscribers for their cooperation. This research met no financial support.

Listen

All relevant data are included in the paper or its Supplementary Information.

Listen

The authors declare there is no conflict.

Ahmad
S.
,
Singh
R.
,
Arfin
T.
 &
Neeti
K.
2022
Fluoride contamination, consequences and removal techniques in water: A review
.
Environmental Science: Advances
1
(
5
),
620
–
661
.
https://doi.org/10.1039/D1VA00039J
.
Google Scholar
 
Baggio
G.
,
Qadir
M.
 &
Smakhtin
V.
2021
Freshwater availability status across countries for human and ecosystem needs
.
Science of The Total Environment
792
,
148230
.
https://doi.org/10.1016/j.scitotenv.2021.148230
.
Google Scholar
Crossref
Search ADS
PubMed
 
Barlow
P. M.
,
Alley
W. M.
 &
Myers
D. N.
2004
Hydrologic aspects of water sustainability and their relation to a national assessment of water availability and Use
.
Water Resources Update
127
,
76
–
86
.
Google Scholar
 
Bermedo-Carrasco
S.
,
Bharadwaj
L.
 &
Waldner
C. L.
2018
Factors associated with drinking and being satisfied with tap water in Indigenous communities in Saskatchewan, Canada
.
International Journal of Circumpolar Health
77
(
1
),
1466605
.
https://doi.org/10.1080/22423982.2018.1466605
.
Google Scholar
Crossref
Search ADS
PubMed
 
Budiyono
B.
,
Pamungkas
C.
 &
Hanani Darundiati
Y.
2020
The customer satisfaction of the drinking water in the coastal of semarang and the risk management
.
E3S Web of Conferences
202
,
05015
.
https://doi.org/10.1051/e3sconf/202020205015
.
Google Scholar
Crossref
Search ADS
 
Chard
A. N.
,
Trinies
V.
,
Edmonds
C. J.
,
Sogore
A.
 &
Freeman
M. C.
2019
The impact of water consumption on hydration and cognition among schoolchildren: Methods and results from a crossover trial in rural Mali
.
PLoS ONE
14
(
1
),
e0210568
.
https://doi.org/10.1371/journal.pone.0210568
.
Google Scholar
Crossref
Search ADS
PubMed
 
Cox
D. R.
1958
The regression analysis of binary sequences
.
Journal of the Royal Statistical Society. Series B (Methodological)
20
(
2
),
215
–
242
.
Google Scholar
Crossref
Search ADS
 
Denantes
J.
 &
Donoso
G.
2021a
Factors influencing customer satisfaction with water service quality in Chile
.
Utilities Policy
73
,
101295
.
https://doi.org/10.1016/j.jup.2021.101295
.
Google Scholar
Crossref
Search ADS
 
Denantes
J.
 &
Donoso
G.
2021b
Factors influencing customer satisfaction with water service quality in Chile
.
Utilities Policy
73
,
101295
.
https://doi.org/10.1016/j.jup.2021.101295
.
Google Scholar
Crossref
Search ADS
 
Ellawala
K. C.
 &
Priyankara
D. P. M. P.
2016
Consumer satisfaction on quantity and quality of water supply: A study in Matara, Southern Sri Lanka
.
Water Practice and Technology
11
(
3
),
678
–
689
.
https://doi.org/10.2166/wpt.2016.073
.
Google Scholar
Crossref
Search ADS
 
Hajian-Tilaki
K.
2013
Receiver operating characteristic (ROC) curve analysis for medical diagnostic test evaluation
.
Caspian Journal of Internal Medicine
4
(
2
),
627
–
635
.
Google Scholar
PubMed
 
Institute of Statistics and Economic Studies of Burundi
2017
Demographic Projections 2010–2050 National and Provincial Level
.
Demographic Studies and Statistics
,
Bujumbura
,
181
p.
Kama
R.
,
Song
J.
,
Liu
Y.
,
Hamani
A. K. M.
,
Zhao
S.
 &
Li
Z.
2023
Water availability and status of wastewater treatment and agriculture reuse in China: A review
.
Agronomy
13
(
5
),
Article 5
.
https://doi.org/10.3390/agronomy13051187
.
Google Scholar
Crossref
Search ADS
 
Li
H.
,
Cohen
A.
,
Li
Z.
,
Lv
S.
,
He
Z.
,
Wang
L.
 &
Zhang
X.
2020
Intermittent water supply management, household adaptation, and drinking water quality: A comparative study in two Chinese provinces
.
Water
12
(
5
),
1361
.
https://doi.org/10.3390/w12051361
.
Google Scholar
Crossref
Search ADS
 
Mahlasela
P.
,
Oke
A.
 &
Madonsela
N. S.
2020
Household's satisfaction with water supply in Johannesburg metropolitan municipality, South Africa
.
Procedia Manufacturing
43
,
183
–
192
.
https://doi.org/10.1016/j.promfg.2020.02.133
.
Google Scholar
Crossref
Search ADS
 
Mandy
D. A.
,
Barkey
R. A.
,
Arsyad
U.
 &
Nursaputra
M.
2020
Comparison of water availability in 2015 and 2022 based on land cover in the Maros river basin
.
IOP Conference Series: Earth and Environmental Science
575
(
1
),
012133
.
https://doi.org/10.1088/1755-1315/575/1/012133
.
Google Scholar
 
Megerle
H. E.
 &
Niragira
S.
2020
The challenge of food security and the water-energy-food nexus: Burundi case study
World Review of Nutrition and Dietetics
121, 183–192. https://doi.org/10.1159/000507488
.
Google Scholar
Crossref
Search ADS
 
Mohammed
A. B.
 &
Sahabo
A. A.
2015
Water supply and distribution problems in developing countries: A case study of Jimeta-Yola, Nigeria
.
Environmental Science
1
(
4
).
In: Proceedings, Mohammed, A. B. Abdulrahman Ahmed Sahabo & Hamma Adama, M. A. 2015 Water Supply and Distribution Problems in Developing Countries: A Case Study of Jimeta-Yola, Nigeria, https://api.semanticscholar.org/CorpusID:133289861.
Google Scholar
 
Ocholla
G.
,
Letema
S.
 &
Mireri
C.
2022
Socioeconomic determinants of water delivery satisfaction in a medium sub-Saharan Africa city: A case of Kisumu, Kenya
.
Water Supply
22
(
12
),
8682
–
8697
.
https://doi.org/10.2166/ws.2022.388
.
Google Scholar
Crossref
Search ADS
 
Okello
C.
,
Tomasello
B.
,
Greggio
N.
,
Wambiji
N.
 &
Antonellini
M.
2015
Impact of population growth and climate change on the freshwater resources of Lamu Island, Kenya
.
Water
7
(
3
),
Article 3
.
https://doi.org/10.3390/w7031264
.
Google Scholar
Crossref
Search ADS
 
Salehi
M.
2022
Global water shortage and potable water safety; Today's concern and tomorrow's crisis
.
Environment International
158
,
106936
.
https://doi.org/10.1016/j.envint.2021.106936
.
Google Scholar
Crossref
Search ADS
PubMed
 
Swets
J. A.
1979
ROC analysis applied to the evaluation of medical imaging techniques
.
Investigative Radiology
14
(
2
),
109
–
121
.
https://doi.org/10.1097/00004424-197903000-00002
.
Google Scholar
Crossref
Search ADS
PubMed
 
Tessema
M. M.
2020
Assessment of customer's satisfaction on water supply service deliveryin Bahir Dar City, Ethiopia
.
International Journal of Engineering Research and Advanced Technology
06
(
10
),
20
–
29
.
https://doi.org/10.31695/IJERAT.2020.3656
.
Google Scholar
Crossref
Search ADS
 
Uakarn
C.
,
Chaokromthong
K.
 &
Sintao
N.
2021
Sample size estimation using yamane and cochran and krejcie and morgan and green formulas and Cohen statistical power analysis by G*Power and comparisons
.
Apheit International JournaL
10
(
2
),
76
–
88
.
Google Scholar
 
United Nations
2015
THE 17 GOALS | Sustainable Development
.
Available from: https://sdgs.un.org/goals.
Zheng
M.
,
Wutich
A.
,
Brewis
A.
 &
Kavouras
S.
2022
Health impacts of water and sanitation insecurity in the global north: A scoping literature review for U.S. colonias on the Mexico border
.
Journal of Water and Health
20
(
9
),
1329
–
1342
.
https://doi.org/10.2166/wh.2022.085
.
Google Scholar
Crossref
Search ADS
PubMed
 
© 2024 The Authors
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).