Water, sanitation and the risk of chronic diseases in old age: Results from the Ghana WHO SAGE 2

Data source

The study used data from the WHO Study on Global AGEing and adult health (SAGE). Wave 2 was a survey conducted in six LMICs, including China, Ghana, India, Mexico, Russian Federation, and South Africa.¹⁶ The nationally representative survey collects data through a stratified multistage cluster design to complement existing aging data sources and inform policy and programs. WHO and the University of Ghana Medical School through the Department of Community Health collaborated to implement SAGE Wave 2 in Ghana in 2014–2015. We used the GhanaINDDataW2 and GhanaHHDataW2 datasets. The INDD data set comprised of individual questions targeted at the main respondent and the HHD data set comprised of questions concerning the household within which the primary respondent resided.

The primary sampling units were stratified by region and location of residence (urban/rural) with samples selected from 250 enumeration areas.¹³ In households identified as “older” for sampling purposes, all household members aged 50 years and older were invited to participate in the study. Individuals were interviewed regarding their chronic health conditions and health services coverage, subjective wellbeing and quality of life, health care utilization, risk factors, and preventive health behaviors, perceived health status, socio-demographic and work history, social cohesion, and household characteristics. Respondents further provided details about the use of water and sanitation facilities, including the source of water and type of toilet facility and whether these facilities were shared with others. Primary data management, checking and quality assurance was undertaken by country survey teams and coordinated centrally through WHO Geneva. The Ghana response rate was 83%. The data are publicly available via the WHO Multi-Country Studies Data Archive. Details about the data and further information can be found at https://apps.who.int/healthinfo/systems/surveydata/index.php/catalog/sage. Weight was calculated to offset the sampling effect.¹⁷

Measures

Outcome variable

The main outcome variable of interest in this study was chronic diseases defined as the existence of one or more chronic conditions in an individual (ranging from 1 to 8). We included all available self-reported chronic conditions queried in the WHO-SAGE data, each prompted by the item “Has a health care professional ever told you that you have … hypertension, diabetes, chronic lung diseases, angina, asthma, tightness in the chest, stroke, and arthritis?”. Individuals who indicated “yes” to these items were recorded as having a chronic disease. We dichotomized the outcome variable into 0 = no chronic disease when the respondent answered “no” to all the items and 1 = presence of chronic diseases when at least one response was affirmative.

Explanatory variables

Three independent variables were considered in this analysis: the source of water, type of toilet facility, and shared toilet facility. First, older adults were asked to indicate the main sources of drinking water. The responses were broadly categorized into 1 = piped private, 2 = piped to yard/plot, 3 = public tap/standpipe, 4 = tube well/borehole, 5 = protected dug well, 6 = unprotected dug well, 7 = protected spring, 8 = unprotected spring, 9 = rainwater collection, 10 = bottled water, 11 = small scale vendor, 12 = tanker-truck/lorry, 13 = surface water (river, lake, etc.).

Based on the Joint Monitoring Program’s classification of water and sanitation technologies,¹ we recoded these sources of water into “improved source” = 1 to include piped private, piped to yard/plot, public tap/standpipe, tube well/borehole, protected dug well, protected spring, bottled water, small scale vendor, tanker-truck/lorry, and “unimproved source” = 2 to include unprotected dug well, unprotected spring, and surface water.

Type of toilet facility was collected on a 12-response scale including 1 = flush/pour to piped sewage system, 2 = flush/pour to septic tank, 3 = flush/pour to pit latrine, 4 = flush/pour to other locations, 5 = flush/pour to unknown, 6 = ventilation improved pit latrine, 7 = pit with slab, 8 = pit without slab/open, 9 = composting toilet, 10 = bucket, 11 = hanging toilet/latrine, 12 = no facilities (bush, field). These responses were transformed into 1 = “improved toilet facility” (flush/pour to the piped sewage system, flush/pour to a septic tank, flush/pour to pit latrine, flush/pour to other locations, flush/pour to unknown, ventilation improved pit latrine, a pit with slab, composting toilet), and 2 = “unimproved toilet facility” (pit without slab/open, bucket, hanging toilet/latrine, no facilities (bush, field). Finally, participants answered on a “no” = 1 or “yes” = 2 scale about whether a toilet facility was shared with others.

Covariates

Sociodemographic and health-related variables were assessed and included for adjustments and as moderators. These included age (years), sex (1 = male, 2 = female), location of residence (1 = urban, 2 = rural), and years of education. Marital status was collected using a four-level measure but collapsed and dichotomized into currently married/partnered = 1, and not currently married/partnered = 2. WHO-SAGE collected data on ethnic background on a 10-level scale but the variable was dichotomized into 1 = Akan, 2 = others due to the limited frequencies for specific categories and subsequent incidence of model over-fitting in the regression analysis. Respondents were asked to rate their own health, on a scale from 1 to 5 (very good, good, moderate, bad, very bad) with the item, “How would you rate your current health state?” A higher score indicated poor self-rated health and we recoded this variable into good (very good, good) = 1, moderate = 2, bad (bad/very bad) = 3 for analytic purposes.

Statistical analysis

Univariate descriptive analysis was first conducted to generally describe the characteristics of the sample. These statistics were reported as mean and standard deviation for continuous variables, or count and percentage for categorical variables. Next, we performed bivariate analysis stratified by sex and location of residence to estimate the relationships between the study variables using non-parametric Pearson’s χ² test for categorical variables and independent t-test for continuous variables. Kendall’s tau-b correlations were run to determine the relationships of relevant exposure variables with the outcome variable. Accounting for the complex survey design, survey weights were used to estimate sex- and residential-specific prevalence of chronic conditions, and water and sanitation indicators.

Given the measurement level of the outcome variable, a series of generalized logistic regression models were conducted in which the presence of chronic diseases was regressed on the major independent variables (water, and sanitation) controlling for the potential confounders. Models 1, 2, and 3 regressed the presence of chronic diseases on the use of water, toilet facility type, and sharing of toilet facility respectively. These crude models estimated the variance explained by the three key independent variables. In addition, the presence of chronic diseases was regressed on the three key independent variables simultaneously in Model 4. Model 5 added the sociodemographic and health-related variables for full adjustment. In Model 6, we included the interaction terms (water/sanitation indicators × sex and water/sanitation indicators × residential status) to investigate the potential modifying roles of sex and location of residence in the associations. In a confirmatory analysis, we fitted separate models to estimate the specific effect of independent variables on each individual chronic disease. We checked for multicollinearity by computing the variance inflation factor (VIF) but none of the VIF scores exceeded the value of 1.5, indicating no challenges of multicollinearity. A p-value of less than 0.05 was considered statistically significant and all analyses were performed using IBM SPSS Statistics v.21.0 (IBM, Armonk, NY) (RRID:SCR_019096) (An open-access alternative that can perform an equivalent function is the R stats package (R Project for Statistical Computing, RRID:SCR_001905)).

Ethical approval

SAGE was approved by the World Health Organization's Ethical Review Board (Reference number RPC149) and the Ethical and Protocol Review Committee, College of Health Sciences, University of Ghana, Accra, Ghana. Written informed consent was obtained from all study participants.

Sample characteristics

The characteristics of the sample are presented in Table 1.¹⁸ Our study population consisted of 4735 adults aged 50 years or over. The average age was approximately 58 years and the majority of the respondents lived in rural areas (59%), were females (59%), and were married or partnered (57%). More than half of the respondents professed to other ethnic groups than the Akan. The mean years of education were 11 years, and the health status of the sample was generally good with 68% reporting “good” and only 8% rating their health as “bad”. Approximately 19% reported at least one chronic condition. Hypertension was the most prevalent individual chronic condition (12%), followed by arthritis (8%), tightness in the chest (4.3%), and diabetes (3%). About 90% and 77% reported using improved water and toilet facility sources respectively, and 77% of the respondents shared their toilets.

Bivariate associations

Table 1 shows the bivariate associations stratified by sex and location of residence. Compared to females and urban dwellers, males and those in rural areas respectively were more likely to be older (p < 0.05).¹⁸ Similarly, males and urban residents were more likely to report higher levels of education than their respective females and rural dwellers (p < 0.001). More males were married compared with females (74% vs 45%, p < 0.001) but the majority of the married lived in rural areas compared to urban areas (61% vs 52%; p < 0.001). Males self-rated their health better compared to females and also in terms of chronic conditions (16% vs 21%; p < 0.001). Whilst the difference in health status was statistically insignificant between urban and rural residents, the former reported more co-morbidity (23% vs 16%, p < 0.001). The tendencies to use improved sources of water and toilet facilities were significantly higher for females and urban dwellers compared to male and rural dwellers respectively. Kendall’s tau-b correlations are depicted in Table 2.¹⁸ Co-morbidity was significantly associated with water source (τ_b = -.033, p < .005), type of toilet facility (τ_b = -.080, p < .001), shared toilet facility (τ_b = -.044, p < .001), age (τ_b = .199, p < .001), sex (τ_b = .054, p < .001), residence (τ_b = -.083, p < .001), years of education (τ_b = .046, p < .001), marital status (τ_b = -.117, p < .001), and self-rated health (τ_b = .201, p < .001).

Main regression results

Table 3 depicts the results of the associations of water and sanitation with co-morbidity estimated by generalized logistic models of the pooled sample.¹⁸ In the unadjusted models (Models 1, 2, and 3), those who used unimproved water (OR = 1.350, CI: 1.037–1.757), toilet facility (OR = 1.716, CI: 1.413–2.083), and shared toilet facility (OR = 1.287, 1.077–1.539) were significantly more likely to report co-morbidity. Accounting for water and sanitation variables in Model 4, the associations in Model 3 persisted but the effect sizes were negligibly attenuated by 0.12, 0.10, and 0.02 respectively. After adjusting for potential confounders (Model 5), the odds of reporting chronic conditions significantly increased for those using unimproved toilet facilities (OR = 1.732, CI: 1.377–5.418), and shared toilet facilities (OR = 1.624, 1.095–1.320). However, the use of unimproved water sources lost its robustness although it had a higher odds of chronic disease risk than the use of improved water sources (OR = 1.552, 95% CI: 0.792–3.042) suggesting the potential role of sociodemographic and health-related factors in predicting co-morbidity risks in later life.

Moreover, we found significant interactions between toilet facilities and sex (Model 6). Compared to males, females who used unimproved toilet facilities were 3.5 times more likely to report chronic conditions (OR = 3.498, CI: 1.744–16.442). Also, the results revealed a moderation effect of residential status on the association between water and sanitation, and co-morbidity: First, using unimproved water sources in rural areas significantly increased the risk of chronic diseases compared to living in urban areas (OR = 5.935, CI: 1.320–26.685). Second, rural residents who used unimproved toilet facilities were twice as likely to have a chronic disease compared with their urban counterparts (OR = 1.998, CI: 1.462–8.642).¹⁸

Additional models

A sensitivity CC-specific estimation showed mixed effects of water and sanitation on CCs (Table 4).¹⁸ The use of unimproved water and unimproved toilet facilities was associated with an increased likelihood of reporting hypertension, diabetes, and chronic angina diseases. Moreover, the shared toilet facilities significantly increased the odds of hypertension and arthritis among older people.

Underlying data

Zenodo: Underlying data for ‘Water, sanitation and the risk of chronic conditions in old age: Results from the Ghana WHO SAGE 2’. https://doi.org/10.5281/zenodo.5829346

This project contains the following underlying data:

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).

The WHO SAGE Wave 2 dataset will shortly be freely available online. Further details can be found at: https://apps.who.int/healthinfo/systems/surveydata/index.php/catalog/sage

Researchers interested in using the data used in this study in the meantime should contact the corresponding author, Dr Razak Gyasi (rgyasi.research@gmail.com), for further information.