Strengths and limitations
The major strength of our study is the community-based identification of the suspected cases (with reported symptoms) combined with hospital-based clinical examination, which maximised the likelihood of accurate case identification. The population of recently pregnant women was identified from a well-identified cohort of pregnant women and was followed over a defined period of time, minimising the risk of selection bias. Detection bias was minimized by the use of a standardized data collection instrument (fistula module) together with clinical examination.
Although nondisclosure of incontinence symptoms in the questionnaire interview was possible, it seems unlikely, given that the data collectors were trained to explain the meaning of urinary incontinence and to explain that treatment would be offered free of charge to those detected and to emphasise that all the information collected would be kept confidential. A risk of underestimating the number of fistula cases should be considered, as two of the 30 mothers who reported incontinence symptoms did not attend for subsequent examination. The ostracism and stigma associated with the condition [
37] might have deterred the interviewed mothers from disclosing their symptoms due to the risk of being excluded from their households and/or communities. A further limitation was that we had no information about the pre- or post-operative circumstances.
Interpretation
The fistula incidence of 1.1 per 1000 recently pregnant women that we found is higher than Adler’s [
11] pooled incidence estimates in LMICs of 0.09 and 0.66 per 1000 recently pregnant women in community-based and hospital-based studies, respectively. It is recognized in the literature that reported rates of obstetric fistula vary widely. Taking into account that our survey was followed by gold-standard gynaecological exams, the variation could represent true differences in incidence. We emphasize the importance of the clinical examination and confirmation of all women who reported fistula symptoms: 78% of the mothers who reported fistula-like symptoms had conditions other than fistula. Similar to our study, a community-based screening for fistulae in Nigeria, using the fistula module questionnaire, found that 53% of the women who reported symptoms did not have a fistula [
38].
Our estimate is close to the fistula incidence of 1.2 per 1000 births found in the only prospective population-based study, which was undertaken in rural West Africa in 1999 [
12]. That incidence was, similar to our study, observed in rural areas where women are at higher risk of labour complications. However, the study from West Africa was undertaken in an area with a much lower proportion of births attended by skilled health personnel (39.6% vs 85%) and lower caesarean section rates (0.7% vs 3.8%). Although the low caesarean section rate could suggests an unmet need [
39,
40], it is important to consider that most of the identified cases in this study occurred amongst women who arrived in a timely manner at the primary health facility (except one who took 13 h). Thus, while several studies have described delays in deciding to seek care (first delay) and accessing a health facility (second delays) as being important factors in fistula formation [
41‐
43], we found most delays occurring at health facilities (third delay). Indeed, the detection of high-risk deliveries and decision to refer women to a hospital with surgery capability was not timely. In addition, delays in reaching the referral hospital, mostly due to unavailability of ambulances, have been observed, similar to those described by Waiswa et al. (2017) [
44] from Southern Tanzania, where there are no ambulances at the primary facility level, which is why emergency transport is a constraint. These findings that the third delay (receiving inadequate care at health facility) contributed to obstetric problems are corroborated by other studies conducted in Tanzania, Gambia and elsewhere [
45,
46]. We have classified the failure to detect obstetric problems in a timely manner at the primary health facility as a third delay, according to the model described by Thaddeus and Maine (1994) [
33] and Berhan and Berhan (2014) [
47]. However, as discussed by Gabrysh and Campbell (2009) [
48], these delays can also be classified as second emergency delays, when complications that cannot be managed at the primary facility require referral to a higher-level facility. Whatever the classification, we consider that the issue is the capacity of the health system in caring for the women in labour.
Prolonged obstructed labour is the most common cause of fistula formation [
2,
49,
50]. Three of the five identified fistulae were probably ischaemic in origin (two type II fistula and one type I). However, in two cases, the type I fistulae were high vaginal, close to the cervix. This type of fistula (equivalent to type 1 on the Goh classification [
51], i.e., more than 3.5 cm from the external urinary meatus) is presumed to be iatrogenic in aetiology [
5]. This could be the case for the type I fistula, which occurred where the duration of labour was less than 24 h. A combination of ischaemic and iatrogenic mechanisms could be suggested for the other type I, where the duration of labour was 30 h. It is important to mention that, for a type I fistula to be considered as iatrogenic, it would either mean that the incision of the uterus was too extensive or that the sutures used for the closure of the uterus would have included the fundus or posterior wall of the bladder and would have to be extended to the vesico vaginal septum. Although the mechanism involved could not be verified in this study, there are reports of an increasing proportion of iatrogenic urogenital fistulae in LMICs, primarily in obstetric interventions, especially caesarean section [
5,
52], and it is possible that this was the case in our study. This calls for more quality of care research on obstructed labour and caesarean delivery.
All four foetuses delivered by caesarean section were stillborn. We do not have information about the circumstances of the management of these cases, but, considering that some of these could have been obstructed labour with foetal death, we should consider whether craniotomy may have been an alternative method of delivering these babies. This procedure is performed in LMICs [
53‐
55] and is especially useful in patients who come from rural areas and run the risk of the complications of an abdominal-route delivery and the risks of rupture of uterine scar during subsequent pregnancies or labour [
56,
57].
Although the fistulae were recognized at the health facilities, no treatment was proposed to these mothers at the time. This could be due to the limited availability of fistulae repair services and a lack of fistulae surgeons, which was reported in several settings in sub-Saharan Africa, including Mozambique, where at least 80% of women with fistula are estimated to have no access to fistulae repair each year [
58]. However, the lack of adequate postpartum assessment and referral of the women with fistulae should be also considered.
This is the first study in Mozambique estimating fistula incidence in rural areas through a community-based survey with suspected obstetric fistulae being examined and treated at a health facility. Our results could be an underestimate for the country’s epidemiology as a whole, considering that our study area has a high density of health facilities compared to other parts of the country where more first and second delays would be expected.