A retrospective case-control study, carried out by the Rwanda Biomedical Centre-TB&ORD division through its Centres for Diagnosis and Treatment (CDTs) in Rwanda. This study is a secondary data analysis using a data base abstracted a data base of a study carried out from June to December 2017among participants aged 15 or over.
Study population and inclusion criteria
Cases were defined as patients aged 15 or over who had a positive TB test (confirmation of diagnosis was done through a sputum microscopy test or a GeneXpert test in the past six months). Controls were defined as any person aged 15 or over who attended the same clinics as the case during the same period who was presumptive TB patient (with clinical signs) with no TB diagnosis made by a clinician. The study team located cases from TB registers, contacted them via phone calls, invited and appointed them for interview. A take all approach has been used for the recruitment of cases. In case of refusal to participate, the case was documented and replaced by another eligible participant. For the control group, a random selection was applied to replace any participant who would refuse to participate. The patients with impaired decision-making capacity due to mental health conditions, children under 15 years of age and prisoners were not considered for this study. A Structured questionnaire translated from English to Kinyarwanda was used for interview. For both cases and controls, face-to-face interviews were conducted by trained healthcare providers.
The study enrolled participants from CDTs/Health Centres in Rwanda. Thirty-nine CDTs in Health centres were included in the study. Referral Hospitals and prisons were not included in the study because they don’t serve a homogenous population.
Sample size calculation
The minimum sample size was calculated basing on simulation approach and controls (free of TB), we considered a matching of controls with cases for sex and age group. We used a simulation approach to identify the adequate sample size with appropriate power (the probability of finding an effect if it is real). The probability of finding an effect if it is real (1-β) was set to 80%; we set a series of 1.2 to 2.5 to detect the Odds Ratio. Each case was given 3 controls (Design 1:3) and we set a series of proportion from 0.01 to 0.9 to calculate the proportion of exposed in the control group (λ). The minimum sample size required from the above estimates was 503 cases. From this sample size, we added 20% for non-response rate; therefore, the sample size of TB cases was 629 cases. We then recruited 3 matched controls for each case recruited totalizing 1887 controls and making a total of 2516 participants.
In order to get at least the power of 80%, we assumed Odds Ratio of 1.5 and proportion of exposed in the control group (λ) to be 0.5. The sample distribution per province and HF was calculated basing on the probability proportional to size (PPS) sampling. A two stages calculation of the weight was applied. In the first stage, the weights has been chosen to compensate for the effect that health facilities do not have the same size where is the sampling probability of each cluster, while for the second stage, weights have been used to compensate for the effect that target groups of participants do not have the same number of individuals.
Data collection method
Study team members received training on data collection tool, ethics and protection of human subjects. Data collectors were also trained to ascertain and document participant eligibility, provide eligible patients with study information, obtain signed informed consents, conduct participant interviews, and provide collected data to the data manager for back up. Prior to the interviews, data collectors obtained signed consent forms from cases and control groups. Interviews were conducted in a secured private place. A structured questionnaire was completed matching one case to three controls. Once all expected cases responded to the questionnaires and matched controls interviewed at the selected sites, the sampling process was ended.
Data management and analysis
For quality assurance purposes, Access programme-based questionnaires have been used. Every participant had a unique study identification code. Signed informed consents were obtained from participants and stored in a lock and key protected location in RBC under supervision of the Principle Investigator (PI) of the study. A dedicated data manager was responsible for data entry and cleaning processes. Data analysis was performed in MS Excel and Stata version 13.1 (Stata Statistical Software: Release 13. College Station, TX: StataCorp LP). For descriptive statistics, variables were tabulated and subsequent proportions analysed by age, sex, frequencies, percentage, medians and inter-quartile ranges (IQR). In multivariable logistic regression, the dependent variable was tobacco smoking. Socio-demographic and other variables were dependent variables. Differences were considered statistically significant with p < 0.05.
The informed consent process was completed prior to conducting the interviews. The information sheet was provided and read to participant and signed consent form obtained following voluntarily agreement of participant. Participants were free to withdraw from the study at any moment without any prejudice. All interviews were conducted in private locations to ensure confidentiality. Data collectors were trained on data collection tools, ethical considerations, privacy and confidentiality. The approval of the study and study instruments were obtained from the National Health Research Committee and the Rwanda National Ethics Committee.