Socio-demographic determinants and prevalence of Tuberculosis knowledge in three slum populations of Uganda

Three urban centres in Uganda at different levels of development namely Wobulenzi Town Council, Lira Municipal Council, and Makindye Division in Kampala City. Their 2010 projected populations were 23,100; 108,600 and 403,700 respectively. These three areas report a high annual burden of TB and have above–national HIV prevalence estimates [22, 24, 25].

A cross-sectional survey was conducted between April and October 2011 as part of larger study assessing socio-economic vulnerabilities to TB in urban slums. A list of 58 villages was drawn based on the following criteria: settlements that were informal, unplanned, with limited social services, of high population density and low income. From this list, 35 villages (Wobulenzi = 10; Lira = 10; Kampala = 15) to be visited were randomly selected using open source software (OpenEpi version 2.3.3) [26]. The Field Officer (FO) liaised with Local Council chairmen to identify the “centre” and boundaries of the village. Once the study team assembled at the “centre”, selection of the first respondent was done by spinning a bottle and following the direction where it pointed. Every third or fifth house was visited depending on the density of settlements to cover all four directions of the village. Respondents, one per house, who were 18 years or older and gave informed consent were eligible. Health workers were excluded.

A survey questionnaire was designed to capture data on respondent demographics and TB knowledge with adaptations from the WHO guide on conducting surveys for advocacy-communication-and-social-mobilization (ACSM) for TB control [27]. The tool was translated into local languages (Luganda and Lango) and piloted by a mix of university students and members of the village-health-teams, fluent in both English and the local languages. This team received training in basic interview techniques prior to the study. They were supervised by experienced FOs to assure responsible conduct and data quality. Data was entered into Epi Info^TM version 3.5.3 (Centres for Disease Control and Prevention, Atlanta, USA).

The main outcome measure was the level of TB knowledge stratified as knowledge of cause, symptoms, transmission, risk groups, prevention, and treatment. Exposure variables included age, sex, marital status, educational level, employment, HIV-testing history, study site, distance from health unit and ownership of a communication device.

A sample size was independently estimated for each study area taking into account a maximum variability of 0.5 in the frequency of TB knowledge, ±0.05 precision, and 95% confidence interval. This yielded 430 participants per study site and 1290 in total after adjusting for a 0.1 proportion of persons not able to be contacted.

Data was analyzed using Intercooled Stata^TM version 11.2 (StataCorp, College Station, Texas, USA). Demographic and outcome data was summarized into frequencies, percentages and measures of central tendency. Responses to the 10 questions about TB were scored (one point for correct and zero for incorrect) and categorized into 0 – 25, 26 – 50, 51 – 75 and 76 – 100 percent corresponding to poor, low, moderate and high TB knowledge respectively as a composite outcome. This distinguishes those with almost no knowledge at all (poor) from those with some knowledge about TB (low). The association between exposure variables and TB knowledge was explored by univariate analysis, with crude odds ratios (OR) and 95% confidence intervals. Tests were two-sided and considered significant if P < 0.05. Potential determinants of TB knowledge with P < 0.25 at univariate level were included in a multivariable ordinal logistic regression model to estimate their adjusted odds ratios (aOR). Multiple imputations by chained equations was employed for sensitivity analysis of the missing data using the Stata command “ice” and “mim”, assuming a Missing At Random mechanism. All missing baseline variables and the outcome, TB knowledge, were included to create five imputed data sets.

This study was approved by the Mulago National Referral and Teaching Hospital Research Ethics Committee (MREC 501) and Uganda National Council for Science and Technology (SS 2571).

We interviewed 1361 eligible participants; 449 (33%) in Wobulenzi, 470 (34.5%) in Lira and 442 (32.5%) in Makindye. Their median age was 30 years (IQR, 25 – 40). The majority of respondents were female (72.9%); and 65.6% were married or cohabiting. A half (48.9%) had attained secondary school, whilst 11% had never attended formal education. Most were self employed (63.7%) with few having paid employment (18.5%), whilst 78.5% said they had been tested for HIV before. Seventy percent lived within 1 km of a health facility; 70% owned a Radio, 30.7% a TV and 37.4% a mobile phone (Table 1). The proportion of missing data was less than 5% and there were no statistically significant differences in baseline characteristics with the complete cases.

Many participants (1139, 83.7%) were aware of TB, initially from some form of contact with a TB patient (51.9%) than from the media (29.3%). Current TB information was commonly received through radio (50%) followed by the formal health system (25.5%). Those who said they had never heard of TB (222, 16.3%) were excluded from further analysis about TB knowledge.

Few respondents identified a germ as the cause of TB (26.7%); with smoking mentioned most commonly (37.1%), whilst a third (27.7%) did not know the cause of TB (Table 2). Coughing was the most frequently mentioned symptom reported as any cough (46.4%) or cough ≥ 2 weeks (46.8%). Half the respondents (54.3%) said TB transmission is airborne. The most frequently mentioned means of TB prevention was avoiding sharing dishes (42.9%), followed by Cough hygiene (34%), which was considered the correct response. Two thirds (68.7%) said that although anybody can get TB, persons with HIV/AIDS (19.5%) were at increased risk followed by alcoholics (12.7%), drug users (10.2%) and the poor (5%).

Eighty five percent said TB is curable, using specific drugs (69.4%). Although 80% knew TB treatment can be obtained from government facilities, only a third (35%) were aware of free-of-charge TB treatment. In fact, 32.2% said TB treatment is expensive and a quarter (25.2%) did not know whether TB treatment was provided free or at a fee. Remarkably, one third mentioned the correct 6 – 8 months duration of TB treatment. Correct responses for TB knowledge are marked (ϕ) and in bold (Table 2).

Age, gender, education level, employment status, HIV testing history, and distance from health unit; study site and ownership of a radio or a mobile phone were significantly associated with TB knowledge at univariate analysis (Table 3). In the multivariable ordinal logistic regression model (category, aOR, 95% CI, P – value) older age (40 – 59 years; 1.73; 1.30 – 2.29, P < 0.001) and residence (Lira Town; 2.02; 1.50 – 2.72, P < 0.001) were associated with higher TB knowledge scores. Whilst, lack of formal education (0.56; 0.38 – 0.83, P = 0.004), unemployment (0.67; 0.49 – 0.90, P = 0.010), never testing for HIV (0.69; 0.51 – 0.92, P < 0.012) remained statistically significant determinants of poor TB knowledge scores (Table 4). In this final model, the assumption of proportionality of odds was not violated when tested using the Brant test, (Chi square = 13.2; p = 0.982).

The following baseline variables were missing: age (18), history of testing for HIV (11), distance (7), employment status (7), marital status (6), gender (1) and education level (1) totaling to 51 (4.5%) incomplete cases. The imputed model was not adversely different from the complete cases analysis (Table 4).