05. LAs TEORíAs LUEGO DE LOs ‘70
5.5. Teoría regulacionista
Study design, participants and sam pling procedure
This was a cross-sectional study where respondents were household heads or their spouses or any other household member who played a significant role in household
decision-making, as well as supplying the bulk o f the money and other household necessities. Using the formula for comparative studies by Altman (1991) based on a power o f 80%, Po o f 0.6 and 5% level o f significance, a sample o f 456 respondents was got. However, the data collection exercise yielded 450 respondents with 222 from affected households and 228 from non-affected households.
Two different sampling procedures were used. For non-affected households, a multi-stage cluster sampling technique was used with a sub-county as the primary sampling unit. A list o f all sub-counties and town councils was drawn and two sub counties were then randomly selected from each o f the districts while the 2 town councils (Mpigi and Kanoni) were purposively selected. Twelve parishes or wards were randomly selected from selected sub-counties and town councils (two parishes per sub-county) and finally twenty four villages/zones were considered (two villages or zones per parish or ward). In each village, a central location was identified, and using a randomly selected interval o f 6, households were selected. After throwing a pen in the air, the direction o f the tip was taken as the starting direction. Prominent features such as a school, church, mosque, market, social centre etc where used as the starting point and the first household after the prominent feature was selected and there after every 6* household was selected for interviews. To decrease chances o f misclassification, households with members receiving tuberculosis or pneumonia treatment for at least one month were excluded.
For affected households, PLWHA residing in study districts were recruited from HIV/AIDS treatment health centres between July 2011 and January 2012. The research team visited the health centres on the days when HIV/AIDS treatment was done. Permission to meet all adult (above 18 years) PLWHA available was sought from the
health facility managers. The nature o f the study was explained to all PLWHA available and only those (PLWHA) that provided written informed consent were interviewed and where necessary followed up in their households. Those that were followed up provided household drinking water samples for microbiological analysis as part o f sub-study IV.
Data collection, m anagem ent and analysis
Data for access to water sources was collected based on the reported distances o f water sources from respondents’ households as well as the reported water collection journey time as was recommended by WHO/UNICEF (2004) and Cairncross and Valdmanis (2004). Access to water was further disintegrated in the different levels o f service (public or community sources and private/house connections). Access to sanitation was assessed in terms o f presence and type o f sanitation facility as well as presence and type o f hand washing facility (see annex 1 for the Household survey Instrument).
The WASH perceptions data was collected along two lines. Sanitation perceptions data was based on some o f the major themes that were suggested by Jenkins (1999) as well as Jenkins and Scott (2007). These included cleanliness; privacy; safety; good health; comfort and convenience and assessment o f the most valued attributes was done. Water perceptions data was based on the studies o f Thompson et al. (2011) as well as Cairncross and Kinnear (1992). These included perceptions on energy and time requirements for collecting water as well as perceptions on water costs and vendor
supply. Buying w ater from w ater vendors is a growing practice in m any developing countries and the determ inants o f the practice have been recom m ended for research A fter collection, data were entered in Epidata version 3.0 software and exported to SPSS version 17 and STATA version 10 for analysis. Data were analysed using both descriptive and inferential statistics. For the descriptive statistics, frequencies and cross tabulations were generated and odds ratios as well as chi-squares were used to derive associations.
For the inferential statistics, bivariate and m ultivariate analyses were computed. B ivariate analysis was done to identify significant variables for consideration in m ultivariable analysis. M ultivariable analysis using logistic regression was used to explore the relationship betw een w ater related perceptions and being custom ers o f w ater vendors. Odds rations (OR) and adjusted odds ratios (AOR) derived from the fmal m ultivariable m odel were generated. All significant factors at bivariate analysis as well as factors theoretically associated with buying w ater from vendors were included in the m ultivariable analysis. The dependent variable was the probability that a respondent would buy w ater from w ater vendors. The independent variables included perceptions such as: fetching w ater takes a lot o f time; fetching w ater takes a lot o f energy; prim ary w ater source is far from home; regular washing o f hands w ith w ater and soap is good practice and existing w ater source is improved.