Study design
We conducted a cross-sectional household survey as part of the mixed methods study in three municipal wards implementing community outreach teams out of 74 wards in the iLembe District, KwaZulu-Natal, South Africa. The exploratory-descriptive qualitative study conducted in 2015, and published elsewhere1,20, was used as formative research to inform the design of questionnaire for the quantitative component. The qualitative study explored user perception of services provided by the outreach teams through 16 key informant interviews and four focus group discussions. Four themes emerged in data analysis, suggesting that community members perceived outreach teams to bring services closer, help organise services, expand the package of services and help form bridges between different sectors of health and wellbeing. The survey was limited to households that were being visited by the outreach teams at the time of the study.
Study setting
iLembe is the smallest of the 11 districts of the KwaZulu-Natal province in South Africa, covering 3 269 km2 area and with a population of 662 413.21, 22 The district is further subdivided into four sub-districts, and 74 municipal wards. The largest part of the district is rural, where service delivery is hindered by geographical constraints and poor socioeconomic status, characterised by low educational levels, high employment rates and subsistence farming under tribal authority.22 There is a small urban centre in the district, located in one of the sub-district called Kwa-Dukuza, and surrounded by semi-urban locations. The nearby sub-district of Mandeni is largely semirural, with surrounding rural parts, whereas the last two sub-districts, Ndwedwe and Maphumulo are mostly rural inland, with deep rural surrounds. At the time of the study, three of the sub-districts had one outreach team each, with the exception of Ndwedwe. As a result, the locales for the outreach team were semi-urban in Kwa-Dukuza, semi-rural in Mandeni and rural in Maphumulo. The iLembe District is served by one provincial hospital, three district hospitals, two community health centres (CHCs), 31 fixed clinics and three gateway clinics located adjacent each of the district hospitals.22 For the purposes of this study, urban areas were defined as those areas that were developed in terms of infrastructure and very densely populated. Semi-urban areas were defined as those areas that were closer to the urban areas, which are also mainly densely populated and have basic infrastructure, with most roads tarred, piped water and flush sanitation. Semi-rural, are those areas where more than 50% of the population live within a five-kilometre radius of tarred roads and piped water, but with limited choice of services. Deep rural, are those areas where more than 50% of the population live more than five kilometres from a tarred road, more than 25% of people use water from streams, dams, rivers or rainwater, and people have a limited choice of services.23
Study sampling
Although South Africa introduced outreach teams in 2011, the implementation had been slow and varied across the country, and the district of ILembe reported their first outreach team launch in 2013. However, only three outreach teams per municipal ward were operating in the three sub-districts, each linked to and supported by a PHC clinic, when the study was conducted in 2016. As a result, only three wards were sampled, although the districts boasts 74 wards. Using a population of 6 699 households in the three wards, a systematic random sampling was used to survey 383 households in the three wards that had outreach teams. Sample size was calculated using the following formula:24
n = Z2(1–α/2)pq/d2
where:
Z(1–α/2) = 1.96 at 95% confidence
p = 0.5 proportion of households who have been visited by a WBOT
q = 1–p
q = 0.5, with a desired precision of (d) ±5%.
Table 1 shows the proportion of households in each ward. In each ward, we identified the first landmark as the starting point, such as schools, clinics, shops, churches etc., and every eighteenth household starting from the left hand side of the landmark was surveyed. The inclusion criterion was those households that have received services from the outreach teams in the past 12 months. The household head or representative was interviewed, and if participation was declined or criterion not met, the next household was selected until the desired sample size was reached.
Data collection
The household survey was done by the researcher and trained fieldworkers between October and December 2016 inclusive, using predesigned and pretested questionnaires. The questionnaire used Likert scales, where questions were balanced with a scale of four and five options in order to allow respondents to choose a negative or positive response, and report neutrality, respectively. The questionnaire included sections on the respondent and household demographic information, provision of services, household satisfaction and perception of services, household experience of outreach team visits. The questionnaire piloted with potential participants prior to the study, and feedback was used to refine the questionnaire. For the purposes of this paper, information on demographic characteristics, effectiveness of the team, about the person that visits the most, perception and overall experience were extracted for further analysis.
Data analysis and modelling
Questionnaires were checked for completeness and quality by the research assistant first and then the researcher, before capturing into an MS Access database and imported into STATA IC 13.1 for analysis. Data analysis was handled as per complex survey design, using the ‘svy’ command in Stata to allow for specification of primary sampling unit as households, strata as municipal wards, and Stata-generated probability weights. We calculated the mean cut-off point of 70% for the nine indicators that we used to measure household experiences. We used frequency distributions in the strongly agreed category to establish the mean point, which was 69.4% rounded off upwards to 70%. Indicators scored 70% and above were regarded as optimal experiences, and those that scored below 70% as sub-optimal experiences. Sub-optimal experiences were treated as areas with greater opportunities for service strengthening and improvement, and therefore further characterised in the analysis and modelling. Sub-optimal household experiences of outreach teams’ visits and services were re-categorised into binary variables, and further analysed to determine covariates. Univariate and multivariate analyses were conducted for satisfaction with care, perceived professionalism, equipment and medication, and confidentiality as outcome variables. The independent variables on demographic characteristics, effectiveness of the team, the person that visits the most frequently, perception and overall experience were extracted for further analysis. The weighted odds ratio was used as the measure of association, with a corresponding 95% confidence interval (95% CI). All explanatory variables, significant at the 5% level, were included in the multivariate logistic regression model. Those variables that could not be regressed against the outcome variable, were excluded. The results were presented as odds ratios (ORs) with P-values and 95% CIs.