Modeling the complexities of water, hygiene, and health in Limpopo Province, South Africa

Abstract

Researchers have long studied the causes and prevention strategies of poor household water quality and early childhood diarrhea using intervention-control trials. Although the results of such trails can lead to useful information, they do not capture the complexity of this natural/engineered/social system. We report on the development of an agent-based model (ABM) to study such a system in Limpopo, South Africa. The study is based on four years of field data collection to accurately capture essential elements of the communities and their water contamination chain. An extensive analysis of those elements explored behaviors including water collection and treatment frequency as well as biofilm buildup in water storage containers, source water quality, and water container types. Results indicate that interventions must be optimally implemented in order to see significant reductions in early childhood diarrhea (ECD). Household boiling frequency, source water quality, water container type, and the biofilm layer contribution were deemed to have significant impacts on ECD. Furthermore, concurrently implemented highly effective interventions were shown to reduce diarrhea rates to very low levels even when other, less important practices were suboptimal. This technique can be used by a variety of stakeholders when designing interventions to reduce ECD incidences in similar settings.

Figure 1

Flow chart of ABM routines. Source water can be of three different types, municipal tap (MT), community piped (CP) and surface water (SW) with related quality and reliability. Households collect water from either a primary or secondary source with a given rate. During storage the water can be contaminated by a biofilm layer or biological regrowth which depends on container type and source water. Water can be further contaminated by water transfer devices or dirty hands for ‘wide neck’ containers. Finally, it can be treated by boiling or water quality can be improved through container cleaning. This water chain determines WQ_i - the water quality of the i^thhousehold - which in turn makes children more or less likely to get ECD. A child’s propensity to get ECD can be decreased through hand-washing frequency or increased through previous ECD cases. Children then grow according to their ECD status. They experience stunted growth during ‘single’ cases of ECD and are further stunted if they get two or more cases of ECD within 8 months.

Figure 2

Single parameter behavior space analysis of model summarizing the results from four different behavior space runs with median daily WQ_i, mean total ECD cases, and mean HAZ₂. The boiling and collection interval experiments are in terms of those behaviors occurring every X days. Municipal tap (MT) water quality is the water quality of the MT water system and is varied according to typical ranges found in field data. Biofilm layer contribution (HHS_i) is the amount of coliform bacteria associated with the water storage container sidewalls and is also varied across ranges typical of the field data. Results indicate that only optimal interventions (e.g. daily boiling and HHS_i = 0 cfu/100 mL) produced large reductions in growth stunting as reported by HAZ₂ values. Other variables such as daily water collection and low coliform levels in the MT water system improved water quality, but failed to decrease ECD rates or child stunting.