Menu
Menu
Menu
Menu
To increase maternal health service use and household water treatment (HWT), free water treatment kits were provided at first antenatal care (ANC) visits and free water treatment sachet refills were provided at follow-up ANC visits, delivery, and postnatal visits in 46 health facilities in rural Uganda. We evaluated the impact by surveying 226 women in the initiative (intervention group) and 207 women who received ANC before the initiative began (comparison group). There was no differences in the percentages of intervention and comparison group women with ≥ 4 ANC visits; however, a higher percentage of intervention group women reported treating their drinking water (31.7% versus 19.7%, P = 0.01), and had free chlorine residual in stored water (13.5% versus 3.4%, P = 0.02) than comparison group women. The intervention did not appear to motivate increased maternal health service use, but demonstrated improvements in HWT.
The evaluation included three components: 1) a cross-sectional household survey of an intervention group of women exposed to the Water for Health intervention in 2013 and a comparison group who received ANC in 2012, 2) abstraction of data from 2012 and 2013 ANC and maternity registries, and 3) a health-care provider survey. The 46 HF where Water for Health was implemented ranged from level II with catchment areas of 5,000 people to district hospitals with catchment areas of 500,000. All women who received ANC in these HF in 2012 and all women who received ANC and the water treatment kit in the same HFs in 2013 were eligible to participate. We based our sample size calculation on sachet use, a principal objective of the implementing organization and the donor. To determine the cross-sectional survey sample size, we assumed that the comparison group would have a 5% usage of sachets, which were available in shops in Uganda before the study began, and that the intervention group would report an increase in sachet use to 15%. To assure an adequate sample size, the estimated baseline use of sachets was higher than typically seen16 and the expected increase in use of free product was modest. A sample of 150 women in each group would provide 90% power to detect the difference in sachet use at an α = 0.05. We chose to include all HFs in the evaluation for several reasons. First, increasing the number of HFs enrolled would reduce the impact of a single HF on results and provide analytical options should some facilities provide influential or distinct results. Second, we minimized the potential impact of the design effect on the variance estimates by maximizing the number of HFs enrolled and selecting a smaller sample of women from each facility. Finally, we aimed for as many HFs as available so that HF variance could be statistically addressed, if needed. To ensure that women from each HF were adequately represented in the sample, a minimum of six women per HF was selected, which resulted in a substantial increase in sample size. We stratified our sample across the 37 eligible HFs and aimed to enroll six intervention and six comparison group women per HF, with one exception of a large hospital in which we targeted 10 intervention and 10 comparison group women for enrollment. Eligible women were selected for interview by random number generation from lists of women whose first ANC visit was recorded in ANC registries in March–April of 2012 for the comparison group and March–April of 2013 for the intervention group. This time frame was chosen to ensure that intervention group women had ample time for four ANC visits and would have delivered before the November 2013 evaluation. This was the primary sampling procedure, but because not all women identified by the registry could be located in a village (a common problem in rural Uganda) we used an alternate enrollment procedure in an attempt to reach the desired sample size. In this case, village health teams generated a list of women who attended at least one ANC visit, and delivered between 1 and 2 years previously (for the comparison group) or delivered less than 3 months previously (for the intervention group). Women were randomly selected, using a random number generator, from the list developed by the village health teams. For both primary and alternate enrollment selection methods, women selected for the comparison group were excluded if they were pregnant and received ANC in 2013. Trained research assistants fluent in two or more local languages used a standardized questionnaire to interview women about demographic and socioeconomic characteristics, pregnancy history, use of maternal health services, HWT practices, knowledge about sachets, and receipt and use of water treatment kits and sachet refills. Observations of data recorded in the maternal passport (a document maintained at home by the women, and updated by health-care providers at each visit to a maternal health clinic) and of home environmental characteristics were also made. Stored drinking water was tested for free chlorine residual using the N,N-diethyl-p-phenylenediamine colorimetric method (LaMotte Co., Chestertown, MD) as an objective measure of HWT. Data abstracted from ANC and maternity registries included monthly totals for first ANC, ≥ 4 ANC, total ANC visits, HF deliveries, PN visits for 2012 and 2013, and, for 2013 only, the number of water treatment kits and sachet refills distributed. The outcomes of ≥ 4 ANC, HF deliveries, and PN visits were divided by first ANC, a proxy for the total number of women since > 95% of women attend at least one ANC visit, to further compare across years. To improve our understanding of the Water for Health implementation process, we interviewed one health-care provider from each HF in March 2014. Data from the cross-sectional survey were entered into a Microsoft Access 2010 (Microsoft Corp., Redmond, WA) database, whereas registry data and health provider survey data were managed in Microsoft Excel (Redmond, WA); analysis was done using SAS version 9.3 (SAS Institute, Cary, NC). We tested the hypothesis that exposure to an offer of free water treatment kits and sachet refills at HF-based maternal health services would increase the percentages of mothers with ≥ 4 ANC visits, HF deliveries, PN visits, reported HWT, and confirmed HWT (positive chlorine residuals in stored household drinking water). Intervention and comparison group data were compared using Rao-Scott χ2 test, adjusting for any potential correlation within HF respondents. The design effect ranged from 0.01 to 1.62 with all but one being < 1.00. We report when the design effect exceeded 1.00 indicating some correlation. We report design-adjusted χ2 results to describe associations, as the intent and study were not powered for describing effect sizes and performing multivariable modeling. The protocol was approved by Institutional Review Boards at the Joint Clinical Research Center in Uganda and CDC (protocol 6482). Informed consent was obtained from all mothers.
N/A
