Fistula screening tool
The OF screening questionnaire, which included 15 test questions on LUTF and LGTF symptoms and 12 questions on other pelvic floor symptoms and basic reproductive history, was initially developed by the late Thomas Elkins (Professor, Johns Hopkins Department of Gynecology and Obstetrics) (Additional file 1: Appendix S1). The questionnaire also included urinary incontinence (UI) and fecal incontinence (FI) screening questions that have been validated in many populations [12, 13]. Prior to piloting the questionnaire in Nepal, face validation was performed and the questionnaire was modified accordingly with consultation from 8 expert fistula surgeons in the USA, Ethiopia, UK, and Nepal. To translate and culturally adapt the OF screening questionnaire for Rwanda, the questionnaire was first translated into two versions by Rwandan fistula surgeons and bilingual language experts. All translated versions were back-translated and then reviewed by local health workers and compiled to ensure that the final translated and culturally-adapted questionnaire reflected our original questionnaire and would be understandable to study participants. The translated questionnaire was pilot-tested on focus groups of women with clinically confirmed OF and women without OF, and further modified as appropriate.
We also included a question frequently used in DHSs in South Asia and Sub-Saharan Africa, including Rwanda, which addressed both LUTF and LGTF symptoms but had not been validated (Question 21, Table 2). The screening questionnaire also included questions for the interviewer on observations of wetness/ soiling and smells of urine/ feces on and around the study participants. Participants were verbally administered forms on demographics, medical and reproductive histories, as well as fistula histories and impact of fistula and UI symptoms on quality of life when appropriate.
Institutional review board approval was obtained from both Johns Hopkins Medicine (Baltimore, MD, USA) and the National Research Health Committee and the Rwanda National Ethics Committee (Kigali, Rwanda).
Study design and sites
We designed and powered this case–control study to primarily evaluate the discriminative ability of questions on LUTF symptoms, as women with UI, which is far more prevalent than OF, may confuse these symptoms with LUTF symptoms. This is consistent with what has been reported in the literature and in our previous study in Nepal [11]. In that study, besides the 2 women with confirmed LUTFs, there were 65 false positive women who reported LUTF symptoms which were determined to be symptoms of UI. This mirrors our past clinical experiences as well as those of our international panel of content experts consulted on screening questionnaire development. Additionally, most large epidemiologic studies on OF prevalence or case series on OF have found less than 1% of all women with OF have an isolated LGTF (6). Our sample size calculation suggests that we need at least 58 LUTF cases to validate the screening questionnaire at 90% sensitivity and specificity with 10% margin of error at 0.05% error level at 90% power. We included two types of controls in a ratio of 1:2:2 (case:control:control): the first type of control was women who did not have a LUTF but did have symptoms of UI (Urinary Incontinence Controls, UC); the second type of control was women who did not have a LUTF and did not have symptoms of UI (Normal Control, NC).
To a priori increase the likelihood of recruiting the needed number of LUTF cases, we conducted the case–control study at Kibagabaga district hospital, in Kigali, Rwanda; parous women with suspected OF symptoms are sent to this hospital by local health centres and health posts for evaluation and care triannually in a government sponsored program with support from a U.S. non-governmental agency (International Organization for Women and Development). After obtaining informed consent, interested women were screened for OF symptoms as well as UI and FI symptoms with our questionnaire (Fig. 1A). All study questionnaires, including the screening questionnaire, were verbally administered by preclinical medical students. Study participants then underwent clinical examinations by urogynecologists who were blinded to the screening questionnaire results. All women found to have an OF on clinical examination were classified as cases. Women found not to have an OF on examination were classified as UC or NC depending on if they had UI symptoms. Besides these women, to recruit the needed number of UC and NC, we also recruited parous women with and without UI symptoms at Kibagabaga and the other four provincial hospitals of Rwanda (North, West, East, South) in approximate proportion to the number of OF patients from each province (Fig. 1A, B). These women were screened for OF, UI, and FI symptoms with our questionnaire, and then examined by urogynecologists blinded to the questionnaire findings. All women received treatment according to their diagnoses.
Clinical examination
All consenting participants underwent a clinical examination by board certified urogynecologists with experience in fistula care, who were blinded to the screening questionnaire results. To safeguard against missing OF cases, additional clinical tests were performed as needed, and confirmation of no OF was made by at least two experienced fistula surgeons. Radiologic studies such as intravenous pyelogram and/or barium enema studies were performed on a case-by-case basis. At the end of each day, we reviewed the screening questionnaire results and performed the aforementioned procedures, if not already performed, on any participants who screened positive for OF symptoms but were not found to have a clear fistula tract on initial examination.
Analysis
Means (standard deviations), medians (ranges) were calculated as appropriate for continuous variables, and frequencies for categorical variables. Chi-square statistic for categorical variables and t-statistic for continuous variables were used to test for heterogeneity in basic baseline demographics and medical and reproductive histories for women with and without OF. The frequencies of the women’s self-reported OF symptoms from the screening questionnaire were compared to the clinical examination diagnosis. The sensitivity and specificity of the questions were calculated, along with their respective 95% confidence intervals [16]. The OF screening questions were compared to determine which questions, or combinations of questions, most accurately identified OF. Receiver Operating Characteristic (ROC) curves were plotted and the area under the curve (AUC) was also calculated for the questions [17, 18]. The data were analyzed in R version 3.3.1 [19].
To more accurately determine OF prevalence estimates, we applied the test characteristics we determined for the aforementioned DHS question to the lifetime prevalences of OF reported in DHS from 18 sub-Saharan countries, including Rwanda. We estimated the unobserved true prevalences utilizing:
$$P=\frac{(p + Sp - 1)}{(Se + Sp -1)}$$
P is the unobserved true prevalence, p the observed prevalence, Sp the question specificity, and Se the question sensitivity. This question had been asked among females aged 15–49 in 23 DHSs from 18 countries from 2005 to 2018; we used the data from the most recent surveys [14]. We excluded surveys that used a different sub-sample than females aged 15–49; however, as Rwanda is our country of interest, we did include the 2005 Rwanda DHS, which only asked OF questions to women who had given birth in the previous five years [15]. Our study is reported in accordance with Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) recommendations. [20].