Community involvement in obstetric emergency management in rural areas: a case of Rukungiri district, Western Uganda

The study was conducted in Rukungiri District, western Uganda. Rukungiri district had a total population of 300,800 inhabitants; over 90% live in rural areas and engaged in subsistence agriculture [33]. The district was composed of two counties namely Rubabo (peri-urban) and Rujumbura (urban), 11 sub-counties, 77 parishes and 825 villages. Rukungiri District had two hospitals, two level IV health centres and 16 level III health centres and 42 level II health centres [36]. In 2008, the district had a total of 12 doctors, 117 nurses/midwives and 137 health staff of other categories providing health services. The district was selected based on its rural location far away (about 400 km) from Kampala capital city of Uganda, population distribution mainly in rural areas and relies on subsistence agriculture, poor maternal health status as indicated by a high maternal mortality ratio, low base of skilled health worker availability and poor nature of road network which is typical of most disadvantaged Ugandan districts.

This was a descriptive study. Participants were heads of households who had lived in the district for at least two years and had children born within 24 months prior to the study. Both Rubabo and Rujumbura Counties were included in the study. The sample size was estimated using the modified Kish Leslie (1965) sampling formula with a 95% confidence interval, a precision of 5%, an estimated proportion of heads of households who make decisions to seek care for obstetric complications of 50% and a non-response rate of 10%. The sample size was calculated to be 428 heads of households but 448 were interviewed. Multistage sampling method was used to select study participants. We used simple random sampling to select six sub-counties namely Nyarushanje, Kebisoni, Rukungiri Town Council, Kagunga, Nyakagyeme and Bugangari. There were 42 parishes in the six sub-counties. From the 42 parishes, 32 villages (15.1%) were randomly selected. From each village 14 households were randomly selected, hence a total of 448 households were selected. The head of household from each selected household was interviewed.

To select the household in each village the interviewer moved to its centre. To determine the initial direction of movement a bottle was spun. The starting household was selected randomly from the number of the households in the pre-determined direction. The heads of households in the selected households were interviewed. From the first household, the subsequent nearest household was selected. However, if in any of the selected household there was no eligible respondent, the interviewer moved to the next immediate household and interviewed the head of the household, and then continued till the required number of respondents in that village (cluster) was obtained. We interviewed a total of 108 (24.1%) women and 340 (75.9%) men which corresponds to the district's sex distribution of heads of household [3].

Data were collected using standardized semi-structured questionnaires. The questions covered several aspects including individual characteristics, community and health services variables. Individual characteristics included sex, age, highest level of education, religion, current marital status, current employment status, tribe, number of wives/co-wives, number of pregnancies had, ever heard of a birth plan, ever/partner used a birth plan, types of messages a birth plan contained, payment for health care and saving funds to use in case of life threatening conditions during pregnancy, delivery or within 42 days after termination of pregnancy. Knowledge about symptoms of obstetric emergencies, actions to take in case obstetric emergency occurs, consequences of life threatening conditions and what communities do when obstetric emergencies develop among women in their communities were also included. Types of community involvement available/undertaken and factors influencing community involvement in the management of obstetric emergencies including health service and community factors and individual characteristics also constituted the questions.

The interviews were conducted in the local language Rukiga. Pre-tested semi-structured questionnaires were administered by locally-trained interviewers who were supervised by the first author. Six locally-trained interviewers (all health workers) were employed for one month. The interviewers' training was conducted for over a five days period prior to the data collection. It covered mainly objectives of the study, sampling, procedures of data collection including dos and don'ts of interviewing and getting informed consent

The data were entered into EPIDATA version 3.1 and analysed using STATA version 10.0. First we carried out factor analysis to compute for the community involvement index. Prior to factor analysis the strength of inter-variable relationship including scale of reliability coefficient was assessed using Cronbach's alpha statistic test. The variable obtained after predict was changed into 3 quantiles i.e. high, medium and low.

The variables were further categorized into binary outcome (dependent variable), one value for high and the other for low community involvement. Associations between dependent variable and independent variables were tested. Independent variables were categorized into individual, community and health service factors. Individual characteristics included respondent's sex, age, employment status, whether saved money to use in the event of an obstetric emergency, perceived rating of general health facility care, rating of maternity health care, decision maker at household for health care seeking, number of wives and use of birth plans and knowledge about symptoms/signs of obstetric emergencies. Symptoms/signs of obstetric emergencies included vaginal bleeding, fits during pregnancies, high temperature within 42 days after termination of pregnancy, labour lasting more than 8 hours and swelling of the feet during pregnancy. In this study setting labour lasting more than 8 hours was considered start of prolonged labour indicating a stage when a medical doctor is expected to take action to accelerate or deliver the mother to avert adverse outcomes. Swelling of the feet though can be a normal haemodynamic physiological change, it was included to highlight the fact that not all swelling of the feet can be indicative of normal body changes but could be of progressive high blood pressure or actual severe pre-eclampsia if consistent with high blood pressure and/or protein found in urine and thus the need for women to seek health care provider's attention where the distinction is made to avoid missing a pathological state. Community factors included sensitization of the community members by their local leaders, distance to nearest health facility, distance to the nearest maternity unit (level III and IV health centres or at hospital), conditions of the roads leading to health facilities, beliefs, community actions that members undertake in the event of obstetric emergencies and availability of community social support programmes. Distance to health facility was considered a plausibly important variable. Health service factors included health education of communities about obstetric emergencies by health workers, availability of communication facility, availability of ambulance for transportation of women, perceived availability of skilled health workers at health units. Odds ratio tests were used to determine the associations between the dependent and independent variables. Multi-variable analysis was then applied to variables that were significant at P-value < 0.05 after bivariate analysis plus those with p < 0.2 but are known to be plausibly important or confounders even though they were not significant at bivariate analysis. Logistic regression was performed in order to identify independent predictors for community involvement in obstetric emergency management.

Factor analysis involving 13 variables was carried out to generate the community involvement score which was later converted into a binary variable. Four factor loadings were retained as they contributed 56.1% of the total eigen values. The community involvement index was categorized into tertiles for ease of analysis using cross-tabulations.

Analysis was carried out to identify variables that affect the level of community involvement in obstetric emergency management. Variables that were important in the model for community involvement in obstetric emergencies' management and had a p value ≤ 0.2 were included. They included saving money for use, rating of quality of general health care, rating of quality of maternal health care, sensitization of communities on signs of obstetric complications by local leaders, belief about lack of money and conducting health education by health workers. The plausibly important variables and confounders were also included: distance to nearest health facility, distance to maternity unit, sex and age even though they were not statistically significant at bivariate analysis.

Institutional consent for the study was obtained from the Makerere University School of Public Health Higher Degrees Research and Ethics Committee, Uganda National Council of Science and Technology and permission from the District Health Officer Rukungiri District. Informed verbal consent was obtained from all respondents. Research Assistants explained the purpose of the study to the respondents. Confidentiality was maintained throughout the study and thereafter. No name was indicated in the data collection tools instead unique identification numbers were used.