Study design and setting
The study design and data collection methods have already been described elsewhere [29]. A baseline survey was conducted between 15th July and 30th August, 2019 as part of a cluster randomized controlled trial of Community Health Workers (CHWs) interventions for reduction of blood pressure among adults aged 25-64 years in Kilombero and Ulanga districts. The study area is approximately 450 kilometers by road South-West of the Tanzania’s commercial capital of Dar es salaam. Kilombero and Ulanga districts cover a total surface area of 28,669 square kilometers. Based on the 2012 national census and the average annual population increase for Morogoro region [30], Kilombero and Ulanga districts were projected to have a total population of 659,810 by end of 2019. The predominant tribes in Kilombero and Ulanga districts are Wapogolo and Wandamba with the main economic activities being maize and rice farming as well as fishing activities on the Kilombero river.
The baseline survey was conducted in 12 randomly selected villages (6 from Kilombero district and another 6 from Ulanga district) in Morogoro region. Study villages were selected from a list of 38 villages which had active CHWs at the time of randomization. For the CHWs intervention study, three villages from each district were assigned to intervention arm and the remaining three to control arm. Conditional randomization was done to ensure equal number of villages are assigned to intervention and control in each district. Based on the local government registries, the number of households per village ranged from 1,587 to 5,929 and a total of 262 households were randomly selected from each of the 12 study villages.
Study participants and eligibility criteria
The study population included young and middle-aged adults 25-64 years, who resided in the selected villages and provided a written informed consent to participate. A resident was defined based on the Health and Demographic Surveillance System (HDSS) definition as an individual who has stayed in the selected household for at least 4 months continuously regardless of whether s/he had slept in that household a night before the interview [31]. A household was defined as group of people who served food from the same pot.
Sample size estimation and sampling procedures
The sample size was estimated according to the methods proposed in the WHO stepwise approach to chronic disease risk factors surveillance (STEPS) [32]. The sample size was calculated for 95% CI (z=1.96) on the basis of a 5% margin of error and an estimated prior population prevalence of hypertension of 25.9% from the national wide representative STEPS survey in Tanzania [33], a presumed design effect of 1.2 and an anticipated nonresponse rate of 10%. Participants were adults aged 25-64 years and categorized into four age-sex groups, resulting into 8 strata. The resulting minimum sample size for the study was 3,145.
We used multi-stage cluster sampling technique with villages as clusters. The villages were stratified by district, to obtain equal number of villages in both districts. Furthermore, a random sample of 262 households was drawn for all villages and at each selected household, one eligible respondent was selected for interview. The next birthday rule was used to select one eligible respondent from a household where the research assistants first noted the months of birth of all eligible individuals in a household in ascending order (January – December). The first person on the list was then selected for interview, and if the selected person was not available after two attempts, then the next person on the list was interviewed. Non-response was considered if an eligible participant refused/did not consent to participate or the selected household had only one eligible responded who could not be found after two follow up attempts.
Data collection procedures
Face-to-face questionnaire-based interviews were conducted by a team of trained research assistants with experience in conducting health and demographic surveys. Information collected during the interviews included socio-demographic and economic characteristics, knowledge of risk factors and warning signs for cardiovascular diseases, medical history, behavioral characteristics and physical measurements.
Socio-demographic and economic characteristics
Socio-demographic information included age, gender, marital status, education level, and occupation. Age was collected as a continuous variable and categorized into 25-34, 35-44, 45-54, and 55-64 years. Education level was measured by asking the highest education level attained as none, primary education, secondary education, college/university. Marital status at the time of data collection was grouped into three groups as never married, married or living together and divorced /separated /widowed. Occupation was assessed and categorized as farmer, housewife, employed (public/private), petty business and others. Economic status of the participants was assessed through ownership of household items such as radio, television, telephone, sofa, refrigerator, bicycle, car, and having working electricity; house ownership, construction materials (floor, walls and roofing materials); source of fuel for cooking and lighting; source of water supply for home use and drinking; and household sanitation facility [34]. All this information was used to generate household wealth index following descriptions in the Demographic Health Survey (DHS) toolkit [35].
Assessment of CVD knowledge
To assess general knowledge, participants were asked if they had ever heard or read about CVDs and their sources of information. Knowledge of risk factors and warning signs was then assessed using open-ended questions (participants were asked to mention risk factors and warning signs spontaneously) followed by closed-ended questions (participants were asked to identify risk factors and warning signs from a list of “yes/no” questions). The questionnaire comprised of 10-items about knowledge of risk factors and 9-items on knowledge of warning signs. Knowledge scores were assigned as 1 (one) for correct response and 0 (zero) for wrong response. Knowledge scores were calculated based on closed-ended questions.
The score points were then summed across to obtain a total score for knowledge of risk factors and warning signs separately. Total scores for knowledge of risk factors were categorized as; good (7-10 points), moderate (4-6 points), poor knowledge (1-3 points) and not knowledgeable at all (0 point). Total scores for knowledge of warning signs were categorized as; good (7-9 points), moderate (4-6 points) and poor knowledge (1-3 points) and not knowledgeable at all (0 point). Internal consistency and reliability of each set of items for assessment of knowledge of risk factors and warning signs was evaluated using Cronbach’s alpha test. The reliability coefficients were 0.826 for the scale of knowledge of risk factors and 0.782 for the scale of knowledge of warning signs. The two scales were reliable as Robinson and colleagues assert, an alpha coefficient above 0.80 is “exemplary”, and that in the range between 0.70 and 0.79 is “extensive” [36].
Total knowledge of CVDs was assessed by combining the total knowledge scores for risk factors and warning signs. The maximum possible total knowledge score was 19 points. Participants who obtained total knowledge score of ≥14 points were classified as having “good knowledge”, those with total knowledge score between 8-13 points were classified as having “moderate knowledge”, those with total knowledge score between 1-7 points were classified as having “poor knowledge”, and those who did not get any score as “not knowledgeable”.
Medical history and behavioral risk factors
Participants were asked if they had ever been diagnosed with hypertension or diabetes mellitus. Lifestyle related CVD risk factors including smoking, alcohol drinking, and dietary habits were assessed using a modified WHO STEPs data collection questionnaire which has been previously used in Tanzania [33]. Questions on smoking probed current and past smoking while questions for alcohol probed current and past drinking. Dietary assessment included frequency of consumption of fruits and vegetables (in days per week), and frequency of use of raw table salt (as never, sometimes or always).
Perceptions about CVD and practices towards CVD prevention
The study questionnaire also contained a section with set of questions that assessed participant’s self-perceived risk for cardiovascular disease, perception about current body weight, general practices, dietary practices, and behavioral practices in the past one year that relate to better cardiovascular health and prevention of CVDs.
Physical measurements
Blood pressure was measured using a digital blood pressure machine (OMRON HEM-712C, Omron Healthcare, Inc). Research assistants were trained to measured blood pressure on the left arm with participant in a seated position. The first reading was taken after at least 5 minutes of resting. The second and third readings were taken half-way and at the end of interview respectively. Participants who had elevated blood pressure had blood pressure measurements repeated on the following day to confirm their elevated blood pressure before being given referral letter to nearby health facility for proper diagnosis and management per standing national guidelines. For these individuals, we used their repeat blood pressure measurements for analysis. An average of three blood pressure measurements was used during analysis. Hypertension was defined as average systolic blood pressure (SBP) ≥140mmHg and/or diastolic blood pressure (DBP) ≥90mmHg and/or currently taking blood pressure lowering medications in accordance with the Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation and Treatment of High Blood Pressure [37]. Anthropometric measurements included weight and height. Body weight was measured to the nearest 0.1kg using a SECA 803 digital scale placed on flat ground with participant wearing light clothing and with no shoes. Height was measured to the nearest 0.1cm with participant in a standing position with heels perpendicular to the portable stadiometer. Body mass index (BMI) was calculated using measured weight and height (kg/m2).
Data analysis
Data were entered and analysed using IBM-Statistical Package for Social Sciences (SPSS Inc., Chicago, IL, USA) version 20 software for Windows. Descriptive statistics were used to summarize continuous variables as means with standard deviations and categorical variables as frequencies with proportions. Comparisons between groups were done using Chi-square test and independent samples t-test (or ANOVA) for categorical and continuous variables respectively.
To perform logistic regression, knowledge of CVD risk factors and warning signs was first dichotomized by combining good knowledge and moderate knowledge as having “adequate knowledge” and those with poor and not knowledgeable as having “inadequate knowledge”. Bivariate logistic regression was then performed to determine the relationship of each independent variable with CVD knowledge score. Predictor variables were categorized as: sex, age, wealth status, marital status, education level, place of residence, occupation, current smoker, current alcohol drinker, use of raw table salt, fruits consumption per week, vegetable consumption and body mass index
All variables with p ≤0.2 in the bivariate analysis were included in the multiple logistic regression analysis to determine factors that are independently associated with overall adequate CVD knowledge. Before fitting multiple logistic regression model, we first checked for correlation and strength of correlation between variables in the regression model to avoid problem with multicollinearity. We used variance inflation factor (VIF) with a cut-off value of ≤5 to rule out multicollinearity. Overall fitness of the model was assessed using the Pearson Chi-square test and Hosmer–Lemeshow goodness-of-fit test. Adjusted odds ratio (AOR) with their corresponding 95% confidence intervals (95% CI) are presented as measures of association. Statistical significance was accepted based on two-sided p-value ≤0.05.